Reflecting Davidson: Donald Davidson Responding to an International Forum of Philosophers 9783110886504, 9783110131802

Table of contents :
Preface
Truth, Meaning and Logical Form
Donald Davidson (Berkeley): Reply to Wolfgang Künne
Davidson on Truth
Donald Davidson: Reply to Richard Schantz
Quine and Davidson on Reference and Evidence
Donald Davidson: Reply to Felix Mühlhölzer
Is Radical Interpretation Possible?
Donald Davidson: Reply to Jerry Fodor and Ernest Lepore
The Philosophical Significance of a Shared Language
Donald Davidson: Reply to Andreas Kemmerling
Norms and Meaning
Donald Davidson: Reply to Akeel Bilgrami
Uses of Mistakes
Donald Davidson: Reply to Joachim Schulte
“What Metaphors Mean” and how Metaphors Refer
Donald Davidson: Reply to Oliver Scholz
How Relational Are Davidson’s Beliefs?
Donald Davidson: Reply to Johannes Brandl
First-Person Authority and Radical Interpretation
Donald Davidson: Reply to Eva Picardi
The Explanation of First Person Authority
Donald Davidson: Reply to Bernhard Thöle
The Architecture and Evidential Base of the Unified Theory
Donald Davidson: Reply to Lorenz Lorenz-Meyer
Reasons, Actions, and their Relationship
Donald Davidson: Reply to Ralf Stoecker
The Explanatory Force of Action Explanations
Donald Davidson: Reply to Peter Lanz
Mental Concepts: Causal because Anomalous
Donald Davidson: Reply to Peter Bieri
Evaluative Judgements
Donald Davidson: Reply to Thomas Spitzley
Problems of Induction: Davidson and Goodman on Emeralds, Emeroses and Emerires
Donald Davidson: Reply to Rosemarie Rheinwald
Selected bibliography
Index

Citation preview

Grundlagen der Kommunikation und Kognition Foundations of Communication and Cognition Herausgeber/Editors Roland Posner, Georg Meggle

Reflecting Davidson Donald Davidson Responding to an International Forum of Philosophers

Edited by Ralf Stoecker

w DE

G

Walter de Gruyter · Berlin · New York 1993

® Gedruckt auf säurefreiem Papier, das die US-ANSI-Norm über Haltbarkeit erfüllt. Printed on acid-free paper which falls within the guidelines of the ANSI to ensure permanence and durability

Die Deutsche Bibliothek —

CIP-Einheitsaufnahme

Reflecting Davidson: Donald Davidson responding to an international forum of philosophers / ed. by Ralf Stoecker. — Berlin ; New York : de Gruyter, 1993 (Foundations of communication and cognition) ISBN 3-11-013180-3 NE: Stoecker, Ralf [Hrsg.]

Library of Congress Cataloging-in-Publication

Data

Reflecting Davidson: Donald Davidson responding to an international forum of philosophers / edited by Ralf Stoecker. p. cm. — (Grundlagen der Kommunikation und Kognition = Foundations of communication and cognition) Revised versions of lectures delivered in Feb. 1991 at the Center for Interdisciplinary Research, Bielefeld, Germany. Includes bibliographical references and index. ISBN 3-11-013180-3 (alk. paper) 1. Davidson, Donald, 1917—Congresses. I. Stoecker, Ralf. II. Series: Foundations of communication and cognition. B945.D384R44 1993 191 —dc20 93-37480 CIP

© Copyright 1993 by Walter de Gruyter & Co., D-10785 Berlin Dieses Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Printed in Germany Satz und Druck: Arthur Collignon GmbH, Berlin Buchbinderische Verarbeitung: Lüderitz & Bauer-GmbH, Berlin

Preface The idea for this book grew out of a conference devoted to the philosophy of Donald Davidson, which was organized by Peter Bieri and myself at the Center for Interdisciplinary Research (ZiF), Bielefeld, in February 1991. It was the aim of the conference (which took its name from Davidson's programmatic paper "Toward a Unified Theory of Meaning and Action") to provide an overall survey of Davidson's philosophical enterprise. The articles in this book are revised versions of the lectures held at this conference. During the conference Professor Davidson opened the discussion of each presentation with a detailed response. These responses were recorded and supplied the basis for his thoroughly revised replies that are added to the respective articles in this book. Many people have helped us in the course of the project. I want to thank Peter Bieri, Andreas Kemmerling, Wolfgang Künne and Peter Lanz for their cooperation in the organization of the workshop and them and all the other contributors for their willingness to include their papers in this volume. My special thanks go to Ernie Lepore for his advice and support in carrying out this project. Consulting him has always proved to be a safe method in order to solve any problem whatsoever. I also want to express my gratitude to the ZiF and the Department of Philosophy, Universität Bielefeld, which generously supported the project, and in particular to Marina Hoffmann, Karin Matzke and Gitta Schmidt for their excellent work in preparing the conference and transcribing the tapes. The book closes with a selected bibliography comprising Davidson's publications as well as some secondary literature. I want to thank Gabi Ruhmann very much for all the labor she invested in compiling, correcting and preparing the data. This book is dedicated to the work of an outstanding philosopher. But it would not have been possible if this philosopher had not turned out to be such a kind and supportive man. So, first and foremost I would like to thank Donald Davidson for his contribution to the conference and to the book and for all his help and encouragement during the last two and a half years. Bielefeld, November 1992

Ralf Stoecker

Contents Preface

V

(Hamburg) Truth, Meaning and Logical Form

1

WOLFGANG K U N N E

Donald Davidson (Berkeley):

Reply to Wolfgang Kiinne

21

(Berlin)

RICHARD SCHANTZ

Davidson on Truth

25

Donald Davidson: Reply to Richard Schant%

36

(Dresden) Quine and Davidson on Reference and Evidence

41

Donald Davidson: Reply to Felix Mühlhöl^er

54

(Rutgers, New York) Is Radical Interpretation Possible?

57

Donald Davidson:

Reply to Jerry Fodor and Ernest Lepore

77

(München) The Philosophical Significance of a Shared Language

85

FELIX M Ü H L H Ö L Z E R

J E R R Y FODOR / E R N E S T LEPORE

ANDREAS K E M M E R L I N G

Donald Davidson: Reply to Andreas Kemmerling

117

(New York) Norms and Meaning

121

Donald Davidson: Reply to Akeel Bilgrami

145

AKEEL BILGRAMI

JOACHIM S C H U L T E

(Bologna)

Uses of Mistakes

149

Donald Davidson: Reply to Joachim Schulte

158

R. S C H O L Z (Berlin) "What Metaphors Mean" and how Metaphors Refer

161

Donald Davidson: Reply to Oliver Schol^

172

OLIVER

VIII

Contents

(Salzburg) How Relational Are Davidson's Beliefs?

175

Donald Davidson: Reply to Johannes Brandl

194

E V A P I C A R D I (Bologna) First-Person Authority and Radical Interpretation

197

Donald Davidson: Reply to Eva Picardi

210

(Berlin) The Explanation of First Person Authority

213

Donald Davidson: Reply to Bernhard Thöle

248

(Hamburg) The Architecture and Evidential Base of the Unified Theory

251

Donald Davidson: Reply to Loren^ Loren^-Meyer

263

(Bielefeld) Reasons, Actions, and their Relationship

265

Donald Davidson:

287

JOHANNES B R A N D L

BERNHARD THÖLE

LORENZ LORENZ-MEYER

R A L F STOECKER

Reply to Ralf Stoecker

(Bielefeld) The Explanatory Force of Action Explanations

291

Donald Davidson: Reply to Peter Lan£

302

(Berlin) Mental Concepts: Causal because Anomalous

305

Donald Davidson: Reply to Peter Bieri

311

(Duisburg) Evaluative Judgements

315

Donald Davidson: Reply to Thomas Spitzley

330

PETER L A N Z

PETER BIERI

T H O M A S SPITZLEY

Contents

IX

ROSEMARIE RHEINWALD (Münster) Problems of Induction: Davidson and Goodman on Emeralds, Emeroses and Emerires

333

Donald Davidson: Reply to Rosemarie Rbeinwald

347

Selected bibliography

349

Index

389

Truth, Meaning and Logical Form Reflections on Davidson's Philosophy of Language WOLFGANG KÜNNE

I. Introduction: Davidson and Tarski The governing idea of Professor Davidson's inquiries into truth and interpretation I take to be this: We would know "what it is for words to mean what they do" if we knew how to construct a theory of "radical interpretation" for a natural language. Think of such a theory as a fairly complex sentence, as something which can replace a sentence-letter in a schema. Then we can say: A theory of radical interpretation, formulated in English, for the native language of most German citizens, would make the following schema true when substituted for 'p': (R.I.)

If somebody knew that ρ he would be able to understand every German sentence, and one can confirm that ρ by evidence available to somebody who does not yet know German.

Davidson finds his model for the structure of such a theory in Tarski's work on truth, that is, in a recursive specification of the truth-conditions of all sentences of a language whose grammar is that of classical first order predicate logic. Tarski showed how, from a meagre set of axioms, there could be deduced a flood of theorems (T-sentences), having the form s is true *-> ρ where the sentence used on the right has the same meaning as the sentence mentioned on the left. Such bi-conditionals may sound trivial when the language of the theory contains the target language, but they are not trivially obtained, for their deduction shows how mastery of a finite set of linguistic elements and constructions can generate a limiteless grasp of conditions of truth for the sentences of a language. And it is this very feature of Tarski's achievement that makes it attractive for Davidson's entirely different project. While Tarski can appeal to the concept of meaning (in the guise of sameness of meaning) without detriment to his aim, Davidson cannot do so, since his project is to elucidate this concept.

2

Wolfgang Künne

In his 'Introduction' to 'Inquiries into Truth and Interpretation' (ITI) Davidson has put the difference this way: while Tarski intended to analyse the concept of truth by appealing (in Convention T) to the concept of meaning (in the guise of sameness of meaning, or translation), I have the reverse in mind. (ITI: xiv)

This way of putting the difference is misleading both for Tarski's and for Davidson's project. For one thing, it sounds as if Davidson was up to analysing the concept of meaning, and this is not a very happy description of his aim. And on the other hand, it is hardly an adequate characterisation of what is going on in Tarski's theory. More recently Davidson himself has said: Correspondence theories have always been conceived as providing an explanation or analysis of truth, and this a Tarski-style theory of truth certainly does not do. 1

And in his Dewey-Lectures he has rehearsed the arguments for this negative claim. 2 Here is my own version of these arguments. Let us take as our target language a slightly simplified version of German — call it " L I " — where something is a sentence if and only if it is either identical with "Schnee ist weiß" or with "Gras is grün". ('S' and 'G' are to abbreviate the quotation-names of these two sentences.) It is now very easy to construct what Tarski would call a definition of 'true in LI': (A)

χ is true in L1 «-• ((x — S, and snow is white) or (x = G, and grass is green)).

This exactly fixes the extension of the truth-predicate for LI, and it is in complete conformity with Tarski's standards of formal correctness and material adequacy. Now compare (A) with the following bi-conditional which exactly determines the extension of 'is a daughter of Laban': (a)

χ is a daughter of Laban

(x — Rahel or χ — Lea).

Now knowing that (a) is entirely useless if you want to know under what conditions the predicate 'is a daughter of King Lear' applies to someone. The predicate on the left-hand side of (a) is a relational predicate, a predicate one can quantify into. A statement fixing the extension of such a predicate explains its meaning only if it can be projected to variants of the predicate where the embedded singular term has been replaced by another one. This demand can easily be satisfied for our predicate in (a): 1

2

Donald Davidson: A Coherence Theory of Truth and Knowledge; Afterthoughts, 1987. In: 'Reading Rorty', ed. A. Malichowski, Blackwell, 1990, pp. 1 2 0 - 3 8 ; p. 135. Davidson: The Structure and Content of Truth. In: 'The Journal of Philosophy', 1990, 87, pp. 2 7 9 - 3 2 8 ; esp. pp. 2 8 5 - 9 5 .

3

Truth, Meaning and Logical Form

(a )

χ is a daughter of Laban (x is female & χ was begotten by Laban).

The bi-conditional (a') makes us see what Rahel and Cordelia have in common. — These complaints about (a) apply to (A) as well: we are not given a hint as to when predicates like 'true in L2', 'true in L3' etc. hold of a sentence. There is a further problem with both (A) and (a) when taken as explanations of the predicates on their left-hand side. According to (a) (b)

Rabel is a daughter of Laban

has the same truth-value as (c)

Rahel = Rahel or Rahel = Lea.

Now in order to know that (c) is true you only have to understand ' = ' and 'or' and to know that 'Rahel' and 'Lea' are singular terms. You do not have to inquire into early Jewish history. Now if (a) were an explanation of the meaning of 'is a daughter of Laban' then (b) would have to have the same epistemic status as (c), — which is absurd. — The same situation arises with respect to (A). According to (A) (B)

S is true in L1 2 , we find it pointless for η > 3 , and soon thereafter we shall find it impolite, ridiculous, inacceptable and, from some value of η on, impossible. — Examples like these raise interesting questions about the status of sentence-meanings. If we find it humanly impossible to grasp a difference in meaning between two given sentences but feel, as truth-theorists of meaning, obliged to acknowledge some such difference — what should we, then, conclude? Should we be sensualistic about these matters, and accept only differences in meaning which we can, as it were, feel? Or should we approach such phenomena like a scientific realist and accept the existence of "imperceivable" semantical differences in the same manner as we accept, without batting an eyelid, the existence of imperceivable quarks? Semantic sensualism clearly has a tendency to undermine the truth theoretical approach to the semantics of natural languages. Scientific realism applied to these issues has a reificational tendency which is certainly unattractive for a good Davidsonian. These are hard problems. But since Davidson ignores them, I shall, for the present purposes, keep up my attitude of naivete: the semantic distinction between phrase books and truth theories is unassailable.

108

Andreas Kemmerling

communication presupposes that speaker and hearer have (and recognize each other to be having) a capability of expressing beliefs which essentially have an indefinitely high degree of specificity. Our capability of having and expressing indefinitely specific propositional attitudes is of fundamental philosophical importance. Not only for the philosophy of language, but for many central areas in philosophy. In ancient days, human beings were characterized by philosophers as creatures which have "logos" or "ratio". Within the last hundred years or so, it has become increasingly manifest that one of the aspects that makes "logos" or "ratio" such a difficult topic is exactly this feature of arbitrarily high specificity of propositional attitudes. These issues are surrounded by fascinating and intriguing questions, for example "How is this capability of indefinite specificity acquired and applied?". I shall try to concentrate on a more modest topic, namely: Assuming a realist attitude towards the concept of belief, how can we make sense of the objectivity of ordinary belief-attributions? By an ordinary belief-attribution, I mean an ordinary case in which a belief-predicate like "believes that it is raining" or "believes that Harvey is a blockhead" is applied to somebody. And by the objectivity of a statement or judgement, I mean basically all that goes with its intersubjectively homogeneous cognitive evaluation, in particular all that is presupposed by the possibility that the (factual) statement or judgement which has been made by one person is definitively corrected by somebody else. I have called the question about the objectivity of belief-attributions a relatively modest topic because it avoids the more fundamental question whether such attributions are ever objectively true. In the following, I shall presuppose that the answer to that more fundamental question is: Yes, ordinary belief-attributions are sometimes objectively true. It has been argued, for example by Quine and Stich, that this is not so. 27 And though I will not try to answer their arguments, let me just point out that their subjectivist accounts of belief-attribution are based on the assumption that it makes no objective sense to speak of two people's sharing a language. This has to do, I assume, with the fact that both Quine and Stich have a tendency to identify what is objective with what can be captured in a naturalist description of the world. Moreover, both seem to take it for granted that belief-states (if there were such things) would have to be individuated individualistically or non-externally. In presupposing the correctness of an objectivist account of beliefattribution, I do deny the appropriateness of at least one of those standards of objectivity. In fact, I deny the naturalist standard. My attitude towards 27

Cf. Willard Van O. Quine, Word and Object, Cambridge, Mass. I960, especially 45; Stephen Stich, From Folk Psycholog) to Cognitive Science, Cambridge, Mass. 1983, Part I.

The Philosophical Significance of a Shared Language

109

individualism (or internalism) is more favourable but I won't go into that here. 28 Just take me to be denying any kind of (reductivist) naturalism, including internalist naturalism. XI. In giving a philosophical account of a concept it is of primary importance to account for simple uncontroversial cases, if there are any, to which the concept clearly applies. So in what follows, I shall restrict myself to simple attributions of belief-properties which are obviously true, if any such attributions are, and I shall inquire into the nature of their objectivity. I have no definition for what is an obvious truth. And I am not extremely confident that the following sufficient condition is very helpful, but at least it will give you an idea of what I am aiming at in the next sections. I want to call an attribution of a property F to an entity χ obviously true if there is a condition (or set of conditions) C, such that (1) something's satisfying C is evidence of the best possible kind for its being F, although (2) something's satisfying C does not entail that it is F, and (3) χ satisfies C. — Here is an example. It is obviously true that Harvey is not in my study, if Harvey satisfies the following condition: he is in my office which is more than two miles away from my study. (There may be possible worlds in which Harvey somehow manages to be in both these distant places at the same time, and therefore his presence in the office does not entail his absence from the study. But as things stand his presence in the office is still evidence of the best possible kind for his absence from the study.) XII. And here comes my claim: The only criteria we have at hand in order to establish the obvious truth of belief-attributions involve the idea of a shared language. It is this wherein I scent the philosophical significance of a shared language (or should I say: the philosophical significance of the idea of a shared language). Any belief-attribution which is obviously true presupposes the idea of a shared language. 29 28

29

I have gone into that, at deplorable length, in a paper entitled "Genau dieselbe Überzeugung" (Intentionalität und Verstehen, Forum für Philosophie Bad Homburg (ed.), Frankfurt 1990). I still think that individualism is feasible, when it comes to trying out a science of the human mind. But my provisos and qualifications have piled up in between. There are passages to be found in Davidson's work where he sounds receptive to this idea. In "True to the Facts" he remarks:

110

Andreas Kemmerling

Let me give you another example of an obvious truth: Jill believes that Harvey is a blockhead. Here is why this is obviously true. First, because Jill satisfies the condition of being a normal, adult, competent speaker of American English who has seriously, reflectively and truthfully uttered the sentence "Harvey is a blockhead", thereby referring to Harvey. Second, because Jill's satisfying this condition is evidence of the best possible kind of her believing that Harvey is a blockhead. And while we are at it, let me give you a third example for a candidate of an obvious truth. Daniela believes that Harvey is crazy. She satisfies the condition of being a normal, adult, competent speaker of Italian who has seriously, reflectively and truthfully uttered the sentence "Harvey e pazzo", thereby referring to Harvey. And Daniela's satisfying this condition is evidence of the best possible kind of the fact that she believes that Harvey is crazy. — Well, on second thoughts, is that really evidence of the best possible kind? I don't know enough Italian and American English, so I can't say. But I can offer another reason for taking it to be obviously true that Daniela believes that Harvey is crazy. First, Daniela is a normal, adult and competent speaker of some natural language. And — that's still the first point — if she were a normal, adult, competent speaker of our natural language (i. e., for the purposes of this paper, American English), she would answer our question "Is Harvey crazy?", in that very wording, with an unqualified affirmation, therein being serious, reflective and truthful. And her satisfying this counterfactual condition is also the right kind of evidence: best possible evidence for the claim that she believes Harvey to be crazy. Certainly it is not evidence which is easy to get. It may be a tough question whether Daniela (who is not a competent speaker of American English) in fact satisfies this counterfactual condition. But if she does, everything is as plain sailing as in Jill's case. Daniela's case invites for a general criterion of belief-attribution. "Somebody believes that so-and-so if the following condition holds: If he or she were a normal, adult, competent speaker of American English, he or she would —if asked (in American English) whether so-and-so— answer

What we can hope to make sense of, I think, is the idea of a sentence in another tongue being the translation of a sentence of English. ... we seem required to understand what somebody else would mean by a sentence of our language if he spoke our language. But difficult as this concept is, it is hard to see how communication can exist without it. (Inquiries into Truth and Interpretation, p. 53; bold italics mine) Decidedly, this could serve as the motto for the rest of my paper. Of course, I have boldly italicized the phrase "our language" in that quotation, whereas Davidson then hasn't and nowadays wouldn't.

The Philosophical Significance of a Shared Language

111

with an unqualified affirmation, therein being serious, reflective and truthful." This is a sufficient condition for ascribing beliefs to people. 30 There are plenty of cases where this criterion is not applicable. It is a general criterion, not a universal one. But we have nothing better in our sight and nothing as good at hand which could serve as a criterion for the obvious truth of belief-attributions. This criterion applies in the first place to actual speakers of our language, i. e. the language which is used in making the attribution. It applies by speculative extension to all other candidates (i. e. possible speakers of our language). For better or worse, it fails to apply to impossible speakers of our language, like e.g. computers, flies, dogs, lions, dolphins and maybe even some coffeehouse intellectuals. This is not to say that they don't have beliefs. It is just to leave open the question how the attribution of a specific belief to them could be warranted. Our criterion holds true under translation. A German version of it reads as follows: "Jemand glaubt, daß so-und-so, falls die folgende Bedingung erfüllt ist: Wäre sie oder er ein normaler, erwachsener, kompetenter Sprecher des Deutschen, dann würde sie oder er — vor die Frage gestellt, ob so-und-so — uneingeschränkt bejahend antworten und wäre dabei ernsthaft, bedacht und aufrichtig". As you will have noticed, in its German version, the criterion turns to competent speakers of German. In its Xese version, it'll concentrate on competence in Xese. The criterion is lingually egocentric, or — if you can tolerate a bit of lingo — it is lingocentric.

XIII. Is lingocentricity not dissonant with objectivity? It certainly is dissonant with objectivity if the language in question were characterized as something subjective. A lingocentric criterion of belief-attribution would be inac30

By the way, as long as we restrict ourselves to this criterion we need not be afraid of certain otherwise impending puzzles. Think of Kripke's Pierre as a normal, adult, competent speaker of American English who happens to be speaking French also. If we asked Pierre (in American English) whether London is ugly, he would say "Yes"; if we asked him (again in this language) whether London is pretty, he would in all desirable consistency say "No". If we asked him "Is London both ugly and pretty?", he would say "Of course not". No puzzle in the offing so far. And our criterion does not reach farther. If we think of Pierre as a normal, adult, competent speaker of French, we should switch to a French version of our criterion. Again, the results would not be puzzling. Our criterion is puzzle-proof in any version. So the question arises if this criterion solves, in a sense, the Kripkean puzzles or if it rather leaves something important out. I think that this is an interesting question and certainly don't want to try to answer it in a footnote.

112

Andreas Kemmerling

ceptable as a way of capturing the objectivity of belief-attributions if it involved reference to my language essentially. If this phrase, "my language", were to imply that I can do within the realm of my language whatever pleases me — and that I can do it without being responsible to any kind of criticism —, then bringing in my language could be of no help in objectifying belief-attributions. (Otherwise, I could rule, e.g., that in my language the sentences "It's raining" and "It's raining considerably" are synonymous. But I cannot rule such a thing for our language, or for English, or for any shared language.) Therefore, bringing in lingocentricity has to be construed as relativization to shared languages. And this makes it desirable to have an account of what it is, for two or more persons, to be competent in the same language. Professor Davidson does not believe in the philosophical significance of a shared language, so probably he disagrees with my lingocentric account of belief-attribution. For this account brings in the idea of a shared language centrally. And if this is — as I claim — the only account we have of making sense of the objectivity of obviously true attributions of indefinitely specific beliefs, then the idea of a shared language takes on the flavour of essentiality. And anything with that flavour smells of philosophical significance. XIV. Or is it just bad smell? Certainly to the nose of an adherent to internalist naturalism. Quine, Stich and others are right: there are no prospects for an internalist naturalist account of what it is, for a group of individuals, to be competent in the same language. Internalism leads to searching for the language spoken by an individual within his or her body-limits. The internalist has a tendency to think of a shared language as something miraculous. How wondrous that two semantic painters should depict the infinite universe of meanings in exactly the same way! Their perspectives and experiences must be so very different! Naturalism leads to neglecting or playing down the whole dimension of normativity. The naturalist has a tendency to think of a shared language as an invention of the theorist. After all, there are only finitely many patterns of firing neurons, and such patterns are —in the last analysis — where all cognitive action is. Hence a phrase book account should do, in principle. At least "if we are limning the true and ultimate structure of reality" 31 , there is no good answer to the question "what are languages, and when do they count as identical or distinct". 32 31 32

Willard V. O. Quine, Word and Object, Cambridge, Mass. 1960, p. 221. Ibid., p. 214.

The Philosophical Significance of a Shared Language

113

What's normativity got to do with all that? It's a complicated story. In my closing remarks I shall just hint at a few highlights of the plot. But before I turn to that, let me briefly take stock. Contrary to what Professor Davidson holds, I take it to be a problem of great philosophical importance to give an account of what it is for two people to share a language. The concept of a shared language comes in essentially as soon as we want to account for the objectivity of obviously true attributions of arbitrarily specific beliefs. Expressing such beliefs and recognizing such beliefs as being expressed is a hallmark of linguistic communication. So the analytical order of my argument, meant to persuade the Davidsonian, up to this point is this. [1] We want to give a philosophical account of successful linguistic communication. [2] It is distinctive of linguistic communication, successful or not, that belief-attributions of indefinite specificity are involved. [3] Such belief-attributions are, at least sometimes, objectively true or objectively false. Objectivity befits in particular those obviously true belief-attributions which are based on evidence concerning language-use. (In such cases, the specific belief is determined by way of determining exactly what has been expressed, in making a linguistic utterance, as being believed.) Hence [4] we want a philosophical account of the objectivity of such belief attributions. [5] A naturalist individualistic view does not yield such an account. [6] The only starting point in sight for an account of such belief attributions refers us to (the concept of) a shared language. Hence [7] the concept of a shared language is of philosophical significance (provided that an account of successful linguistic communication is of such significance). XV. Given that a naturalist individualistic position fails for our purposes, how could we hope to give a credible account of a shared language? It is at this point that, with no aspiration towards originality, I want to bring in normativity as the theoretical saviour. The concept of normativity I want to bring in, at this point, is a surprise packet. It contains glamorous items like rights and duties, commitments to (and impositions of) truth conditions, acceptations of corrections, and intersubjective agreement. On closer inspection, it may turn out to be Pandora's box. But let's give it a try. And in doing so, let's leave aside all troublesome features of languages like ambiguity, indexicality and what have you. Forget even about troublesome features of human communicators such as confusion, rashness, thoughtlessness, playfulness, unseriousness etc. Leaving all this out for a

114

Andreas Kemmerling

moment is no cheating. It just helps us to concentrate on the basic problem, namely: giving an account of what it is for two speakers to share one language. To make things easy, we'll deal with boringly virtuous speakers and boringly poor languages. We are approaching the holy places of language use: the sanctuaries of homophonous interpretation. XVI. Any account of what a shared language is must involve some theoretical device which delivers assignment of truth conditions to an infinity of sentences, regardless of who of the language-sharers uses any of these sentences. This theoretical device should be insensitive to the psychological peculiarities of individual speakers and just capture their semantic joint possession. One such device is the concept of a speech community. Here is an off hand definition: Two speakers belong to the same (homogeneous) speech community, if their utterances are treated (and are to be treated) equally with regard to resulting rights and duties. That is to say, in a speech community, speakers are treated semantically without respect of person. Pure semantics have to do with what you achieve purely linguistically, as just another speaker of a given language. Therefore, by uttering the sentence "It is raining", all speakers of the same (homogeneous) speech community commit themselves to be expressing exactly the same truth condition. And here comes the important tag: even if they have different ideas of what rain is.33 Insofar as truth conditions of utterances capture functional roles of their utterers in the network of social rights and duties, attributing truth conditions has little and only indirectly to do with individualist psychology. XVII. Think of betting. A few days ago, I bet Harvey five bucks that it was raining. Harvey's very words were "It's not raining", I bet against him. We went out to check the weather. It was raining. But Harvey said, "That's no rain, dude. That's just light drizzle. Gimme the money." Harvey would never lie to me. So by his utterance he really meant to be accepting a bet the winning conditions of which were satisfied by the actual weather. But he had accepted a bet, the exact wording of which was "It's not raining". 33

That's certainly Frege's idea about these matters. Probably, it's the Tractatus-Wittgenstein's idea. It's pretty obvious that even the late Wittgenstein cherished this idea (if you do not insist on /rw/A-conditions).

The Philosophical Significance of a Shared Language

115

He was committed to winning conditions different from those he thought he was committed to. It was raining. Losing conditions of Harvey's bet obtained. According to the relevant truth theory, Harvey had bet that it was not raining. So he lost. He had to pay me five bucks. He payed. Why was Harvey committed to those winning conditions? Because he stuck around for a sufficiently long time with the relevant bunch of people to have achieved the status of a competent bettor. They imposed their winning conditions on his English sounding bets with them. He accepted that. They all did that to each other. They still do it. That's what makes them having a shared communication system for betting. The essential point is not that a communication system for betting be shared between any two bettors. Such a system would make betting things particularly easy. But the kind of objectivity which is essential for betting can be established without such a shared system of verbal or gestural signs. What counts is that the two bettors commit themselves to the same winning/losing-conditions, by whatever signs this identical commitment is achieved. (The one bettor may have used the sound "hutchee", the other may have used "hutchoo" when they betted. With a little help from their respective friends they may be committed to accept any case of snee2ing as a relevant betting conditions.)

XVIII. Among other things a language is a system which allows its speakers to have arbitrarily specific bets. Many of the assertive uses of linguistic means can be profitably viewed in close analogy to bets. (Something certainly needs to be said about what the stake is, and in which sense the hearer of an assertion resembles somebody who takes up a bet. But let that go.) The whole institution of betting presupposes the concept of two people's being committed to the same winning conditions by the same words. It is of the essence for a bet to take place that the two bettors share a commitment to the same conditions even if they have different ideas of what these conditions are. But who fixes those conditions they are committed to? If no quarrel comes up about who has won, everything goes smoothly. In case of conflict, an "impartial and well-informed spectator" 34 , a "judge" 3 5 or a "group" 3 6 comes in handy. Of course, he or she or they may be hard to 34 35 36

A d a m Smith, The Theory of Moral Sentiments, L o n d o n 1 7 5 9 , section 2, chapter 1. J.L. Austin, How to Do Things with Words, O x f o r d 1 9 7 5 (second edition), p. 122. Eike v o n Savigny, The Social Foundations of Meaning, Berlin / Heidelberg / New Y o r k 1988, chapter 2.

116

Andreas Kemmerling

find. But it is even harder to have a bet without accepting the idea that a possibly ensuing quarrel could be settled in this manner. Even if the analogy between betting and linguistic communication may be a bit strained, I think it points to an important aspect. Let me try to bring out this point by adding another twist. Consider the question: Where exactly does a shared language come into the picture? As I said before, I think that Professor Davidson is fully right in claiming that linguistic communication can be successful even if speaker and interpreter do not share a language. The idea of a shared language lurks in the background of linguistic communication. It concerns the interpreter's linguistic competence of his language, i. e. the language in which his interpretation is expressed. It is this language which the interpreter has to share with other people (not necessarily with his interpretandus). It must be possible for him to make mistakes in his interpretative attempt. It must be possible that such mistakes could be corrected by others. But such corrections could only be made by somebody who is competent in the interpreter's language. (It would be irrational or unreasonable to accept criticism which is assumed to be incompetent.) In this sense, linguistic communication presupposes the idea of a shared language. A shared language is needed, not so much as a condition of interpretative success but rather as a precondition of the objectivity to which we aspire in interpretation. 37

37

Thanks to the members of the seminar on problems in analytic philosophy at the university of Munich, and to Donald Davidson, Hanjo Glock, Brian Loar, and Bernhard Thöle for comments on earlier versions of this paper. Special thanks to Katia Saporiti and Stephen Schiffer. Sebastian Edwards, Margot Magowan and Mark Helme were kind enough to check and improve my English.

Donald Davidson: Reply to Andreas Kemmerling A thought which determines much of what I say about linguistic communication is this: when we talk of what words mean, what names name, what it is for an utterance to be true, we must start with cases where communication succeeds, by which I mean occasions on which an audience interprets a speaker's words as the speaker intends and expects those words to be interpreted. Unless there are such occasions, talk of meaning is empty. How we use the notion of meaning in cases where a speaker's intentions are for one reason or another frustrated seems to me relatively unimportant as long as we keep cases distinct. This is the thought which leads me to find the idea of a shared language, and the dependent concepts of literal meaning and convention, unimportant to understanding what grounds linguistic meaning. Kemmerling, like some of my other critics, enjoys quoting my words "There is no such thing as a language" while ignoring my careful explanation of how I am using the word "language" in this context. Of course I don't deny that there are languages like English, German and Basque or that, in some ordinary sense, people speak them. What is clear to almost everyone, however, is that no two speakers speak in exactly the same way; each speaker has, as Chomsky has maintained, a personal ideolect. In spite of the differences, both in practice and in implicit "theories" about what expressions mean, people who speak differently often understand each other. If it were not for this, we would have no way of defining the concept of language we appeal to when we call Russian or Amharic a language. Kemmerling clearly thinks the order of explanation goes the other way; he wants to treat personal idiolects as deviating from a norm or set of rules or conventions, where it is the norm which we treat as basic. He scoffs at my suggestion that perhaps the rules are set by the standards of "schoolma'ams", but he offers no alternative. What does determine "conventional" meaning according to Kemmerling? Unless meanings are laid up in heaven, they must depend on the actual practice of someone or some group. If we want to be serious about the idea of a standard or conventional meaning we must decide what determines these meanings. I don't see the philosophical interest in making such a decision. Kemmerling complains that my concept of "first meaning" is nothing like the usual notion of literal meaning, and so cannot substitute for it.

118

Donald Davidson

Kemmerling identifies literal meaning with standard or conventional meaning, and so for him literal meaning depends on the unstated choice of a reference group of speakers; first meaning is tied directly to a single speaker (or to be more precise, to communicative occasions), and so deliberately diverges from what Kemmerling calls literal meaning. I "substitute" first meaning for literal (in Kemmerling's sense of "substitute" and "literal") only in that I base my theory of meaning and communication on the former and not on the latter. (I confess that having explained what I meant, I have sometimes allowed myself to substitute the phrase "literal meaning" for "first meaning".) Kemmerling thinks that when we engage in a linguistic exchange we enter into some sort of implicit pact to speak the same language. Why should we do this? To enter into a pact we would have to understand each other first, and if we understand each other, what's the point of the pact? Kemmerling's myth of two speakers picking up the same book —the book that contains the "truth-theory" they are to agree upon — reveals the question-begging assumption; for how do we or they know the speaker and interpreter will interpret the book in the same way? (This is an old point of Wittgenstein's.) I think the idea, which I have also come across in writings of Michael Dummett and Tyler Bürge, that as speakers we have an obligation to the language, or the community, or our audience, to speak according to some standard, is simply wrong. Such obligations may sometimes exist, but they are irrelevant to communication. (I owe it to my school, or my parents, to talk in a way that will not embarrass them; I owe it to my children to give them an example that will give them some desired status.) Such obligations are irrelevant to communication because if I engage in a linguistic exchange I necessarily intend to speak in a way that will be understood. If I think this requires that I speak as others do, I will do so without any pact or obligation. If I think that by speaking in a way that differs from the way those unspecified "others" speak I will be better understood, I'll deviate accordingly. Unless, of course, I am more eager to attain some social status than to be understood. Kemmerling contends that we would be unable to make objective statements about beliefs if we did not speak in the same way as others, that is, mean the same things by the same words. I am not sure I follow his argument, but the conclusion, if I have it right, is to my mind false. Though he takes a detour through a question about belief ascriptions, the point of which is lost on me, his idea seems to be Kripke's, or the view Kripke attributes to Wittgenstein, that there would be no check on the correctness of the use of an expression if there were not others who used the same expression in the same way. I can think of no reason to believe this. There is no reason to think it is easier for me to determine that your

Reply to Andreas Kemmerling

119

use of an expression is the same as my use of that expression than it is for me to determine that your use of some expression is the same as my use of a different expression. When I come to realize that you use some expression in the same way I use another expression, the question whether one of us is misapplying an expression is open to the same sort of check as when we figure out that we are using the same expression in the same way. I am with Kemmerling in holding that interpersonal communication is the basis of our concept of truth and objectivity; but the agreement this requires can be established whether or not we use the same words to express the same thoughts. Finally, there is the question about bets. There are areas of language use, of which bets are perhaps one and the law another, where we sometimes don't let somebody off the hook just because they didn't understand how their words could be taken or would literally be taken by a judge, a policeman, or somebody with whom they are making a bet. These are legalistic uses of language, and here we sometimes do hold people to standards which are only indirectly related to communication and the roots of language, but have a lot to do with enforcing bets or enforcing the law. However, if I were convinced that in apparently agreeing to a bet someone had understood certain words of mine differently than I thought he understood them, I wouldn't make him pay — unless he was my enemy. There is a similar issue in the law: if you get someone to sign a contract, for example, and it contains phrases that he has no way of understanding, the law sometimes holds that he has not committed himself to what those who have a different understanding of the words would have committed themselves by using those same words. These are interesting matters and they do involve rights and duties. But they do not prove that in general we have an obligation to mean by our words what others mean by those same words. A final remark: I would distinguish between the claim that linguistic communication depends on sharing a communal language and the claim that such sharing depends on "conventions". Kemmerling says nothing here about convention except that he thinks David Lewis has it right. I have questioned, not Lewis's analysis of convention, but whether that analysis applies to language. If the purpose of using language is to communicate, there is no need of a convention to make you talk in a way you think will be interpreted as you intend. To put this the other way around: if you are pretty sure that somebody is going to interpret you in a "non-standard" way, then you are foolish if you don't speak in a nonstandard way.

Norms and Meaning AKEEL BILGRAMI

Davidson's philosophy of language and mind, among other things, has taught us to think about the nature of language without a central place being given to the idea of convention, but with a central place nevertheless to be given to the idea of communication. Instructed by this lesson, I want to look at Saul Kripke's recent commentary on Wittgenstein with a view to making a negative point against the recent focus on normativity in the study of meaning. 1

I. The Apparent Relevance of Norms Let me first walk you very briskly through two tediously familiar examples and a common philosophical conclusion that they have suggested to those who've devised them. Here's the first example. A man, call him Bert, living in one of our English-speaking societies, from time to time sincerely and literally says "I have arthritis in my thigh." Intuition is supposed to tell us that he has made a certain kind of mistake. He has misapplied his concept, misused his term 'arthritis' (I shall use the words 'term' and 'concept' more or less interchangeably, as equivalent counterparts in the study of language and mind, respectively). Now imagine another man who is not different from Bert in any physiological, phenomenological or functional respect but who lives in another society where, the same phonetic vocable and the same inscription, 'arthritis', unlike as in one of our English-speaking societies, is applied to a wider class of ailments, affliciting joints and ligaments. This man says the same thing, Ί have arthritis in my thigh'. Intuition is supposed to tell us that this man makes no mistake at all. He has correctly applied his concept, which is not the concept of arthritis at all, because arthritis only afflicts joints. It is another concept deserving of another name; which might as well be 'tharthritis.' ' See Davidson (1984) and (1986), as well as Kripke (1982) and Wittgenstein (1953).

122

Akeel Bilgrami

What is the conclusion we are supposed to have drawn from all this? It is that an individual's concepts are constituted by the linguistic practices of his society and that is what permits us to say that he is applying his words correctly or applying them incorrectly. If we saw Bert, our Englishspeaker, as having his own idiosyncratic concept we could not see him as making a mistake. In fact we would never be able to detect mistakes of this kind if we did so. Tyler Bürge calls a philosopher who sees Bert this way, an individualist. An individualist attributes concepts to an individual simply on the basis of regularities in the individual's behaviour; in the case of Bert, on the basis of the conditions under which Bert regularly uses the word 'arthritis'. Bert, as we know, applies the word not only to conditions in joints but also to conditions in his thighs. Thus individualism attributes to him his own idiosyncratic concept, the concept of tharthritis. The price to pay since he cannot any longer be counted as making a certain kind of mistake, is that individualism leave out an essential normativity from the study of meaning. 2 The point of the conclusion can be, and has been, made in terms of rules. That is, one can think of the use of terms and the application of concepts as being governed by rules, which are the standards by which particular uses and applications can be judged to be correct or mistaken. In recent years much of the discussion of normativity, and its relation to meaning and content, has concentrated on the nature and possibilities of following linguistic rules. That brings me to the second example, which I adapt from an adaptation by Kripke of an example of Wittgenstein's. (By the way, there is an obligatory and ritualistic warning that every writer on this subject must pronounce because there are a multitude of inconsolable authors and coauthors who have protested that Wittgenstein did not draw the conclusion Kripke does from the example. You are hereby warned that I have no interest in that question, and proceed as if both intend the conclusion.) Imagine a man, I will call him KWert, who for several years has used the word arthritis (as we, his fellows, do) to talk of a disease of the joints only. Then on and from January 1st 1990, he is heard to say things like Ί have arthritis in my thigh.' We are inclined to say that he is making a mistake. He is not applying the rule for the concept of arthritis correctly. But we are asked, with what authority we say this? Why do we not say that he is quite correctly applying the rule and indeed was all along 2

There is a widespread tendency to conflate the distinction between individual and social with the distinction between private and public. The two distinctions do not coincide and there is no denial of publicness at all in my sketch of the individualist position here. For more on this and the other themes of this paper see chapter 3 of Bilgrami (1992).

Norms and Meaning

123

applying the rule for the concept of quartbritis. The rule for the concept of quarthritis, unlike the one for the concept of arthritis, does not say: "Apply the term 'arthritis to rheumatoid ailments with properties x, y, z, and which afflict the joints only." Rather it says something more bizarre and complicated: "Until January 1st 1990 apply the term 'arthritis' to rheumatoid ailments with properties x, y, z, and which afflict the joints only, after that apply it to rheumatoid ailments with properties x, y, z, and which afflict the joints as well as the ligaments." Kripke, unlike Bürge, presents his example as posing a sceptical problem about the very notion of meaning. The sceptical problem is just this: there is no fact about KWert that will decide for us that he has the concept of arthritis rather than of quarthritis. And this is a perfectly general problem. It is quite possible that any given concept of ours is afflicted with this sort of future possibility of what, to our common sense, might appear as lunatic deviant application. In that case we are not sure what any of our concepts are at any time. Various candidates for facts about KWert which might be thought to decide the matter about which of the two rules is being followed (which concept he really has) are canvassed and rejected by Kripke. In particular, facts about his neurophysiology, images of one kind or another, an inner conviction that he meant to be following one rule rather than another, and a variety of other sorts mental occurrence which KWert may have undergone when he applied the term. All those facts are as much in need of interpretation as the behaviour they are supposed to deliver a verdict on; and there is the problem that if such purely inner things gave our words meaning, we may render meanings private. 3 This is just Wittgenstein's version of anti-internalism. ("An inner process stands in need of an outward criterion.") Crucially, another sort of fact, not, in its essence, internal and so one might think acceptable to Wittgenstein, is also rejected. These are the dispositional facts about KWert, how he is disposed to use the term 'arthritis'. But that won't help either. Dispositions are descriptive of the person to whom they are attributed, and they are read off his behaviour. If so, we are never in a position to say that his behaviour on January 1st 1990 counts as a mistake. To appeal merely to dispositions is to appeal to something that does not allow us to say anyone is mistaken in this way. Equally, therefore, it does not allow us to say of KWert's linguistic behaviour and usage before that (or any body at any time), that it is correct. 3

It may be thought that neurophysiological and inner functional states are perfectly public, despite being inner, and so if there is a reduction of meaning to such states then publicness at least is not threatened. This would be to totally miss the point underlying the insistence that meaning and content are public.

124

Akeel Bilgrami

But to be able to judge correctness and mistakenness, Kripke says, is essential to our notion of rules, meanings and concepts. These are essentially normative notions. No amount of refinement of the idea of disposition will help. People do what they do and they say what they say and they do so in relatively uniform and regular ways; and their dispositions are attributed to them on the basis of that. But these are just facts about regularities in their behaviour and projections based on them. They cannot settle an issue which is, in its essence, an issue that goes beyond facts into matters of correctness and mistakenness. That is, into considerations of normativity. 4 Nothing about KWert, then, decides the issue. Not his internal states nor his dispositions. And this is not just a difficulty about finding out which concept he has. There is nothing to find out, no fact of the matter. Thus it is a scepticism about our very talk of concepts and meanings. Kripke admits that there is no solution to this sceptical problem, if the solution is to take the form of those rejected above. All those efforts fail and so if there is a solution, it must proceed from a recognition of something principled in those failures. Acknowledging the failure of these efforts is to concede something important to what the sceptic is saying. For this reason Kripke calls his own solution a 'sceptical solution'. 5 To put it roughly it is this. The scepticism consisted in saying that nothing about an agent determines which concept he has. The solution does not, as I said, combat this claim. It concedes it. It proposes instead not to look merely at the agent for the solution, but to look at the community in which he lives and speaks. If the agent applies the concepts to things in a way that squares with the way the community applies it, then he is correct. If not, not. If his answers to addition problems, his judgements of the location of disease, square with the community's answers and judgements, then there will be no reason to count them as mistaken. If they don't square, then there will. The sceptical solution appeals to the community to bring in the normativity and thereby to bring in the fact of the matter which will decide which rule is being followed by KWert. 4

5

Kripke has other criticisms of the dispositionalist answer but I will focus only on this criticism invoking normativity since my interest is primarily in the relevance of normativity in motivating a certain externalist view of content. Kripke himself is very explicit that it is this criticism which he takes to be the deepest and the one that is ultimately devastating of the dispositionalist answer. Kripke sometimes characterizes the 'sceptical solution' as a position which gives up on providing truth-conditions for sentences attributing meaning and rules, a position which resorts to providing only assertibility conditions for them. By contrast a 'straight solution', if there was one, would be a position which succeeded in giving truth-conditions for them. I don't think this way of putting things adds any clarity to the distinction between the t w o kinds of solution. See (Bilgrami 1986) for a scepticism about the possibility of making out a viable distinction between truth and assertibility conditions.

Norms and Meaning

125

If he is a member of our community he is following the rule for arthritis and he is mistaken on January 1, 1990. It should be obvious where the role of normativity enters in Kripke's dialectic. Normativity is presupposed, by Kripke, as being essential to the meaning of terms; and it is introduced into the dialectic of his sceptical problem at the point where one thought one could get away with an answer to his sceptic, an answer which appealed to the regularities in individual agents' behaviours and the dispositions attributed to agents on that basis. It is the 'dispositional' view of the meanings of terms which falls afoul of the demand for normativity. It does not allow us to say of KWert's answers after January 1990 (or any uses of terms, generally) that they are mistaken. We could just as easily see him as following the more bizarre and complicated rule mentioned earlier, and see him, therefore as making no mistake at all. But if we look to the members of his community and find that his answers after that date don't any longer square with the answers they give, we do have the resources to say that he is mistaken. Looking to the community, thus, captures the normativity. In both Bürge and Kripke, then, the community plays an essential role. But the underlying common factor in both is the stress on normativity as a motivation for their appeal to the community. That one of them is attacking a certain form of individualism and the other is attacking individualism as a way of responding to a sceptical problem is a minor difference of presentation compared to this shared conception being presented. In this paper I am not much concerned to combat the social element that these philosophers appeal to. The theme of this paper is the underlying demand for normativity that motivates the appeal to the social. II. The Irrelevance of Norms Individualism, we've been told looks at just the regularities in the individual's linguistic behaviour and therefore cannot retain the normativity essential to meaning. In any given case, it sees no reason to say that an individual must be attributed the concept of his social fellows. Of course, if his regular uses coincided with his fellows' regular uses, one could talk of their concepts overlapping but the point is that the constitutive relation of social use to individual meanings are eschewed. More could be said about this individualist position and many have said it, most notably Davidson himself. The question I am concerned with is how they should respond to the question of normativity. Individualists have wrestled with this question, trying to provide some other source for normativity than the social, and others

126

Akeel Bilgrami

despairing of providing an alternative source have tried to say that normativity of meaning is sui generis. It is without source. I will argue that they are all, in one sense, as wrong as Kripke and Bürge. One should instead treat the question of normativity with contempt. Normativity is irrelevant to the meaning of words. Once one grants the need for normativity, it would seem that Kripke's appeal to the community's practice is impeccable (though see the different interpretation of Kripke that I discuss at the end). It would be impeccable at least so long as one does not see the normativity as coming from some set of internal mental states (inner feelings of conviction etc.). Or alternatively as coming from some set of Platonistic abstractions (like 'Forms' or 'Meanings') or Platonistic metaphors (such as 'rails which are set up independently of our linguistic behaviour and which our linguistic behaviour tracks'). We have already seen that Kripke, following Wittgenstein, rejects the internalist ways of providing the fact of the matter at least as vehemently as he rejects the dispositionalist picture. And ever since the middle-period of his writing, Platonism is as much the target of Wittgenstein's attack as the internalist picture and the dispositional picture, so Kripke does not want the right to norms to come from the Platonistic sources either. What then can the source of the desired normativity be but the social element? However, a question that does not seem to have occurred to Kripke is this. Does the Platonist's mistake lie merely in thinking that the norm is afforded in the wrong sort of thing, an abstract entity or the 'rails' metaphor; or is it rather that Platonism's mistake lies deeper — in the fact that it hankers for such normativity at all, hankers for such final judgements of rightness and wrongness in the use of words? It is possible, I think, to argue that the erecting of metaphysical abstractions or the metaphor of independent rails which our linguistic behaviour tracks, is a relatively superficial shortcoming of the Platonist picture; and it is possible to argue that a deeper analysis would show that the real trouble with Platonism lies in the desire to see our concepts and meanings as governed by normativity in the sense of being demanded by Kripke and Bürge. It lies in the desire to find our use of terms right and wrong in the sense being demanded by them. Kripke and others never stop to ask what the compulsion is to conceive of the use of words in this way. They uncritically presuppose that it must be so conceived and then argue that unless we saw meanings as governed by abstractions or exaggerated metaphors on the one hand and inner or private mental states on the other, we would have to appeal to the social considerations of the sort we more recently find in Bürge as well. Without this presupposition, however, the regularity-based view of concepts and contents being objected to is not obviously inadequate.

Norms and Meaning

127

This may sound as if I am agreeing with Kripke that there is no nonsceptical solution to his problem and am offering an alternative sceptical solution to Kripke's own sceptical solution? But I think the more accurate description is that I am taking a way out that does not quite acknowledge that there was a terribly big problem in the first place and that I am giving a diagnosis as to why it has been thought to be a problem needing a solution of one kind or another. That is, I am not offering either a sceptical or a non-sceptical solution, I am rejecting the problem as being based on a misguided philosophical dogma, a residual Platonism. The diagnosis I am giving for the anxiety which leads to posing a sceptical problem is this: in rejecting the abstractions and metaphor of Meanings and 'rails' on the one hand and the internalistic mentalism of inner facts of the matter on the other, one has not yet succeeded in rejecting what in Platonism underlies the search for these things being rejected. Without rejecting this deeper urge, one will no doubt find another such thing to gratify the Platonist urge and indeed one has found it in society. This deeper urge underlying Platonism is precisely the drive to see concepts and terms as governed by such normativity. It is only because of this view of concepts, after all, that we are not able to be indifferent to the question: which concept does KWert really have, arthritis or the more accommodating notion of quarthritis? It is only because of it that we see the question as raising a 'problem', a sceptical problem about meaning and concepts. It is only if you have this residual and, in my view, deeper Platonist picutre of meanings and concepts that you will think that KWert raises a sceptical threat about them. For it will matter to you that he have one or the other concept. If it did not matter, you could not find him ever making the sort of mistake that the drive for normativity demands. And it is only because this matters to you that you will find some ersat2 version of the Platonist picture, cleansed of its unpalatable metaphysics and metaphors. Notice that this criticism of Kripke is not the, by now, familiar criticism of him which takes him to have failed to fully understand the full force of his own demand for normativity. 6 Philosophers have, like I am doing, opposed Kripke's social view of meaning and concepts but on altogether different grounds from mine. John McDowell, for instance, has argued that the way Kripke brings in the social is just an extension of the normativity-denying position of the dispositionalist because all Kripke does is bring in the dispositions of other members of the society to account for an individual's meanings. So he says if something was missing in the 6

See McDowell (1984) for a most forceful and detailed statement of this criticism. See also Blackburn (1984) for the same convincing criticism of Kripke's social view even though in the end Blackburn, unlike McDowell, opts for an individualist perspective.

128

Akeel Bilgrami

individual dispositionalist account in the first place, then it will be missing in the social extension as well. This criticism seems to me to be fair enough, if one accepts the normativity demand as one finds it in Kripke and as one finds it in these others who think that Kripke has himself failed to live up to that demand. But I do not accept the demand in the first place. So mine is a much more fundamental criticism of Kripke. In my view, one should repudiate the 'Platonism' altogether, (even in its ersatz forms) and in doing so give notions like meaning and concepts a much lower profile, whereby it does not matter very much that one is not able to say that KWert is making a mistake on January 1st 1990 or that Burge's protagonist has all along made a mistake when he applies the term to a condition is his thigh. If KWert regularly persists with his new usage and makes accommodating changes with the use of other terms conceptually related to his term 'arthritis', then on the individualist view we will say that he has a new concept after January 1990. But Kripke's sceptic may persist and say the question asked is not whether he has a new concept but whether he all along had the concept of arthritis or all along had the weird concept of quarthritis. To this question, the answer should be that it makes no difference to anything at all, which answer we give. His behaviour is equally well explained no matter what we say. There is no problem, sceptical or otherwise. Taking one line, one could explain all his relevant behaviour before and after 1990 by attributing contents containing the concept of quarthritis. Or taking another line, one could explain his behaviour prior to 1990 with contents containing the concept of arthritis and after that date with contents containing the concept of tharthritis. (The concept of tharthritis, you will recall from the discussion of Bert, is the concept of a disease which afflicts both the joints and the ligaments. It differs from the concept of quarthritis). Of course the behaviour itself would get somewhat different descriptions depending on which line we took. The point is that these would just amount to different but equivalent ways of describing the same behaviour (explanandum) and the same intentional states (explanans). There is no urgency any longer to the question which concept does KWert really have. No more urgency, as Davidson says in a slightly different context, then there is to asking whether something is three feet long or a yard long 7 .

7

The context in which Davidson invokes an analogy with different scales of measurement is one where he is discussing Quinean indeterminacy of interpretative theories rather than rule-following scepticism. The difference of contexts is not by any means trivial because rule-following scepticism in Kripke and others is often tied to examples where the behavioural evidence is restricted, whereas Quine usually discusses indeterminacy in situations where the totality of behavioural evidence is assumed to be in.

Norms and Meaning

129

III. Is Any Kind of Normativity Relevant to the Study of Mind? It is important to be clear about what exactly I am denying when I reject the notion of normativity here. There are two points to be made in order to be clear about this. First, I am only denying normativity in a certain domain, the domain of the meanings of words. This does not mean that I am denying norms of other kinds which might affect our notion of intentionality, such as in logical reasoning. And second even in the lexical domain of the meanings of words, I am only denying norms with a high philosophical or constitutive profile, i.e. what might be called intrinsic norms. I am not denying what might be called extrinsic norms. I'll elaborate on each of these points in turn. 1. There are no lexical norms. My rejection of Kripke's Wittgenstein's and Burge's insistence on normativity in the study of the meaning of terms is exactly that: a rejection of normativity in the study of the meanings of terms, in matters of the lexicon. It should not be taken to imply a similar rejection of normativity in other matters that affect the study of intentionality. The point can be put very crudely at first and then qualified. It can be put by saying that I am only rejecting the idea that norms have a very high profile, when it comes to certain matters of language.8 But there may be very good reason to think that the notion of norm has a much higher profile when we are talking, not about the meanings of words, but about the codifications of, say, deductive rationality. (Or of decision-theoretic rationality, possibly even of inductive rationality.) For here we do not leave things to the regularities in the speech behaviour of individuals but indeed assess individual behavioural performances themselves in terms of norms which are in place autonomously of such regularities. And, I think it is fair to conclude, in general, that if something which appears to be a norm is attributed merely on the basis of regularities in an individual's behaviour, then it is not a norm in any interesting sense. If it is not derived

8

See also Chomsky (1986), ch. 4 for a rejection of norms in the study of language. I should nevertheless distance my position from his because I know that he has objections to my positive remarks on the regularity-based, externalist-holist nature of concepts. In (Bilgrami 1985) and (Bilgrami 1987) I had made some very primitive remarks, first to dissociate normativity in Kripke's sense from the question of meaning and content (suggesting instead that normativity in the sense that is involved in logic were far more relevant.) and second, to separate the question about the normativity of meaning from the publicness of meaning. In making them I was clearly influenced by the general individualism that both Chomsky's and Davidson's philosophy of language make possible, though I was not aware that Chomsky had written explicitly on Kripke.

130

Akeel Bilgrami

from or attributed on the basis of such regularities, if it is autonomous from such regularities, then it has more of a right to be called a norm. By this I am not suggesting that when it comes to logical concepts and principles, we should endorse what I have been opposing in Kripke's Wittgenstein's proposal about meanings generally, viz., that the norms involved are determined by the practices of the community. The right thing to say is that they are autonomous in some further sense. This cannot be the place for me to explore the normative status of logic. The idea of the sort of autonomy we might allow for the norms of logic has already been explored in different ways by different philosophers. Davidson's writings suggest one path for exploration. He has suggested that we could not make sense of an agent, could not attribute intentional states to him, if we did not see those states as satisfying those norms to a very great extent. This idea certainly puts those norms at an autonomous distance from the regularities in the behaviour of individuals since it is a constraint on the assignability of propositional contents to agents based on the observation of their behavioural regularities. In order even to make sense of the regularities we have to impose these norms. We have no choice in the matter. So the norms cannot possibly be based on an observation of those regularities. They form an a priori constraint on the very possibility of interpretation. But when it comes to the lexicon, it is perfectly possible to attribute concepts and meanings on the basis of observation of regularities in the behaviour of agents. As I have been saying there is no compulsion, therefore, to impose the norms dictated by social practice on an individual agent's concepts and meanings. We do have a choice in the matter. We do have the choice as to whether Bert has the concept of arthritis or tharthritis. We do have a choice as to whether KWert has the concept of arthritis first and then later tharthritis or whether he has had the concept of quarthritis all along. It is possible to account for their behaviour no matter which of these alternatives we attribute to them. It is possible, therefore, to think of our attribution of such concepts as something done (holistically, of course) on the basis of observed regularities. Normativity, thus, simply does not have the same grip in this domain. 9 That was, as I said, a crude statement of the point. It needs to be made less crude because there is an obvious objection to it: I have made a distinction between lexical matters or matters having to do with concepts on the one hand and matters of logical inferential relations between sentences 9

My use of the word 'lexicon' in drawing the distinction with logic might be a little misleading since logical words like 'and' and 'or' may be said to be part of the lexicon. What I am really distinguishing are the logical terms and the non-logical primitive predicates and singular terms.

Norms and Meaning

131

on the other, but I have failed to mention that there might be conceptual inferential relations which are relevant to the lexicon. That is, there might be inferential relations between items of the lexicon. And these too, it might be said, force a normativity which is autonomous from regularities in behaviour and which is akin, therefore, to the normativity I am granting to logical matters. The objection is that one cannot distinguish too strictly between the formal norms of logical inference and the material norms of the inferential relations that hold between concepts and meanings; and one cannot do this because it is absurd to suggest that for all the sentences of someone's speech, one knows that this, that and the other sentence are logically related to one another but one does not know what any of them mean. Detection of logical inference in an agent is not altogether independent of detection of material inference. Nobody can deny this point. But I still think that the comparison is inexact to a fault. It is still possible to argue, for instance, that the role that the requirement of (say) logical consistency plays as an a a priori constraint on interpretation is very different from the role of the requirement which says that in order to have one concept one must have some other beliefs and concepts. The idea of material inference is just the idea that one cannot have a concept (say, the concept of table) without having some other concepts (say, the concept of a middle-sized object). But that merely says I cannot attribute my concept of table to someone who fails to have certain other beliefs or concepts. It does not require that any particular concept that he has must be the concept dictated by a norm set by my linguistic practice or by social linguistic practice more generally. The former requirement of logical inconsistency, therefore, still seems properly describable as an imposition of norm whereas the latter requirement does not. The latter merely says something rather general, viz., that an agent's concepts are interdependent on his other concepts and beliefs; but the former has relatively clear and stable codifications as to what amounts to consistency and what does not. It, therefore, makes much more sense to think of the former as involving norms. There will be a further protest which I must address. Someone will say that I have failed to stress the autonomy (from regularities in behaviour) that attaches to the idea of material inference. I have merely said that there is an interdependence of any given concept on other concepts. They will say that Davidson himself., whom I have invoked to make the point of the autonomy of logical norms, commits himself to a similar autonomy for lexical norms when he makes his famous (infamous) claim that an interpreter and interpretee, in radical interpretation, must by and large share concepts and beliefs if there is to be interpretation at all. That is, in interpretation, material disagreements or failure of overlap in concepts make sense against a background of agreement and conceptual overlap.

132

Akeel Bilgrami

So the radical interpreter is (by and large) willy-nilly imposing the material inferential relations between his concepts on to those whom he is interpreting. If this too is being done, not on the basis of the observation of regularities in the interpretee's intentional behaviour, but is part of the very idea of making sense of those regularities, then by my own criterion (autonomy from regularities of behaviour!), one has just as much right to say that normativity governs the domain of concepts and the lexicon as the domain of logic. The idea is that it is not just that there be material inferential relations, not merely that there be interdependence among an interpretee's concepts, but rather that there be overlap between the concepts of the interpretee and interpreter. So the protest is: why not say that the interpreter imposes not just logical norms but also his own concepts and beliefs. Does this not amount to conceding that there are norms of the lexicon? Does it not amount to conceding that both logical and lexical norms are autonomous of regularities in the interpretee's behaviour and both are imposed on the regularities of an agent's behaviour in order to assign contents at all? The idea that there must be by and large conceptual overlap between interpreter and interpretee is of course controversial. But even if we grant the idea, it may still be possible to retain our qualms about seeing the lexicon as governed by norms. It may still be possible to say that Bert is not even violating a non-social norm, not even a violating a norm generated by my (the radical interpreter's) idiolectical linguistic practice. I do not have a strict argument here, no strict criterion of the difference between the genuine norms of deductive rationality and the so-called (in my view spurious) norms of the lexicon, but I will briefly try and convey a rough sense of the difference. One way of conveying it is to look at the different character of the exceptions to or violations of the 'norms' in the two cases. When someone (an interpretee) falls short of deductive rationality, it is perfectly clear in advance what the norms will have to be which are violated. When there is a failure of conceptual overlap, that is, when there is disagreement in material inference, there are no such norms, clearly statable in advance, which are violated. For if there are norms involved at all in the domain of material inference, then which one is violated in the case of a failure of overlap will differ with each particular failure. By my lights, some interpretee may violate the norm for 'snow' by thinking that snow was warm, another by thinking that it was green, yet another by thinking that it grew underground, and so on and on. Things are not like that with failures of deductive rationality. With regard to violations of deductive rationality, we can say something like: There is a norm, such that for any violation of deductive rationality, it is a violation of that norm.

Norms and Meaning

133

But, if I'm right, with regard to the failure of conceptual overlap one has to say something quite different, something more like: For any failure of conceptual overlap (or material inference), there is a norm, such that the failure is a violation of that norm. This shift in the quantifier makes vivid what the difference is. It is only apparently the case that there are norms governing concepts, because — as the example of 'snow' above makes clear — the material inferential relations that fail to hold, in the cases of failure of conceptual overlap, are so highly contextual, so highly situated in each given case that there is no sense in which it is useful or necessary to call them norms and to say that it is norms that are being violated. On the contrary, logical norms are highly context-free and one can say quite exhaustively and in advance what the principles are that are being violated in the case of logical failure. 10 Here, two points must be made by way of qualification. First, I am assuming that there is no clear analytic-synthetic distinction. If there were analytic truths, and in some uninteresting cases there may be, 11 then there would be clear norms governing concepts (norms of the lexicon) and the violations of material inferences would not be subject to the same situatedness as they are when there is no commitment to analyticity. Second, indeterminacy about some of the logical norms slightly spoils the distinction I am making between them and the so-called norms of the lexicon, but it would be utter hyperbole to think that they spoil it to the extent that the distinction loses its point. Both these qualifications may show that there is a spectrum rather than a sharp distinction. But the rough sense of the distinction between the genuineness and the spuriousness of norms in these different domains can be conveyed by looking at the two ends of the spectrum. If one just considers the principle of non-contradiction at the one end and the so-called rule for 'snow' at the other, it should be absolutely obvious what the distinction amounts to. So, nothing about lexical norms follows even if we granted the Davidsonian point that if there is massive failure of conceptual overlap in interpretation one has violated an a priori constraint of rationality on interpretation. The a priori demands for logical consistency are wholly unlike this sort of a priori constraint. The latter does not warrant the claim that the meanings of our terms are governed by norms in a way that the former shows that thoughts are governed by a norm of logic. It is much more reasonable to say instead that particular cases of failure of conceptual It is also probably arguable that the codifications of inductive logic lack this freedom from context. " There may also be cases of norms governing some concepts which may not quite amount to analytic definitions such as transitivity for the concept of length. 10

134

Akeel Bilgrami

overlap are not a violation of norms but just a matter of there being different meanings and concepts. That is just what I have been saying about Bert against Bürge and KWert against Kripke. I believe that a number of philosophers have unconsciously run together (or at any rate assumed an easy extension of) the undeniable normativity that attaches to the principles of logic and what they take unjustifiably to be the normativity ('rules') governing the meanings of words. I have tried in the last few paragraphs to diagnose why they should unconsciously fall into this assumption by looking to the similar status both sorts of norms might be thought to have as prerequisites for successful interpretation. 12 2. There are no intrinsic norms Does all this mean that I am committed to saying that there is absolutely no sense of norm attaching to matters of the lexicon on my picture of intentionality? That of course is a matter of nomenclature. So long as one is clear that it is not the sense of norm that is driving Kripke, Bürge and McDowell and others, I am happy to admit that there are norms with a much lower profile at play in the unified, individualist picture of intentionality. To repeat, what I am denying is only the conception of norms which drives either the anxieties about being able to say that Bert is making a mistake or which finds a sceptical problem in KWert. What sense of norm, then, do I accept? It emerges from the fact that individuals intend to speak like others in the community speak and they intend to speak as they have in the past rather than waywardly. This means that individuals do intend to speak in a way that is natural to describe as, and that they themselves describe as, correctly. It is a sense of norms that emerges from purely pragmatic considerations. By that I mean that there are very good utilitarian (and of course historical) reasons for us to speak like others in our society do or for that matter to speak with regular habits, i.e. to speak as we have done in the past. And if one sees linguistic norms as motivated by pragmatic reasons, there is no reason to think that Bert has the same concept as his fellows. It is not that Bert possesses the same concept as his fellows and has made the mistake of misapplying a concept he possesses, rather he has made the 12

I think the philosophers who are most clearly guilty of this are John McDowell and Philip Pettit in (McDowell and Pettit 1986). See especially p. 13. Here they explicitly invoke Davidson's arguments for the imposition of norms of consistency and coherence in order to defend Burge's social externalism about terms such as 'arthritis' and other natural kind terms. If all that I have been saying about the difference between Davidson's interest and the interest in the lexicon is correct, the kind of defence they argue for is simply unavailable.

Norms and Meaning

135

mistake of not speaking as his fellows do, not having the same meaning and concept as his fellows when that, for pragmatic reasons, is what he intended to do and have. Thus we not only have a normative element in the individualist picture of content, we have a norm that is in fact perfectly sensitive to the social linguistic practices that surround an individual. But the pragmatic explanations underlying the norm make clear that it is a norm in a sense that is quite different from the one that Bürge says is violated by Bert. It is an extrinsic norm which necessarily has a hypothetical formulation. It says "Speak like others do, if it pays to do so." or more specifically it might say "I ought to use words as others do, if I want to be easily understood."^ This norm is so obvious that we never make it explicit to ourselves until perhaps we leave our own communities and go to others ('When in Rome ...'). And perhaps not even then. Turning from Bert who deviates from social linguistic practices to an agent who departs from his own regularities, he too has an extrinsic norm which he violates. The norm says, "I ought to use words as I have used them in the past, if I want to be easily understood." 14 And to speak as I have in the past (or as others do), in turn, of course, means that I ought to use the word 'arthritis' to talk of this sort of thing rather than that or any other sort of thing. These norms are necessarily extrinsic. If this is right, the idea that we ought to use the word χ and not y to convey z, because ζ is what X means, is strictly reducible the idea that we ought to use the word χ to convey ζ because we have done so in the past (or because others do so) and therefore using them will allow us to convey it without causing strain. The connection between 'ought' and 'mean' here is none other than this lowprofile connection. The normativity involved, if it is strictly reducible in this way to something extrinsic like this, is certainly not groundless or sui generis. It is firmly grounded in our desire to communicate without strain. The 'oughts' I am admitting are extrinsic in the sense that if one has a certain ends and a certain view of the means to achieve the ends then one ought to pursue those means. There is nothing special about the norma13

14

Of course, often, as when we are being poetic or perverse, we may deliberately flout this tacit hypothetical norm. Those, however, are cases essentially different from Bert. Notice that I am not saying that this hypothetical norm is sufficient for meaning. Someone may think I am saying that since I am denying that the categorical norm I am contrasting it with is necessary. But to say it is sufficient would be false since, I suppose someone may not ever have the desire to be easily understood and so the antecedent of the conditional would be false — and here one would want to deny that there was meaning even though the conditional remained true. Rather my point simply is that there are no categorical or intrinsic norms of the lexicon, but, given the desire to communicate, one often implicitly thinks one ought to do whatever it is that fulfils and facilitates that desire. If that is a norm, then it is at best a hypothetical norm in precisely the sense that Kant defined and criticized. It is a norm which tells us how best to bring about something extrinsic which we want.

136

Akeel Bilgrami

tivity which attaches to the meanings of words. Meanings are normative in the sense that a lot of other things might be — such as that I ought to take my umbrella if I don't want to get wet, or that I ought to listen to songs by Hugo Wolf, if I want to get depressed. Similarly there are certain things I ought to do if I want to be easily understood — I ought to speak as I have in the past or I ought to speak like others do. And, of course, in turn, if I want to speak like I have in the past or like others do, I ought to use words in some ways and not others. The point of these extrinsic norms is that neither intentionally conforming to others usage nor even intentionally sticking to the past ways of one's own usage, is a sign that meaning has an essential or primary aspect of correct and incorrect usage. The notion of correctness is entirely secondary to the desire and intention to communicate without causing strain, which underlies the notion of meaning. Actually it is more than a little misleading to talk of these imperatives as being hypothetical imperatives in the way that I have been. So also it is not enough merely to talk of ends and means to speak of norms. This is because the imperative "If you want to speak English, do not apply 'arthritis' to diseases of the ligaments' is hypothetical in form also and it too may be seen in terms of ends and means. But qua imperative it appeals to something intrinsic to language, it appeals to rules of language. It says that one ought to say χ rather than y in certain circumstances because that is what χ means. Meaning is the resting point in the unpacking of this ought. But it should not be the resting point. The hypothetical imperatives must appeal to something extrinsic, something it pays to do, such as not causing strain in understanding. So it is really extrinsic norms that I am stressing, rather than hypothetical. This pragmatic element introduced by such expressions as " e a s i l y understood" or 'understood without strain" is not a cancellable part of the norm. If I had left this element out and instead formulated the hypothetical as "You ought to use χ in a regular way, if you want to be understood", that would still be to appeal to an intrinsic consideration because it would be too much like resting with an unpacking of the ought in the notion of meaning. Why am I insisting on the extrinsic, pragmatic qualifications to the idea of understanding in these imperatives. Because, as Davidson makes clear it is a plain fact that for any given term, I can be understood even when I do not use it in a regular way. It's just that I will probably cause strain and not be understood easily. I may make a listener sit up but its use can, with effort, be understood. Someone might protest my insistence on bringing in the pragmatic element into the norms on the following ground. They might say that an individual must behave in regular ways not just in order to be understood without strain, but to be generally understood at all. If there was not by and

Norms and Meaning

137

large regularity and stability in our linguistic usage we would not generally be understood at all and would not be counted as linguistic agents. I don't deny this last point. However the point does not support the protest at all. A claim about how for us to be generally intelligible there must be by and large regularity in our usage does not in any way contradict the point I just made that in any given case, one may use a term which is not in accord with any regularity and yet be understood. The crucial point is that norms attach to particular terms. And they assess usage for correctness and incorrectness on particular occasions. They allow us to say such things as "You ought to use the word 'plus' as you have in the past" or "Bert used 'arthritis' incorrectly when he said Ί have arthritis in my thigh'". And I am denying that there need be anything as specific needed to be generally understood. I can use a term once and never use it again and be understood on that occasion. I have said "The gavagai is too loud" and found that a friend of mine (who had never heard the sound 'gavagai' before) turned the volume of the radio down; there is the famous story about a Cambridge philosopher who is supposed to have smiled politely as she was leaving a party and said 'Fuck you very much' to the hosts, who we can imagine said "Come again" ... Examples can and have been multiplied. 15 I never before used the term 'gavagai' in such circumstances and probably never will. For all I know the Cambridge philosopher never uttered the word 'fuck' at all in any circumstance before or since that occasion. So: no need for regularity with any given term in order to be understood. Of course the utterances of these terms would not have been understood without a background of regular usage of a lot of other terms, but that is a general requirement as I said which does not yield any particular norm for any particular term, no specific intrinsic lexical imperative such as "Use 'gavagai' under conditions if you want to be understood!" Not even "Use the word 'loud' under conditions if you want to be understood". The next moment after my 'gavagai' utterance I might have said looking at the table in front of us, "The loud is too messy" and my friend might have tidied up the table. This last utterance would certainly not have violated any so-called norm for loud, which was allegedly operative in the previous 'gavagai' utterance when I said "The gavagai is too loud." The second use of loud has no interesting connection with the previous use of loud at all. And if it is not a violation of a norm, there is no sense in claiming that I got away with my use of 'gavagai' instead of 'radio' there because at least the norm for 'loud' was operative. The fact that the next moment I used 'loud' in an utterance to mean 'table' without violating the so-called norm for 'loud' suggests that the norm for 'loud' is merely 15

See Davidson (1986) for examples put to similar use.

138

Akeel Bilgrami

so-called. It is not a norm in any sense because such idiosyncratic uses of that word are not a violation of it. So the concession to by and large regularity needed for the intrinsic goal of general intelligibility does not translate at all into a claim about how one ought to follow this or that norm in order to be understood. The latter (i.e. particular imperatives for particular words rather than general demands for overall regularity) is all that deserves to be called 'a norm' and it is the latter which I, following Davidson, am denying. Nothing but philosophical conflation and confusion is gained by insisting that the sort of thing that Kripke, Bürge and McDowell demand (norms for the meaning of particular words, such as 'arthritis' or 'plus') in the examples and puzzles they devise is the same as the demand that there be by and large regularity in our linguistic usage in order for us to be understood and counted as linguistic agents. So just in case you were getting the impression that the point of the paper so far has been to replace a notion of norms based on social practice with a notion of norms based on individual regularities, this last point should have made it clear that the position on norms is more radical than that. The last point establishes that though regularities, in general, are important to interpretation and meaning, they do not yield any sense of norms for the meaning of words. If I am right, meaning no longer has the normativity of that highest deportment demanded by what I called ersatz Platonism. Burge's and Kripke's view demands a more intrinsic or categorical sense of norm. It demands that Bert and the other agent are counted as mistaken not because they confuse or strain or mislead others (and because they have a tacit norm directing them not to confuse, mislead or strain them). That would be too instrumental a reason for counting them mistaken. They are counted as mistaken because they have violated something more constitutive and intrinsic to the idea of meaning. Therefore the idea of extrinsic or hypothetical norms will not count as norms at all for the ersatz Platonist. They will not have a high enough profile. From his point of view one either has intrinsic norms or one has given up on normativity. If so I think we should give up on normativity.

IV. Norms and Intentions So far I have applied the lesson learnt form Davidson's attack on the notion of convention to criticize Kripke's demand for normativity in the study of meaning. In doing so I have interpreted Kripke's demand in a way that the lesson seemed to me applicable.

Norms and Meaning

139

But at the beginning of this paper I had mentioned two Davidsonian lessons. I now turn briefly to the second though not unrelated lesson, one having to do with the importance of communication in the study of meaning; and I apply it to a somewhat different interpretation of Kripke's demand for normativity in the study of meaning than the one I have given so far. I have so far assumed that it is natural to see connection between Kripke's demand for norms in the study of meaning and his own positive solution to rule-following scepticism, a solution which appeals to social practice as being constitutive of the meanings of words such as 'plus', 'loud', etc. I have done so because it is a charitable reading of an author to interpret a demand he lays down in the study of a subject in such a way that the positive view he develops about the subject fits in with what the demand is demanding. But John McDowell has argued that Kripke's demand for normativity is quite different from what I have taken it to be. If his interpretation is right, then there is no way that the social practice as Kripke spells it out can even be seen to be relevant to the demand of normativity. If he is right, Kripke demands one thing and offers something which has nothing to do with it. I will not try here to defend my way of reading Kripke because there are one or two passages in the Kripke text which might suggest McDowell's interpretation. And in any case this paper is not intended primarily as a commentary on Kripke's book, so much as an exploration of how Davidson's conception of meaning may be exploited to undermine the normativity that Kripke demands in the study of meaning and intentionality. McDowell reads Kripke as simply saying that if one means anything at all by ones words, then intentions are involved in a way that brings a certain kind of normativity with it. Roughly speaking, his reading is as follows. Let us take someone who means fly by 'fly'. This requires that (when the person is speaking sincerely and non-figuratively) the person intends to apply the word 'fly' to flies, in his usage. By usage, I mean his use of sentences containing that term. Now it is the nature of intentions that only some actions fulfil the intention. Others will count as failures to carry out one's intention. This distinction between successful and unsuccessful actions has an internal or conceptual relation to our very notion of intention. An intention is the sort of thing which by its very nature demands a division of actions into those which do and those which do not carry out that intention. This general point about intentions applies therefore to the special case of one's intentions to use words. And when it does it yields a sort of normativity because it divides the application of words into correct and incorrect applications, i.e., into linguistic actions

140

Akeel Bilgrami

or uses which do and do not apply a given word as it is intended. Thus were I to apply the word 'fly' to a black spot, I would have failed to live up to my intention to apply it to flies only. I would have made a mistake. This notion of failure or mistake, and such possibilities of mistakes, are essential therefore to our notion of meaning. A purely causal or dispositional account of meaning cannot capture this essential aspect of meaning since it can have no place for right and wrong applications in this sense. Hence the constitutive normativity in the study of meaning. Notice first of all that this notion of normativity is perfectly generally true of all intentions and not just to the intentions to use words that are relevant to language and meaning. This normativity attaches just as much to my intention to go for a run this afternoon. If I go to a rock concert instead, I would have violated a norm in no more, no less and no different sense than my use of the word 'fly' in the presence of a black spot. And McDowell thinks that this is all that Kripke is demanding when he says that any account of meaning must meet the demand of normativity. If this is the right reading of Kripke, then it is altogether inessential to ask whether my present or future use of a term squares with my past use of a term. There is no longer any real parallel with the problem of induction, and there is no interesting debate about the limited or finite evidence of past usage and what can be projected on its basis. None of this is relevant since if I intend something by a word, then right now various things I might have said by using it will be either right or wrong uses in this sense of right and wrong. Notice, then, that as a result of this neither Kripke's appeal to social practice nor an individualist regularist position on normativity speak with any relevance to this notion of norm at all. This notion of norm simply follows from the notion of intention, its built into it. It is not necessary or possible to consolidate it in social practices or regular uses on the part of individuals since the notion of intention will already be needed to describe the social practices and the regular individual uses. The notion of intention must be understood prior to even appealing to these other things, so the normativity is already there prior to the appeal. One might, of course, ask what an intention is, but that has nothing specifically to do with meaning at all. And though it may be true that we could not be said to have intentions unless there was a society of other creatures with intentions, the fact remains that this appeal to society does not help with any particular intention such as the intention to use 'plus' in a certain way or to use 'arthritis' or 'fly'. It is an appeal to society which is altogether different from Kripke's since it is of no help at all with Kripke's question, which is how shall we account for the fact that some applications of 'plus' seem to us to be right and others to be daft. Kripke's appeal to social practice, whether one agrees with it or not, addresses this latter question about right and wrong. But the other appeal

Norms and Meaning

141

to society I just mentioned does not have any relevance to this latter question. If all there was to the demand for normativity was the one McDowell's reading suggests, it is not clear how we can make any sense of why Kripke appeals to the community's usage of words like 'plus' or 'fly'. So if this interpretation of Kripke's demand is right, Kripke changes his own subject when he tries to meet the demand by appealing to social practice in the way he does. Whereas on my interpretation of what Kripke was demanding, Kripke was not confused about what he was demanding, though as I tried to argue he was wrong to demand it. As I said, I am not going to pause to discuss the virtues of the differing interpretations. Rather I will go on to consider whether one should even demand the kind of normativity that McDowell is demanding for the meanings of words. I don't believe we should. To make this denial is not to deny McDowell's point that intentions have the sort of internal link to what is to be counted as successful and unsuccessful actions. It is not to deny that there is built into the notion of intention the idea of what it is to act in accord and discord with the intention. But the application of this point to our intentions to use words is much more nuanced than McDowell's view takes account of. And once the nuances are elaborated I think it will be clear that it is highly misleading to think that it amounts to any kind of norm even in the sense that McDowell's interpretation suggests. Let me explain. In order to make my criticism of McDowell's view of normativity, let me first invoke an assumption which is central to Davidson's conception of language, and though there may be some controversy about it, it does not seem to me something that McDowell would dispute. This is the assumption that any account of language must take it as primary that meaning emerges and is to be understood in the context of communication. We don't just have meanings or intend things with our words (I intend fly by 'fly') independently of the general fact of communication, independent of the general fact that we intend our words to be interpreted or understood in a certain way. Perhaps there are cases in which we mean something without really communicating anything, perhaps there are cases where we communicate something to ourselves but in which we do not need to interpret what is being said since it is to with ourselves we are communicating. But these are all rightly regarded as in some way degenerate cases in the sense that even making sense of such phenomena requires that the paradigm of communication is in place. Since McDowell's notion of normativity about meaning turns on the intentions involved in meaning, let us turn to what sorts of intentions are involved once we invoke this Davidsonian assumption. In the context of communication, two kinds of intentions are centrally involved in the use

142

Akeel Bilgrami

of words. By uttering the words, say, "That is a fly" I intend that I convey to someone that that is a fly, I intend that someone come to believe that that is a fly. But this intention is not itself a linguistic intention and is not what is of primary and direct interest when the subject is meaning. The intention which is crucial to meaning, and which Davidson calls 'linguistic intention' is the intention to utter words (and to be interpreted as uttering words) with certain truth-conditions, i.e. to utter words and to be interpreted as uttering words which are true if and only if that is a fly. This, of course, carries over to sub-sentential components, viz., I intend to utter a word with such and such reference. (My point here is not to emphasize Davidson's truth-conditional view of meaning, the linguistic intention could be reformulated with any other meaning-giving condition if one does not favour a truth-conditional view). Let's then turn to McDowell's claims for the normative element that intentions bring by themselves to meaning. McDowell's claim is that if I were to apply the term 'fly' to a black spot, then that would be a mistake since it would be a failure to live up to a certain intention relevant to meaning. The first thing to notice is that if the foregoing specifications of intentions in communication are correct, then McDowell has misformulated the intention that is relevant to meaning. These linguistic intentions are not to be formulated as "I intend to apply 'fly' to flies" or "By 'that's a fly' I intend to talk about a fly". Rather the intention that is relevant to meaning is to be formulated as: by 'That is a fly' I intend to say something and to be taken to be saying something which has certain, say, truthconditions, i.e. which is true if and only if that is a fly. But notice now that if one is clear about this then the sorts of things that McDowell wants to count as mistakes do not count as mistakes that are relevantly mistakes about meaning at all. And since his case for normativity turns so much on finding this sort of mistake as being a mistake about meaning, to that extent the case is undermined. Suppose I were (sincerely and non-figuratively) to say 'That's a fly' pointing to a black spot. Given McDowell's formulation of the intention that is involved in meaning, this would certainly count as failing to carry out one's intention, and therefore as making a mistake in the sense of norm that we are now considering. But if one formulated the intention relevant to meaning in the way I suggest above, it is not a mistake to say 'That's a fly' in the presence of a black spot. If I intended to talk about a fly when I said that, I certainly failed to do so. I did not apply 'fly' to a fly, I did not in that sense talk about a fly, I talked about a black spot. But this failure is not a mistake having to do with the intention that is relevant to meaning. For I certainly did not fail in my intention to say something with certain truth-conditions, just

Norms and Meaning

143

because what I was pointing to was a black spot. I still said something which was true if and only if that was a fly. The fact is that McDowell does not pay enough attention to the communicative context in which the intentions involved in meaning must be understood. As a result he talks simply of intentions to apply words to things, one's intention to mean fly by 'fly'. Now it is true that in communication one might also intend to speak the truth. We often do. And sometimes a failure to speak the truth involves the misapplication of a concept or word, such as in the present example. But a failure to live up to our intention to speak the truth in a particular case cannot amount to a mistake about meaning since our meaning is perfectly well communicated, as in the present example. It may be thought that I have distorted McDowell's point by stressing the communicative element too much and thereby restricting his point to being one about the meanings of words only rather than about thoughts and concepts as well. Its possible that what he has in mind is a perfectly general point about the normativity of thoughts and the application of concepts and not just about the conveying of meanings with words. Thus, he might point out that if I were to think that that is a fly, while focusing perceptually on a black spot, I have failed to live up to my intention to apply my concept of a fly to flies. The relevant intentions which bring in the normativity in meaning are first and foremost about this sort of intention, he might say, and only derivatively about meaning in communication. First of all, if that was the point it is highly misleading to express the point by saying that I intend fly by 'fly'. The quotation marks around the second occurrence of the word fly would seem inappropriate to make the point. It may not be misleading if one were convinced that there was a language of thought, though I doubt that McDowell has such a conviction. But I think even if there was a language of thought, nobody should be inclined to say that we have intentions regarding its particular expressions since this language is not a language we use. And in any case the idea of misapplying my concept of a fly when I think that a black spot is a fly, is nothing over and above the idea of my misperceiving something. It is simply reducible to misperception and to think of it as a failure of living up to an intention seems factitious. I'm certainly not going to deny that we sometimes misperceive things or that we have other similar kinds of false beliefs And no doubt when we do these amount to some kind of failure or wrong. But to think of it as the sort of failure that Kripke's discussion of normativity centrally demanded seems to me to be changing Kripke's subject (and McDowell's and mine) altogether. One may grant this normativity and wonder what it has to do with meaning at all, no matter how interesting and close one took the

144

Donald Davidson

relation b e t w e e n language and t h o u g h t t o be. T h u s n o t e v e n the minimal n o r m a t i v i t y that M c D o w e l l rightly f i n d s internal to the n o t i o n o f intention is c o n s t i t u t i v e o f m e a n i n g in the w a y that M c D o w e l l requires. I conclude, then, that neither i n t e r p r e t a t i o n o f K r i p k e ' s d e m a n d f o r n o r m a t i v i t y in the study o f meaning — neither mine n o r M c D o w e l l ' s — yields a d e m a n d that w e must feel c o m p e l l e d t o meet, o n c e w e h a v e learnt the lessons o f D a v i d s o n ' s o v e r a l l c o n c e p t i o n o f language.

REFERENCES

Bilgrami, A. 1985: Comments on Loar. Contents of Thought, eds. R. H. Grimm and D. D. Merrill (Tucson, Arizona: University of Arizona Press). Bilgrami, A. 1987: An Externalist Account of Psychological Content. Philosophical Topics 15. Bilgrami, A. 1992: Belief and Meaning (Oxford: Basil Blackwell). Blackburn, S. 1984: The Individual Strikes Back. Synthese 58. Bürge, Τ. 1979: Individualism and the Mental. Midwest Studies in Philosophy 6: Studies in Metaphysics, eds. P. French, T. Uehling, H. Wettstein (Minneapolis: University of Minnesota Press). Chomsky, N. 1986: Knowledge of Language·. Its Nature, Origin and Use (New York: Praeger). Davidson, D. 1984: Communication and Convention. Inquiries into Truth and Interpretation (Oxford: Oxford University Press). Davidson, D. 1986: A Nice Derangement of Epitaphs. Philosophical Grounds of Rationality eds. R. Grandy and R. Warner (Oxford: Oxford University Press). Kripke, S. A. 1982: Wittgenstein. On Rules and Private Language (Oxford: Basil Blackwell). McDowell, J. 1984: Wittgenstein on Following a Rule. Synthese 58. McDowell, J. and Pettit, P. 1986: Introduction to Subject, Thought, and Context (Oxford: Oxford University Press). Wittgenstein, L. 1953: Philosophical Investigations (Oxford: Basil Blackwell).

Donald Davidson: Reply to Akeel Bilgrami I have nothing but praise for this paper. It expresses views that I have long held, and expresses them better than I have sometimes expressed them myself. Naturally, then, it gives me a lot of satisfaction to hear them stated in such a convincing and subtle way. I would be delighted if I thought that Bürge and his followers, and Kripke and his followers, would be convinced by Bilgrami where they have not been convinced by me; but on this point I am not filled with confidence. Bilgrami's attack on the notion of linguistic norms seems to me extremely effective, and his suggestion, that the yearning for such norms is a nostalgic hangover from dependence on a Platonic conception of meaning, is convincing. He is right, too, in claiming that the only philosophically interesting concept of meaning must derive from cases of successful communication. We cannot define successful communication in terms of shared meanings, practices or conventions since we have no idea what meanings are until we can abstract them from occasions of use. (I take this to be something Wittgenstein pointed out long ago, but somehow the lesson was not really taken to heart except by those who thought it licensed them to give up on the search for a systematic theory.) Bilgrami doesn't quite put it this way, but it is clear from what he says that for him, as for me, it is successful communication from which notions like meaning, reference, translation, and interpretation draw their substance. The point of this observation is that it is not enormously important to decide what someone's words meant when communication succeeds by what we think of as non-standard means, or when communication fails. Whatever vitality the concept of meaning has depends on the successful cases; we can describe other cases as we please. I do not know whether Bilgrami will agree with two further, relatively minor, comments. The first concerns what Bilgrami calls "conceptual overlap" between interpreter and interpretee, which he says I consider a condition for successful interpretation. He calls this view controversial, and seems to distance himself from it. I'm not certain what the thesis is which is controversial; is it the claim that we could not understand someone with whom we did not share a large number of (fundamental) concepts? I know of no one who denies this, for how could an interpreter grasp, much less formulate, the truth conditions of an utterance which she lacked

146

Donald Davidson

the resources to conceive? Bilgrami seems, however, to be thinking of my insistence that understanding another requires a sharing of (non-logical) beliefs, the idea which Wittgenstein expressed by writing, " I f language is to be a means of communication there must be agreement not only in definitions but also (queer as this may sound) in judgments". This thesis is, perhaps, controversial, but as far as I can see it follows from the first. The reason is simple: quite apart from any connections between concepts that someone might consider analytic, most empirically grounded concepts can only be identified by (perhaps among other things) their relations with other concepts. This doesn't mean, as I have long insisted, that there is any fixed list of connections which if lost would alter the identification of the concept, but if enough are lost, it must be another concept. This has long been accepted doctrine concerning "theoretical concepts", but it also applies, in varying degree, to all concepts. I agree with Bilgrami that this thesis, or these theses, about conceptual "overlap" have nothing directly to do with norms, and certainly not lexical norms. They tell us nothing about how people should or must speak to be understood. They impose no obligations on speakers; they are descriptive facts about what people do mean by what they say, about what is involved in (correctly) understanding them. This brings me to the second comment. There seems to me no fundamental distinction to be made, so far as norms are concerned, between what I just said about non-logical connections among concepts and the logical properties of the logical constants (whatever we decide those constants are). We can interpret a speaker's truth-functional sentential connectives, for example, only by noting the truth tables that would account for the speaker's verbal behavior. This interpretive practice may seem to hold the speaker to objective norms of good reasoning; in fact, however, it is only one more example of how the speaker's verbal behavior determines what the speaker means. It creates no opportunity for judging the speaker's use of language correct or incorrect. Norms do, of course, enter into our understanding of what people say in many ways. We may have trouble deciding whether someone who generally applies the word "arthritis" as we do, but deviates from time to time, means what we do by the word and disagrees with us about the facts, or is using the word with a slightly non-standard meaning and agrees with us about the facts. Further evidence may resolve the problem, but we should remember that there can be no guarantee of this; asking the speaker helps only to the extent that we already understand many of his other expressions. Ultimately the interpreter's norms come into play in deciding which interpretation is the most reasonable. But reasonable in what respect? There is on the one hand the question whether it is more likely the speaker is using a word in a slightly unusual way or is mistaken

Reply to Akeel Bilgrami

147

about some cases of what we would call arthritis. But on the other hand • there may (perhaps also) be a question about the reasoning of the speaker; we may find it difficult to avoid an interpretation which convicts the speaker of an offense against the (our) norms of good reasoning. Should we now revise the interpretation? Our norms as interpreters are unavoidably involved in the decision. But though our norms guide our judgments of the reasoning ability of a speaker and these judgments in turn affect our understanding of what the speaker means, none of these normative considerations should tempt us to say that the speaker has failed to follow the norms of language, for as Bilgrami insists, there are no such norms.

Uses of Mistakes JOACHIM SCHULTE

Jonathan Bennett, in his book Linguistic Behaviour, mentions the possibility of an occurrence which he regards as highly unlikely and which he proceeds to illustrate in the following way: I doubt if I have ever been present when a speaker did something like shouting 'Water!' as a warning of fire, knowing what 'Water!' means and knowing that his hearers also knew, but thinking that they would expect him to give to 'Water!' the normal meaning of 'Fire!'1

This remark is quoted on the first page of Davidson's paper "A Nice Derangement of Epitaphs", and Davidson there expresses his disagreement with Bennett by claiming that such things, which in Bennett's view happen but rarely, in reality "happen all the time". "In fact", Davidson continues, "if the conditions are generalized in a natural way, the phenomenon is ubiquitous". 2 What, however, is the kind of thing which Bennett regards as so extremely uncommon while Davidson thinks that it is virtually universal? In order to answer this question we should have to find out what exactly it is that the case described by Bennett is supposed to be an example of. Davidson mentions the remark from Bennett's book in the context of deliberate puns and malapropisms of various kinds. He quotes a piece from an article by Mark Singer, in which a large number of cliches and set phrases are misused in such a way that we none the less understand what the writer wishes to say, and besides Sheridan's Mrs. Malaprop he mentions Shakespeare, Lewis Carroll and James Joyce. Examples of malapropisms are the phrase in the title of Davidson's paper ("a nice derangement of epitaphs", which in Mrs. Malaprop's use should be read as meaning "a nice arrangement of epithets") or the joke about people needing a few laughs to break up the monogamy. What all the cases intended by Davidson have in common is, according to him, that "the hearer has no trouble understanding the speaker in the way the speaker intends". 3 1 2

3

Jonathan Bennett, Linguistic Behaviour, Cambridge University Press, 1976, p. 186. Donald Davidson, "A Nice Derangement of Epitaphs", in R. E. Grandy and R. Warner (eds.), Philosophical Grounds of Rationality, Oxford: Clarendon, 1986. Davidson, op. cit., p. 158.

150

Joachim Schulte

How it is possible that the hearer will find it easy enough to understand a speaker uttering malapropisms, puns, or jokes of the kind Davidson cites is explained by pointing out that in such cases the hearer will notice that the standard reading of the expressions used by the speaker will not do as a plausible interpretation of his words and that, in order to make sense of the speaker's utterance, it will be helpful to try a different interpretation suggested by similarity of sound or contextual features or likelihood of intention. The hearer is warned by the palpable inappropriateness of the speaker's utterance, and the linguistic or non-linguistic context contains a sufficient number of indicative elements to enable the hearer to guess a natural intended meaning. In short, the type of case envisaged by Davidson is one of successful or fairly successful communication. And it is for this reason that he can go on to generalize and argue as he does that this type of case speaks against the widely held belief that the competence of speaker and hearer is, as Davidson puts it, "learned in advance of occasions of interpretation and is conventional in character". 4 I wonder, however, if it is all that easy to generalize here. It is after all not obvious that the type of case intended by Davidson can be described and delimited without running into more problems than expected, nor is it clear that all, or at least the majority of, the cases which can adequately be grouped together are cases of successful communication. Of course, it would be foolish to try to throw any doubt on the claim that there is plenty of successful communication involving malapropisms, puns, etc. All I want to say for the time being is that at first blush I fail to see the natural boundaries of the class of cases Davidson wishes to use as a basis of his argument in his paper on malapropisms. More specific questions concern the case quoted from Bennett's book. Is that really an example which Davidson can usefully employ to promote the persuasiveness of his argument? And what exactly is it an example of? The case described by Bennett is one where the speaker shouts the word "Water!" as a warning of fire, although he knows the meaning of the expression "Water!" and knows that his hearers also know it. The speaker none the less thinks that his hearers expect him to give to "Water!" the normal meaning of "Fire!". There are at least two but probably more ways of understanding this description, depending on what you take the crucial word "know" to mean. If you take "know" as implying awareness at the moment of utterance, then the whole description becomes a piece of incoherent nonsense because it does not fit the situation. It is one thing to be fully aware of the meaning of the word "monogamy", to use it in jest and to think that your audience will expect you to give it the normal 4

Davidson, op. cit., p. 161.

Uses of Mistakes

151

meaning of "monotony", while it is a completely different thing to be fully aware of the meaning of the word "water" and still use it in a dangerous or even disastrous situation thinking that your audience will expect you to give it the normal meaning of "fire". So this reading is to be excluded as fatuous. If on the other hand we choose to read "know" in a purely dispositional sense, meaning roughly that the speaker will in normal situations be able to use and understand the word "water", the case as described is in a way too easy to make sense of because it is the non-linguistic surroundings of this utterance which determine the speaker's assumptions as to what he will be taken to mean as well as the audience's reactions. If a man comes running from a burning house gesticulating in a certain way, practically anything he may shout will be understood as meaning "Fire!". If, on the other hand, he slowly passed you in the street und said in a normal tone of voice "water", neither he nor you would have any reason to suppose that he is using this word to mean "fire". The case as described by Bennett serves no recognizable philosophical purpose; and this is due to the exceptional character of the situation imagined. But if we try to adapt it to more normal circumstances, we shall run into difficulties. As long as we stick to one-word sentences and normal words like "water", the overall situation must be such that the meaning of possible utterances is to a large extent determined by it and its peculiarities if speaker or hearer are to have the slightest reason to treat an utterance of the word in question as meaning something different from what it usually means. And if the situation is thus, the interpretation to be assigned to the word will exclusively depend on non-linguistic features of the specific situation and will hence be irrelevant to considerations of meaning (except in so far as it serves as an example of the possibility of this sort of dependence). If however we allow more complexity and more normal conditions of utterance, we are, I think, faced with quite a few possibilities. Tentatively one may want to distinguish three types of cases. The first is the one which Davidson is interested in. This possibility comprises situations where features of the linguistic or non-linguistic context indicate that a certain expression is not to be taken in the standard sense. In such cases we speak of irony, malapropisms, jokes, etc. The second is that of a straightforward mistake or error. And the third is that of wilful obstinacy in continuing to use an expression which is not the one standardly used to convey the intended meaning. I shall now look into the last two possibilities in order to see if they yield anything useful. Now, what about straightforward mistakes? Here I want to turn to a philosopher who was greatly interested in mistakes of all kinds and who would probably not have been satisfied with such a broad category as the

152

Joachim Schulte

one alluded to here. The philosopher I mean is Austin, who often urges us to choose our examples well and to describe them adequately. After our complete failure to make any but negative use of Bennett's case no advice could appear more apposite. Austin writes in "A Plea for Excuses": By imagining cases with vividness and fullness we should be able to decide in which precise terms to describe, say, Miss Plimsoll's action in writing, so carefully, 'DAIRY' on her fine new book: we should be able to distinguish between sheer, mere, pure, and simple mistake or inadvertance. 5

Let us follow Austin's lead and have a look at the vividly and fully described case of Miss Plimsoll. The story comes from Harold Nicolson's book Some People. It is told from the point of view of a young boy whose governess, Miss Plimsoll, receives from the boy's family, which is staying at the British embassy at Tangier, a birthday-present in the shape of a "large quarto manuscript book bound in white vellum with leather tags". This book, writes Nicolson, was presented to Miss Plimsoll, who became ecstatic. 'Now,' she said, clasping the volume to her flat little bosom, Ί really shall write a diary.' She loved that book. She crooned over it. And one day she produced from the English shop in the town a paint-brush, a mapping nib, and a bottle of Indian ink. She was adept at calligraphy, and across the top she wrote in old English characters 'Edith Plimsoll, Morocco, 1899.' Then in the centre of the vellum cover, and in larger and even older English lettering, she sketched out in pencil the word Diary. I watched her doing it. She measured each letter with the ruler, and there was a great deal of india-rubber needed, and then she brushed and blew away the filings that the indiarubber had left. Then very carefully she uncorked the Indian ink and began outlining the letters with the paint-brush. Her tongue peeped out beyond her pale little lips, following, now that careful down-stroke, now that up-stroke, which had to be more careful still. I myself had noticed, when the design was yet only in its pencil stage, that she had written DAIRY instead of DIARY. But I held my peace. With a fearful joy I saw the Indian ink descend upon that premature A, pass on to that belated I. The outlines were finished: the brush was carefully wiped: the mapping pen was dipped (her little finger crooked) into the Indian ink. The letters were then crosshatched. I looked over her shoulder. 'But Miss Plimsoll,' I said, when she had got irretrievably to the R, 'you don't spell "diary" like that.' 'Oh ...' she exclaimed. 'Oh, oh!' At this I snatched the book from her and danced from the room. I burst into the Chancery, waving it above my head. 'Look!' I shouted, 'this is how Miss Plimsoll spells Diary!' Everybody, including the native clerk, was thoroughly amused. My triumph became positively orgiac. I danced round the garden shouting 'Dairy! Dairy! Dairy!' All my repressions were suddenly released. As usual Miss Plimsoll dissolved into tears. 6

This, of course, is not the classical speaker-hearer situation but that, I suppose, does not matter. The story can quite easily be modified in such a way that it conforms to whatever speaker-hearer model you prefer. One 5 6

J. L. Austin, Philosophical Papers, Oxford University Press, second edition 1970, p. 198. Harold Nicolson, Some People, Oxford University Press, 1983, pp. 15f.

Uses of Mistakes

153

thing, however, is clear: Miss Plimsoll evidently means to write "diary" but in fact she writes "dairy". She does not, however, believe for a second that "dairy" means a diary. She is perfectly aware that "dairy" means dairy. As soon as she realizes what she has written she sees that she has made a mistake, and the realization of this mistake makes her suffer. It is not obvious if her mistake is to be counted as a sheer, mere, pure or simple mistake. But the question of a classification in accordance with these headings proposed by Austin need not be pressed here. What we should try to discover is whether the story of Miss Plimsoll indicates any distinguishing features of mistakes which may help us to see a general difference between our supposed second category, that of straightforward mistakes, and our first category, that of malapropisms, irony, etc. A person who makes a mistake by using a linguistic expression which is not the one he intends to employ does not use this expression as a wrong or inappropriate expression. Such a person is — as Miss Plimsoll's case makes sufficiently clear — simply not aware of the fact that he is using this expression. On the contrary, this person believes that he is using a different phrase from the one he is actually using. A speaker who makes a deliberate pun or expresses himself ironically or plays with words knows what he is doing; in fact, doing it depends on the speaker's paying a lot of attention to the expressions he is using. But, of course, Miss Plimsoll too is paying plenty of attention to what she is doing. That is part of what makes her look so pathetic. She takes such pains, the whole operation takes so much time, and still she makes such an obvious mistake. She certainly believes that she knows what she is doing. Most of us who have ever seen a paper or a book through its various stages of preparation, revision, proof-reading etc. will certainly remember their chagrin upon discovering in the printed text some stupid blunder which has survived in spite of all our efforts. It is surely a mistake but we surely knew — or thought we knew — what we were doing. So the attention paid by the speaker (or writer) to his words cannot be the distinguishing mark of mistakes as opposed to puns, irony, etc. On reflexion, however, we shall soon notice that this is no way to find any differences between the several categories we thought we might be able to distinguish. For you may quite easily adapt the Plimsoll story to cases of the first category. You may tell a very similar tale about someone taking great pains over an attempt at being ironical or making a subtle pun and failing dismally. This is even more clearly so if you take malapropisms into account. Think of a Miss Plimsoll who wants to show her brilliance by using uncommon words — and then she confuses epitaphs and epithets. She will no doubt dissolve into tears again. On the other hand, it is quite possible to write "dairy" instead of "diary" without having to suffer for it in the way Miss Plimsoll did. If you read

154

Joachim Schulte

a letter or an article in the newspaper where these two words are mixed up, you will tacitly and without any bother supply the right word and read on. The point is simply that in these cases the speaker or writer gets away with it, just as Mrs. Malaprop got away with it, whereas Miss Plimsoll did not. Davidson makes a lot of this possibility of our getting away with it. Actually, this is exactly the point of introducing Mrs. Malaprop because her case illustrates the possibility of getting across what one wants to get across in spite of using the wrong words as well as our capacity of understanding a speaker's meaning even if he puts things in an awkward, misleading or straightforwardly mistaken way. This we may grant, but we are still wondering about Miss Plimsoll and the likes of her. The outstanding characteristic of her performance is that her intentions are thwarted but, curiously enough, it is not her intention to communicate something which is thwarted. Young Harold Nicolson, the native clerk, and all the other people at the British embassy understand extremely well what she wants to say. It is because they understand it that they can make fun of her, as the gulf between what she wished to achieve and what she actually managed to accomplish is so strikingly manifest. Miss Plimsoll does not get away with it, not because she does not succeed in getting her meaning across, but because her mistake is not only noticed but made use of. It is used by her audience in order to make her appear ridiculous. The people who make use of mistakes of this kind are typically not those who make the mistakes but those who notice them. They use them in order to gain an advantage over those who make them: to frustrate their plans, to punish them, or to poke fun at them, as in the case of Miss Plimsoll. If we now turn to compare the example of Miss Plimsoll with that of Mrs. Malaprop we shall immediately see that the same thing could easily have happened to the latter. Davidson is interested in her because as a matter of fictional fact she gets away with her outrageous utterances. But we must not overlook that Mrs. Malaprop too could be made an object of ridicule and thus come to see that she has made a fool of herself. In that event she would probably share Miss Plimsoll's lot and suffer for what she has said. These comparisons should alert us to a fact which I shall only mention in passing, although it could surely be exploited to some effect. The fact I have in mind is this: that Mrs. Malaprop's getting away with her utterance and Miss Plimsoll's not getting away with hers has not got all that much to do with communication or success in communication — unless communication is understood in the minimal sense of getting on with our exchanges of linguistic expressions. We can let Mrs. Malaprop get away with it even if we only have a vague idea of what she means. And on the

Uses of Mistakes

155

other hand we do not let Miss Plimsoll get away with her performance inasmuch as we use her mistake in such a way that as a result she will be sorry for having said or written what she did say or write, and we do not let her get away with it although we understand her meaning perfectly. Our not letting her get away with it does not depend on whether or not she succeeds in communicating what she wants to communicate. From these examples and comparisons we may surely draw the conclusion that no real boundary between our alleged first and second categories exists. The difference in the reactions to Miss Plimsoll's and Mrs. Malaprop's utterances is entirely due to their audiences deciding, in the second case, to let the speaker get away with her blunder and, in the first case, not to let her get away with it. Our only reason for hesitating to call Mrs. Malaprop's utterance a mistake is the fact that her audience does not make use of her mistake in the way Miss Plimsoll's audience makes use of ber mistake. And that reason may on second thoughts, come to look a little too feeble in the face of mounting evidence in favour of calling it — perhaps not a sheer, mere, pure or simple mistake — but a mistake nevertheless. Now, in order to complete my short inquiry, I shall briefly examine the question whether our third tentative category really has a life of its own or whether it too will collapse into that shapeless mixture constituting the first — and in that event possibly only — category. This time I shall turn for help, not to John Austin, but to Jane Austen. In Northanger Abbey her heroine, Catherine Morland, hears her host, General Tilney, asking his son Henry, who lives in a house of his own in the neighbourhood, if he (the General) together with his guest and his daughter could come over to see him (Henry) and have dinner at his house. The General explicitly tells his son: 'You are not to put yourself at all out of your way. Whatever you may happen to have in the house will be enough. I think I can answer for the young ladies making allowances for a bachelor's table.' The next thing Catherine hears is that Henry is about to leave Northanger Abbey two days earlier than planned. When she asks him why he is leaving so soon he answers: 'Why! — How can you ask the question? — Because no time is to be lost in frightening my old housekeeper out of her wits, — because I must go and prepare a dinner for you to be sure.' 'Oh! not seriously!' 'Aye, and sadly too — for I had much rather stay.' 'But how can you think of such a thing, after what the General said? When he so particularly desired you not to give yourself any trouble, because any thing would do.'

In reply to this Henry merely smiles and leaves to go about his preparations. The author continues:

156

Joachim Schulte ... it being at any time a much simpler operation to Catherine to doubt her own judgment than Henry's, she was very soon obliged to give him credit for being right, however disagreeable to her his going. But the inexplicability of the General's conduct dwelt much on her thoughts. That he was very particular in his eating, she had, by her own unassisted observation, already discovered; but why he should say one thing so positively, and mean another all the while, was most unaccountable! How were people, at that rate, to be understood! Who but Henry could have been aware of what his father was at?7

Simplifying a little, the situation appears to be the following. General Tilney says one thing and he says it very clearly. He does not confuse words, he does not use inappropriate expressions, but still he really does mean something different — in fact, more or less the opposite — of what his words would in ordinary cases be taken to mean. And what is more, he also gets away with it, as is shown by Henry's immediate action. An attentive hearer like Catherine has no means of guessing at the General's real meaning: Neither the linguistic nor the non-linguistic context provide any clue which could alert the hearer to the fact that the General means something different from what he is saying. For this reason Catherine is right when she arrives at the conclusion that only Henry "could have been aware of what his father was at". No doubt such cases are quite common. We all are familiar with domestic tyrants who can rely on being understood the way they want to be understood by the people concerned even though their words indicate something different. Here the correct understanding depends on the hearer's being familiar with the speaker's personality and his ways. But even in a situation where the hearer has had no personal acquaintance with the speaker similar things can happen. Thus it may for example be that the hearer guesses from what he knows about the otherwise unfamiliar speaker's social status that he will expect to be obliged in certain respects although he explicitly avows indifference. Not all speakers of this type get away with it in all situations. The family tyrant may be disobeyed, the party official's declaration may be taken literally by his subordinate, and so their intentions may be frustrated. In this respect the General Tilneys of this world are in the same position as the Mrs. Malaprops and the Miss Plimsolls. One might think that in cases like those of the General language plays virtually no role; after all, the hearer is expected to react in a certain way, no matter what the speaker will say. But that is not really so; the hearer would not know what to do if the words did not indicate the direction of the response expected from him. It is only through the specific mention of dinner that Henry can learn that he is expected to provide a particularly good one even though he is told that a frugal one would suffice. Had the General mentioned horses 7

Jane Austen, Northanger

Abbey, Harmondsworth: Penguin, 1985, pp. 21 If. (chapter 26).

Uses of Mistakes

157

instead, Henry would have known that he had to do something about horses. But language is still needed to get it across to the hearer what it is that he is expected to do something about. Looking at the situation this way we are surely not tempted to speak of a mistake on the General's part. If at all, it would be Henry who could make a mistake by not reacting as he is expected to react. But if, on the other hand, Henry learned to stand up to his father and simply ignored his expectations and then managed to put up with the General's indignation at having his implicit intentions thwarted, then the mistake would be the General's because he does not say what he means, while Henry could make use of this mistake to further his own interests. If you look at the broad outlines of the cases we have considered it appears that there is not much difference between General Tilney, Mrs. Malaprop and Miss Plimsoll. What they have in common is that they all, in one way or another, do not say what they mean. Two of them get away with it, one does not — but it could, as we have seen, easily have been the other way around. Our inclination to speak of a mistake varies according to the speaker's good or bad luck in getting or not getting away with it. What we still have not found is an independent criterion which would allow us to identify or distinguish different types of cases belonging to the group Davidson has used for his argument. These cases belonged to a rag-bag to start with and that, I am afraid, is where they still are.

Donald Davidson: Reply to Joachim Schulte I like the examples in this paper. I wish I had thought of some of them myself. And I agree with Joachim Schulte that for many purposes it is important that examples be given in detail, especially if the purpose at hand requires making precise distinctions and applying them. But is Schulte right in thinking my thesis in "A Nice Derangement of Epitaphs" is threatened because I was not careful to distinguish cases? I can not see how. My thesis was that it is common for people to say things which are not in any sense "standard" expressions in the language they are speaking, and yet communication is not impaired. To make this point I didn't need to be precise about what is standard in a language — in fact I suggested this couldn't be done, and for this reason claimed there are no natural languages if one means by a natural language a system of signs with a precisely defined syntax and semantics. Nor was it important for my thesis to draw a sharp boundary between cases where it makes sense to speak of mistakes and cases where it doesn't. All I needed was to describe a variety of situations most people would accept as cases where communication succeeded, i. e. a person was interpreted as he or she intended to be interpreted, and yet where there was no reason to think that the interpreter was prepared to interpret the speaker in that way in advance. I was especially careful to emphasize the fact that there are many sorts of cases in which this happens, and that there is no specifying in general how we manage correct interpretations in such cases, since we draw upon just about everything we have — we employ our wit, imagination, sympathy, knowledge of the world, insight into motives, and acquaintance with languages not at the moment being spoken. While I had no reason, in "A Nice Derangement of Epitaphs", to make fine distinctions within the vast array of examples where communication is not damaged by non-standard utterances (however "standard" is construed), I did want at least a crude distinction between such cases and ones which do not raise this issue, and to this end I devoted two full pages to distinguishing what I called "first meaning" from other things that have also been called meaning, such as what is sometimes called speaker's meaning, or (by Grice) implicature. Curiously, this essential distinction is lost on Schulte. As a result, his pleasant example drawn from Jane Austin's Northanger Abbey is simply irrelevant to my thesis, since it

Reply to Joachim Schulte

159

is entirely concerned with an intended conversational implicature which fails. The words are intended to be given their "standard" interpretation, and they are given the intended interpretation. What happens after that was beyond the scope of my discussion. Schulte's other detailed example, that of Miss Plimsoll and the misspelled word, is interesting because although she knows how to spell "diary" she intentionally writes an "a" and then an "i" after the "d". This is something (not exactly, of course) like a slip of the tongue, and it certainly does illustrate the sort of phenomenon I have in mind. But Schulte must have misunderstood much of "A Nice Derangement", since he considers this a case of not "getting away with it". I used this phrase for cases where the speaker (or writer) is interpreted in the way she intended to be interpreted. I was limiting the relevant level of interpretation to first meaning, and on this level, Miss Plimsoll got away with her "dairy"; if she had not, she would not have been an object of fun and scorn. Many levels of intention are involved in every speech act, as Austin was the first to emphasize, and every level is subject to potential interpretation. Getting away with it, as I used the phrase, was not a matter of being interpreted as one intended on every level, but only on the level of the literal meanings of words. (I did, however, give a non-standard meaning to the notion of the literal; hence my neologism, "first meaning"). Finally, I should thank Schulte for inadvertently supplying me with one more example of a non-standard expression which was (I think) correctly understood: in my copy of his paper I read, "[H]e slowly passed you in the street und said in a normal tone of voice 'water'".

"What Metaphors mean" and how Metaphors Refer 1 OLIVER R .

SCHOLZ

In the "Introduction" to "Inquiries into Truth and Interpretation" Davidson writes: No discussion of theories of meaning can fail to take account of the limits of application of such theories. The scope must be broad enough to provide an insight into how language can serve our endless purposes, but restricted enough to be amenable to serious systematization. (...) Essay 17, 'What metaphors mean', is mainly devoted to the thesis that we explain what words in metaphor do only by supposing they have the same meanings they do in non-figurative contexts. We lose our ability to account for metaphor, as well as rule out all hope of responsible theory, if we posit metaphorical meanings, (xix) 2

What I have to say about "What metaphors mean" can best be seen as a bundle of provocative questions that arose from the attempt to understand these remarks and, of course, the article itself. It will be convenient to list some of the questions, in advance: (1) Is the scope of Davidson's theory of meaning really broad enough to provide an insight into how language can serve our endless purposes when it comes to non-literal discourse, especially metaphor? (2) Do we really lose our ability to account for metaphors if we ascribe semantical properties (truth and reference, if not meaning) to them, which are additional to literal meaning? (3) Is the way Davidson deals with metaphors in his essay "What metaphors mean" the best way to deal with them in a broadly Davidsonian framework? But let me start with some exegesis. The primary purpose of the essay "What metaphors mean" is a pro domo purpose. Davidson tries to show that

1

2

I want to thank Donald Davidson, Catherine Z. Elgin and Johannes Brandl for valuable comments. Quotations by simple page numbers are from Davidson 1984.

162

Oliver R. Scholz

in dealing with metaphor his project in semantics is not confronted with a special problem (or a problem at all). 3 The central thesis is to be seen in connection with the pro domo objective (namely, to show that metaphor does not endanger his semantic program). It runs: "metaphors mean what the words, in their most literal interpretation mean, and nothing more" (245). Since Davidson provides us with a theory of literal meaning, and metaphors do not have anything especially semantic about them (beyond literal meaning), they can be no challenge to Davidson's project. The crucial question is whether Davidson succeeds in making plausible that all that is special about metaphor does not concern their semantical properties but only "what they are used to do" (247) in a special sense of the ambiguous word "use". To assess this question I will discuss some of Davidson's "limited positive claims" (247) and point out some of their shortcomings (I). After that I will sketch an account of metaphorical truth and reference that promises to explain more about what we particularly want to know about metaphor (II). Finally, I will indicate (somewhat vaguely I'm afraid) how to deal with metaphor in a Davidsonian theory in a better way (III).

I. Davidson proceeds in his article mainly by criticising theories assuming special semantic properties for metaphors. (In addition, he has some, as it were, direct arguments for his main thesis; later on, I will comment briefly on one of them.) He argues most of the time against (a) special (metaphorical) word meanings and (b) special (metaphorical) sentence meanings. Those are the main targets. Moreover, he also rejects accounts that rely on the acceptance of (c) metaphorical truths and (d) metaphorical reference (or application, extension). More specifically, Davidson attacks the following positions: (i) in metaphor certain words take on new, or 'extended', meanings; (ii) metaphor is a kind of ambiguity; (iii) metaphor is a special kind of ambiguity — similar to that operative in puns; (iv) Fregean views of metaphor; (v) the figurative meaning of a metaphor is the literal meaning of the corresponding simile; (vi) a metaphor is an elliptical simile. I agree 3

Thus, "What metaphors mean" belongs in the same line as, e. g., "Moods and Performances", where Davidson suggests that whereas grammatical mood is of concern to a theory of what words mean, the force of utterances is beyond the reach of a systematic theory of meaning.

"What Metaphors mean" and how Metaphors Refer

163

with most of Davidson's criticisms of these approaches; but in the second part of my talk I will defend a variant of the ambiguity view. Let me now comment briefly on one of Davidson's direct arguments against the necessity of postulating special metaphorical meanings. It is an analogy argument: There are many devices (e. g., similes) that serve to direct our attention to similarities without there being any need to postulate special second meanings (255 f.). What words do with their literal meaning in such devices must be possible for them to do in metaphor (256). One problem with this line of argument is the following: To the extent that it is successful, it becomes less obvious what is special about metaphors. True enough, metaphor is not the only device that is able to invite us to search for similarities between heterogeneous things. As Goodman has emphasized, this is obvious also from simpler examples than the ones cited by Davidson (he cites similes and T. S. Eliot's poem "The Hippopotamus"). Nonmetaphorical sentences of the form "Compare χ with y!" or "x has important features in common with y" can fulfil this function (cp. Goodman 1984:74). But then the capacity to perform such a function and to cause the corresponding effects do not seem to be defining properties of metaphor. Let us turn now to some of the "limited positive claims" under the following headings: USE: "metaphor belongs exclusively to the domain of use" (247). TRIGGERS: "most metaphorical sentences are patently false" (258). — "Patent falsity is the usual case with metaphor, but on occasion patent truth will do as well" (258). EFFECTS: "A metaphor makes us attend to some likeness, often a novel or surprising likeness, between two or more things" (247); "a metaphor merely nudges us into noting" something (253); "there is no limit to what a metaphor calls to our attention" (263); "much of what we are caused to notice is notpropositional in character" (263); "What we notice or see is not, in general, propositional in character" (263). UNDERSTANDING: "understanding a metaphor is as much a creative endeavour as making a metaphor, and as little guided by rules" (245). SUCCESS: "A metaphor implies a kind and degree of artistic success; there are no unsuccessful metaphors" (245). In addition, Davidson suggests some analogies between metaphors and such diverse phenomena as (a) dreams or dreamwork (245, 262); (b) jokes (245, 262); (c) pictures (262, 263); (d) a bump on the head (262); (e) seeing as (263). Let us begin with some comments on TRIGGERS: Every adequate account of metaphor has to provide an answer to the question how we recognize or identify metaphors (and how speakers manage to signal that the words should be taken metaphorically). Which clues in the sentence or in the context trigger a metaphorical re-interpretation?

164

Oliver R. Scholz

Davidson mentions in the first place only literal falsity, especially patent falsity. Later on he adds patent truth, but emphasizes that "patent falsity is the usual case with metaphor" (258). More to the point is his remark in a footnote: "Since the negation of a metaphor seems always to be a potential metaphor, there may be as many platitudes among the potential metaphors as there are absurds among the actuals" (258n.l0). Davidson's account of triggering metaphorical interpretations is incomplete in several ways. (A) Although patent falsity (often as a result of some sort of category mistake) and obvious truth are indeed very important clues, they are not the only ones. On occasion, irrelevance (relative to the foregoing topic) will do as well (cp. Scholz 1988: 276). (Β) Another incompleteness is more serious. It should be clear that not every kind of obvious truth or falsehood will be re-interpreted as a figure of speech, much less as a metaphor. Consider "Saddam Hussein is a pacifist". Though this is false, and obviously so, you will feel little inclination to take it as a metaphor. On special occasions you might take it as a bit of sarcastic irony; but more probably, you will come to the conclusion that it is simply false — full stop. (Perhaps, in addition, you will begin to doubt whether the speaker knows what the word "pacifist" means.) The moral is: We still need to make the principles explicit on which competent language users differentiate (1) sheer obvious falsities or platitudes from non-literal uses of words, (2) metaphor from the other forms of non-literal discourse. (Not that the language users consciously employ such principles, but we who reconstruct their ability to identify and interpret metaphors should make them explicit. Simply saying, like Davidson does, "The ordinary meaning in the context of use is odd enough to prompt us to disregard the question of literal truth" (258) is not enough; the question is why we do that, or better what reasons we do have for reacting this way.) If Davidson's account on how non-literal interpretations are triggered is incomplete, still less has been said so far about how the hearer manages to exclude non-literal readings other than metaphorical ones (e. g., ironical readings) and how he might work out or, at least, restrict the specific metaphorical reading. Hints like "a metaphor makes us attend to some likeness between two or more things" (247) are of little help here. We mentioned above that this is not an exclusive privilege of metaphors. Now, we have to point out additional difficulties. As Davidson accentuates himself (in his criticism of the two different "simile theories" of metaphor): so long as likeness is not refined "everything is like everything, and in endless ways" (254). Although this remark is true enough (in a sense, at least), it makes clear at the same time why Davidson's appeal to likeness (similarity) in his positive account is problematic.

"What Metaphors mean" and how Metaphors Refer

165

Aside from the well-known fundamental problems with the "concept" of similarity — and even if we could define an appropriate, non-trivial similarity concept —, there would remain insuperable difficulties. Metaphor cannot, in general, be a matter of factual likenesses between two or more objects. (1) There is a problem with "factual likeness": Very many metaphors rely on false stereotypes, e. g. "Richard is a gorilla". This metaphor will nudge us into noting — among other things — that Richard is brutal. Now biologists found out that that gorillas are by no means brutal animals. Accordingly, Richard and the gorillas are not similar in the relevant respect. A different account of metaphorical likening is needed. (Cp. Searle 1979: 89 & 108.) (2) There is a problem with "between objects". Think of metaphors containing fictional terms, such as "Richard is a Don Juan". There being no (literal) Don Juan the metaphor cannot attend us to a likeness between Richard and Don Juan. As Israel Scheffler stresses: metaphor is not wholly objectual in outlook. Its routes of comparison are often circuitious, touching not only on the objects in question and their features, but also on various representations of these objects. [...] We live, after all, in a world of symbols as well as other objects. Our view of objects and our knowledge of their representations serve alike as resources for interpretation. (Scheffler 1986:394)

(3) Likeness (similarity, resemblance) is symmetrical, metaphors, however, are directional in a clear sense. "Richard is a gorilla" should prompt an insight about Richard, not about gorillas. This directionality has to be explained somehow. In this respect, Davidson's analogy between metaphor and seeing-as might help, since seeing-as has the requisite asymmetrical character. But, the difficulty with this analogy is that an unexplicated notion of "seeing-as" is not any clearer than metaphor itself. Now, one last comment about the topics USE/EFFECTS/UNDERSTANDING: Davidson says: "metaphor belongs exclusively to the domain of use" (247). Unfortunately, "use" is a highly ambiguous term; and I am not sure about whether Davidson employs it in a uniform fashion throughout "What metaphors mean". Most of the time, however, he seems to refer to what Austin called the perlocutionary act, which is a matter of non-conventional effects caused by the utterance. Now, as briefly noted above, the causing of such effects (such as noticing a likeness) is hardly restricted to metaphors (even if the non-propositional aspect is added); accordingly, this feature cannot be used to characterize metaphor. But, there might even be a more serious problem. It is hard to see how, on Davidson's account, there could be such a thing as understanding a metaphor (despite the fact that he repeatedly uses this phrase). If, on the one hand, "understanding" would be taken as "grasping the meaning of', then this sort of understanding could only consist in grasping the literal meanings

166

Oliver R. Scholz

of the words, for that's all the meaning there is, according to Davidson. Thus, in this case we would have understanding, but not understanding of a metaphor as such. In addition, Davidson tells us, the words with their literal meanings have certain effects on the hearers when used in a metaphor, effects comparable to those which a bump on the head (262) — or, we can add, a drug — might have, too. 4 Now, such effects would be goings-on caused by the metaphor as such, but goings-on that would hardly deserve the label "understanding". (I know that there are many more senses of the term "understanding", but I see none especially suited for Davidson's speaking of an understanding of metaphor — given what he says on the working of metaphor.) Let us take stock, then: Davidson's positive claims about metaphor are insufficient in several respects: The account of how we recognize metaphors is incomplete; what Davidson says about understanding metaphors seems to be even more problematic. The success of metaphor in communication and in science remains an enigma, a miracle. Our astonishing competence in recognizing and interpreting metaphors — and the high degree of convergence in our verdicts — has to be explained somehow in an adequate theory of metaphor. Davidson's account fails to achieve this. I know two approaches that promise to provide a more adequate account of the identification and interpretation of metaphors. (1) A Gricean theory of conversational implicatures where metaphors are conceived as a special case of exploiting one or more of the so-called conversational maxims. (Grice himself makes the mistake to mention metaphors only in connection with exploitations of the first maxim of quality. Quantity and relevance, however, have to be considered, too.) Since I have sketched a Gricean theory of metaphor in an earlier paper (Scholz 1988), I focus on a second approach: (2) The extensionalist, or even nominalist, theory of symbols and reference (Goodman, Scheffler, Elgin), I want to sketch now. 5

II. It has the following remarkable features: — it does not assume the existence of (coded, stable) metaphorical meanings (not even the existence of stable literal meanings); 4 5

Cp. also K ü n n e ' s critique of Davidson's positive account in Kiinne 1983: 192 f.. Cp. G o o d m a n 1968: 6 8 - 8 5 & 1984: 7 1 - 7 7 ; Scheffler 1979 & 1986; Elgin 1983: 5 9 - 7 0 & 1 4 6 - 1 5 4 ; Elgin/Scheffler 1987.

"What Metaphors mean" and how Metaphors Refer

167

— it assigns the concepts of "metaphorical truth" and "metaphorical reference" (application/extension) a central place in the theory of metaphor, thus accentuating semantic (and cognitive) features of metaphor; — it explains metaphor as a kind of ambiguity while making clear how it differs from simple ambiguity; — it illuminates ways in which "the primary or original meanings of words remain active in their metaphorical setting" (something Davidson himself demands, cp. 249); — more than that, it illuminates the ways in which the original semantic properties of words guide or constrain our interpretation of metaphors; — it provides a theory of metaphorical likening that does not rely on an unanalysed notion of likeness or similarity; — it is able to deal with certain problematic cases (most theories are unable to accomodate) —most notably: (a) metaphors containing fictional terms (b) metaphors based on false stereotypes. Because of limitations of time, I cannot demonstrate all the varied virtues of this approach; so I will concentrate on a few of those that may be most relevant to a comparison with Davidson's proposals. Let's begin with metaphorical truth. Consider once again Davidson's warning: "It is no help in explaining how words work in metaphor to posit metaphorical or figurative meanings, or special kinds of poetic and metaphorical truth" (247). To be sure, views that posit two or more different species or kinds of truth (with fanciful names like "poetic truth", "aesthetic truth" etc.) are justly criticised. But aside from these pitiful evasions, there seems to be a perfectly straightforward and harmless sense of "metaphorical truth". The adjective "metaphorical" in the construction "metaphorical truth" need not be taken as functioning like, e. g., "red" in "red apple". Accordingly, the phrase "metaphorical truth" need not mean that the truth of the sentence is of a special kind, namely metaphorical, but only that the sentence taken metaphorically is true (cp. Goodman 1984:71; Scheffler 1986:392). In addition, let me dwell a bit on how intuitive the idea is that metaphors are truth candidates (cp. also Black 1979): — People do hold sentences containing metaphors true. — You can use sentences containing metaphors to make genuine assertions. (People can agree or disagree with these assertions.) — You can draw certain consequences from such statements (some of them containing metaphors as well, some of them don't). — Reasons can be offered for and against them. Thus, I think it is very intuitive to admit that at least some (correct) metaphors are true in a normal sense. (In the third part of my paper, I

168

Oliver R. Scholz

will add the suspicion that the denial of metaphorical truth is alien to the spirit of an empirical theory of interpretation.) If the harmless sense of metaphorical truth sketched above is granted, we are now in a position to pave the way for the idea of metaphorical application or reference. On this behalf, it is crucial to take notice of the fact that metaphorical truth does often coexist with literal falsity without there being a logical contradiction involved. Take an example: (S) Davidson's articles are gold mines. Taken literally, this is false, even obviously false: Articles are not, and cannot be, gold mines. Nevertheless, taken metaphorically, it is true. Thus, we seem to be committed to "Davidson's articles are gold mines, and they are not gold mines". But, we will not plead guilty to entertaining contradictory beliefs about Davidon's articles (or at least not in this respect). The way out is obvious, though. "Gold mines" applies to different things (or in different ways) in the two occurrences. In its original (literal) application, the predicate "gold mine" serves to sort out geological formations. In a different application (why not call it "metaphorical"), "gold mine" is used to classify intellectual products. Calling an article convincing or a gold mine seem to be simply different but equally legitimate ways of classifying it. Metaphor and ambiguity are, then, akin in that ambiguous terms likewise have two or more applications. The crucial question is, of course, how metaphor differs from sheer ambiguity. Roughly, in that a literal application precedes and, more importantly, influences or guides a correlative metaphorical application. Davidson emphasized, correctly, I think, that "an adequate account of metaphor must allow that the primary or original meanings of words remain active in their metaphorical setting" (249). In addition, he argues that the postulating of special metaphorical meanings (or coded contents) will not help in explaining how metaphors work. But, he does little — too little, in my opinion — in illuminating the ways in which the literal meanings of the words guide or constrain our interpretation of metaphors. In this connection, it is essential to take notice of the fact that predicates or other labels do not function in isolation but as members of families of alternatives — families which sort the objects in a domain. Let us, following Goodman and Elgin, call such a family a schema. The realm associated with a given schema consists of all the things to which any of the schema's predicates apply. A realm is, then, the aggregate of the ranges of extension of all the labels in a schema. Thus, e. g., the schema consisting of "odd" and "even" is assigned to the realm of integers; the schema consisting of "animal", "vegetable", and "mineral" is assigned to the realm

"What Metaphors mean" and how Metaphors Refer

169

of physical objects; and so on. (A schema applied to a realm is called a system.) (Cp. Goodman 1968: 72; Elgin 1983: 37) Now, metaphor normally involves a transfer of a schema (or of parts of it) to a new realm. On occasion, metaphor can also consist in applying a schema to its home realm in such a way as to effect a novel sorting in it. An example analysed by Catherine Elgin will make all that clearer: The entire scheme need not, of course, be explicitly mapped onto the new realm. In the metaphorical as well as the literal case semantic systems are systems of implicit alternatives. If we label a new social program a 'war on poverty' we introduce a novel network of terms for characterizing our responses to social conditions. With some we are at peace. Among these are allies in the current war (e. g., public education), as well as potential foes (perhaps industrial pollution) with which we are not yet ready to do battle. Our various attempts to alleviate poverty can be described in terms of campaigns, battles, skirmishes, and be evaluated as victories or defeats. These need not be made explicit. By the simple fact of calling our program a war, we have made the descriptive resources of the system to which 'war' belongs available for describing the social realm to which poverty is part. (Elgin 1983:62)

The recognition of this sort of contextual character is essential for explaining the workings of metaphor. It partly explains how the original (i. e., literal) application influences the metaphorical application, how the latter is guided by, or patterned after, the former. As Goodman writes: "A schema may be transported almost anywhere. The choice of territory for invasion is arbitrary; but the operation within that territory is almost never completely so. We may at will apply temperature-predicates to sounds or hues or personalities or to degrees of nearness to a correct answer; but which elements in the chosen realm are warm, or are warmer than others is then very largely determinate. Even where a schema is imposed upon a most unlikely and uncongenial realm, antecedent practice channels the application of the labels." (Goodman 1968:74) To sum up: The new application of a label, as happening in a metaphorical re-classification, is guided partly by its place in the whole schema. What is transferred is not just one label, but a set of labels (something like a lexical field) — together with some of the relations, i. e. part of the structure of the schema. This is why metaphor can have organizational power that makes it attractive not only for the poet but also for the scientist. (A good example is the very efficient metaphor "The mind is a computer".) An additional virtue, I can mention only briefly, is that the GoodmanElgin-Scheffler approach considers the whole referential hierarchy, that is: not only reference to objects, but also reference to labels (descriptions, representations), to labels of labels and so on. This is necessary for handling metaphors relying on false stereotypes ("Henry is a fox", "Richard is a gorilla" etc.) and metaphors containing fictional terms ("He is a Don Juan") where the routes of comparison touch on various representations (texts, fables, myths, pictures, etc.), rather than on objects.

170

Oliver R. Scholz

III. Since it is unlikely that Davidson will become converted to a nominalistic theory of reference and since I promised to come back to my earlier question on whether the way Davidson handles metaphor in "What metaphors mean" is the best way it can be dealt with in a grosso modo Davidsonian framework, I would do well to finish with some more constructive remarks after all this indolent censoriousness. It is convenient to distinguish two elements in or aspects of Davidson's semantic program: a) the emphasis on the empirical character of theories of meaning and interpretation; b) the formal structure provided by a suitably revised Tarskian truth theory. The Tarskian criteria do not seem very hospitable to the semantic treatment of metaphor, at least prima facie. Nevertheless, Catherine Elgin has argued at considerable length that — once you have accepted that metaphorical truth is genuine truth and metaphorical denotation genuine denotation — Tarski's truth definition "applies indifferently to literal and metaphorical sentences" (1983:65η.3). I will not go into the details of this part of the problem of accomodating metaphors in Davidsonian theories of meaning (but cp. Elgin 1983: 65 — 68). Instead, I will focus on the empirical side. In a nutshell, I have two rather vague suggestions: (a) I have the impression that Davidson's denial of metaphorical truth is alien to the spirit of his empirical theory of radical interpretation. (b) I recommend to investigate the roles principles of charity might play in the identification and interpretation of metaphor. I read with approval in the "gold mines" that "All understanding of the speech of another involves radical interpretation" (125) and that Davidson intends to "apply the principle of charity across the board" (xvii). I propose to apply these insights to metaphor. Take again our example "Davidson's articles are gold mines" (S for short) and put yourself into the shoes of the radical interpreter. Could you not detect that I, e. g., hold that sentence to be true when you inspect my behaviour (and without knowing before-hand what S means). Accordingly, you will arrive at the hypothesis that I hold S to be true. Following one version of the principle of charity, you are obliged then to take this as prima facie evidence that the sentence is true (152). On the other hand, considering further utterances containing the noise "gold mine" (and employing a lot of theory), you will arrive at the hypothesis that the standard application of "gold mine" is to (literal) gold mines.

"What Metaphors mean" and how Metaphors Refer

171

Suppose you come back to the interpretation of S later on. In the meantime, you found out what the phrase "Davidson's articles" denotes. Then, you face a problem: what I have said seems to be a patent falsehood; on the other hand, charity obliges you to take it true, and to rationalize my linguistic behaviour. After excluding irony etc. you might try to reinterpret "gold mine" (contextual clues might lead you to an adequate interpretation that makes my statement true and my behaviour rational). This caricature of radical interpretation was intended to have the following moral: A principle of charity may do duty in the process of recognizing metaphors. As we saw above, patent falsity or truth are not sufficient to trigger a metaphorical re-interpretation as long as no principle has been added according to which it seems reasonable to try a re-interpretation. A principle of charity may play a role in explaining why we do not dismiss utterances of "Davidson's articles are gold mines", "Juliet is the sun" or "No man is in island" as sheer nonsense and leave it at that. Principles of charity will also help to restrict the non-literal interpretations. What we are seeking in order to avoid the conclusion that "Juliet is the sun" (asserted by Romeo) is sheer nonsense, is not just any different reading, but a reading that rationalizes the linguistic behaviour of the speaker, a reading that accords best with the principles of charity.

REFERENCES

Black, Max: "How Metaphors Work: A Reply to Donald Davidson", in: Sacks, Sh. (ed.): On Metaphor. Chicago 1979, 1 8 1 - 1 9 2 Davidson, Donald: Inquiries into Truth and Interpretation. Oxford 1984 Elgin, Catherine Z.: With Reference to Reference. Indianapolis 1983 Elgin, Catherine Z./Scheffler, Israel: "Mainsprings of Metaphor", in: The Journal of Philosophy 84 (1987), 3 3 1 - 3 3 5 Goodman, Nelson: Languages of Art. Indianapolis 1968 Goodman, Nelson: Of Mind and Other Matters. Cambridge/Mass. 1984 Künne, Wolfgang: '"Im übertragenen Sinne'": Zur Theorie der Metapher", in: Conceptus 17 (1983), 1 8 1 - 2 0 0 Scheffler, Israel: Beyond the Letter. London 1979 Scheffler, Israel: "Ten Myths of Metaphor", in: Communication & Cognition 19 (1986), 389-394 Scholz, Oliver R.: "Some Issues in the Theory of Metaphor", in: Petöfi, J. S. (ed.): Text and Discourse Constitution. Berlin-New York 1988, 269 — 282 Searle, John R.: Expression and Meaning. Cambridge 1979

Donald Davidson: Reply to Oliver Scholz Oliver Scholz contends that I did not give an adequate or complete account of metaphor. He is right; I didn't attempt to give such an account, nor claim that I had. My paper on metaphor didn't essay a general theory about what triggers a metaphorical interpretation, and it didn't try to make a formal distinction between metaphor and other figurative uses of language. One reason I didn't attempt these tasks is that I thought, and still think, that they are not subjects suitable to the sort of formal treatment of language with which I have been concerned. I also argued that any attempt to give a purely semantic analysis of metaphor would be doomed to leave out an important — perhaps the most important — aspect of metaphor. I had two main purposes in writing that article. The first one was to emphasize what I thought to be the emptiness of the notion of metaphorical meaning and of metaphorical truth. I thought these concepts have no explanatory value. The second thing that I was interested in was the dependance of the metaphorical effect on knowing the literal meaning of the words that are being used, something I thought many theories inadvertently lost to sight. With respect to the second point, Scholz and I seem to be in agreement, but not with respect to the first. However, nothing that I've heard has persuaded me that we can seek a serious formal theory of metaphorical truth. "Davidson's essays are gold mines." Is this metaphorically true or is it metaphorically false? It depends upon how the theory of truth deals with it. To make it true, a second meaning must be given the phrase "gold mines". To participate in a serious theory of truth, this second meaning would have to be made explicit. I see no harm in this: my dictionary actually gives a second meaning to the phrase: "gold mine: a source of great wealth or profit". "Gold mine" is, then, officially ambiguous, and so a sentence containing it can be true on one interpretation and false on another. But what has this to do with metaphor? When a use of a word or phrase has settled to the point of being listed in the dictionary, that use can be treated in a serious account of truth, but that is just because the metaphor has died. Dead metaphors rise from the grave as literal meanings, as I was at pains to point out in my paper on metaphor. There is obviously no listing in advance all the possible metaphorical meanings an expression may be given; if it were possible, creating and interpreting

Reply to Oliver Scholz

173

metaphors would not be the interesting tasks they sometimes are. One can, of course, put forward various necessarily vague principles guiding the interpretation of metaphors. But as we all know, this sort of interpretation can demand insight, taste, knowledge of history, familiarity with literature, and so on, without end. Such interpretation is the kind of thing that literary critics are supposed to be, and sometimes are, good at. But the idea of turning our literary skills into some kind of imitation of a serious semantical theory seems to me to be a ludicrous mistake. Here I should make a remark about Gricean implicature. One important motive in Grice's treatment of conversational implicature was the same as my motive in saying what I did about metaphor: to separate those aspects of communication which can be treated only informally from those aspects which can be given formal semantic treatment, namely, the relatively literal which underlies all the rest. I have nothing but admiration for what Grice did in that direction. It seems to me to be one of the classical defenses of the possibility of a serious theory of meaning. To my surprise, Scholz dismisses my proposal that metaphors often make us see one thing as another, because my proposal depends on a notion of "seeing as" which is, in his words, no clearer than metaphor itself. I thought Wittgenstein and others had made the notion of seeing as clear enough to make their (and my) point: there are important experiences that cannot be reduced to one way or another of grasping a propositional content. If it is a central function of (fresh, active, live) metaphors to induce such experiences, no theory of reference or truth can cope with what is distinctive about metaphor. Nothing I have said is meant to suggest that there aren't endless interesting things that can be said not only about individual metaphors but about metaphor in general. The idea of adding further bits of wisdom and wit is one that I can only applaud. But I confess that on the topic of metaphor I have found the writings of some literary critics infinitely more useful and insightful than the wooden attempts of philosophers to reduce metaphor to system.

How Relational Are Davidson's Beliefs? JOHANNES B R A N D L

I. Introduction The idea that beliefs are relational states of mind has two historical sources. On the one hand there is Brentano's theory of intentionality according to which all mental states, events and processes contain an object given in presentation. On the other hand there is the view derived from Frege and Russell that holding a belief means to take a certain attitude towards a proposition. These two sources can be linked by identifying the object contained in a belief with the proposition towards which the believing subject takes the attitude of holding-true. Thus we get the following explanation of belief: (Bel)

To believe something means to stand in the relation of holdingtrue to a proposition which is given in an act of presentation.

This explanation is contentious for various reasons. Brentano himself rejected propositions as objects of belief, and he came to doubt that a genuine relation holds between a subject and a presented object since the latter may not exist. Russell contested the assumption that in holding a belief one is related to a single object. However, both Brentano and Russell took it for certain that whatever the objects of beliefs are, they must be present to the mind of the believing subject. This basic assumption is challenged by Donald Davidson in his recent paper "What is Present to the Mind?". There Davidson raises a difficulty for the view expressed in (Bel) by posing the following dilemma: On the one hand there is the fact that to have a belief or other propositional attitude is to be related to an object of some sort; on the other hand there is the fact that there seems to be no satisfactory account of the psychological relation a person must be in to the appropriate object in order to have the attitude. 1

The psychological relation which seems to admit of no explanation in (Bel) is the relation of 'being given a proposition in an act of presentation'. Described in the active mood, it is the relation of 'grasping a proposition'. ' Davidson (1989a), p. 8.

176

Johannes Brandl

The difficulty Davidson sees for explaining this relation is how to exclude a misidentification of the proposition. Unless presentations are infallible in what they present to the subject, the believer might not know which proposition he grasps, and hence might not know what he believes. To admit such lack of knowledge of one's own mental states seems unsatisfactory. Yet, an infallible method of identification, Davidson argues, is not available. Thus the second horn of the dilemma is vindicated for him. The question I want to discuss in what follows is what happens with the first horn of the dilemma if Davidson's argument for the second horn is sound. Can the intuitive idea be sustained that having a belief means to be related to some sort of object? The conflict between this intuitive idea and the possible misidentification of the object of belief, Davidson indicates, may only "apparently" exist. Does he see a chance for saving the relational view even if no explanation of the psychological relation of 'grasping a proposition' is forthcoming? Davidson's view of the attitudes, it is generally assumed, derives from his paratactic account of indirect discourse, i. e. from his analysis of the saying-that locution. 2 So it comes that Davidson is criticised for holding a sententialist theory of belief. 3 According to this type of theory what a believer is related to is not a proposition 'grasped by' or 'presented to' him, rather it is a sentence which expresses his belief. Usually this move from propositions to sentences is made for the sake of ontological parsimony. Davidson, however, seems to have a different motive. He emphasizes that the sentence expressing someones belief need not be known to that person. Not only might she misidentify the sentences she holds to be true, these sentences may not even be "within her ken", as Davidson puts it. 4 Thus, it seems, the dilemma dissolves. There are objects of belief, namely sentences, without there being a psychological relation that waits for explanation. All we have to do is to exchange sentences for propositions in (Bel) and drop the reference to an act of presentation. In what follows I want to challenge this interpretation of Davidson's view. I think it is a grave misunderstanding of the paratactic analysis of belief-ascriptions to make it the basis of a theory of the mental states themselves. Davidson, I shall argue, is not a proponent of a sententialist theory, nor does he take utterances to be the objects of belief. Rather he dismisses the search for objects of belief in general. Getting the interpretation of Davidson right will be the task of sections 2 and 6. In the intervening sections I consider some objections to the paratactic theory. Partly these objections can be overcome, I argue, by 2 3 4

See Davidson (1968). See Schiffer (1987), ch. 5., and Barcan Marcus (1990). Davidson (1989a), p. 9.

How Relational Are Davidson's Beliefs?

177

making adjustments in Davidson's overall project of a unified theory of meaning and action (Section 3). Partly the objections remain intact as directed at a deflationist theory of meaning and belief. In section 4 I argue that Davidson is committed to such a deflationist theory, once the samesaying relation is recognized as playing a fundamental role in Radical Interpretation. This role in the justificatory task of Radical Interpretation must sharply be distinguished from the explanatory power inherent in a sententialist theory of belief (Section 5). I conclude that Davidson neither advocates a relational view of belief, nor does he offer an adequate alternative for it. II. The Logical Form of Belief-Ascriptions I can give here only a very brief summary of the paratactic account, leaving aside its intimate connection with Davidson's treatment of quotation. 5 First of all, there are two demands which Davidson places on an account of indirect discourse: (Dl) It should explain why certain logical moves cannot be justified by the principle that coreferring terms (or coextensional expressions 6 ) are substitutible salva veritate. (D2) It should explain these logical facts without invoking intensional entities like propositions or Fregean senses. To illustrate the logical moves which are problematic take the following example: Premise A: Premise B: Conclusion C:

Peter said that Mary is pregnant. Mary is John's girlfriend. Peter said that John's girlfriend is pregnant.

There are three possible reactions to this argument. (1) The tough-minded extensionalist accepts the argument as valid. He admits that there may be pragmatical reasons for preferring report A to report C, particularly if Peter does not know about Mary's relationship to John. But these pragmatic reasons, he holds, do not make a difference to the validity of the argument. The principle of substitution mentioned in (Dl) justifies the given inference. 7 (2) The Fregean intensionalist rejects the argument for semantical reasons. He admits that coreferring terms are substitutible salva veritate, but he 5 6

7

See Davidson (1968), pp. 97 f. and (1979), and McDowell (1980). I shall restrict my discussion to coreferring singular terms. The generalisation to any kind of coextensional expressions is straightforward. This position is defended by Scott Soames (1987).

178

Johannes Brandl

adopts a notion of coreferentiality according to which the terms "Mary" and "John's girlfriend" are not coreferring in the context of premise (A). Hence for him premise (B) offers no justification for moving from (A) to (Q·8

(3) Finally, there is Davidson's reaction which tries to steer a middle course. With the extensionalist Davidson shares the contempt for intensional entities, with Frege he shares the demand for a semantic explanation of the intuitive failure of the argument. This motivates the two demands stated above. How can they be satisfied? Davidson offers a solution in three parts. First he notes that principles of substitution are governing individual sentences. Substituting coreferring terms should preserve the truth-value of the sentence in which the substitution is made·, it may change however the truth-value of some other sentence. Next Davidson claims that (A) actually consists of two complete sentences which are connected by a demonstration. One of these sentences should express the content of Peter's utterance, the other just states that this is what the second sentence is supposed to do. Thus, in a more perspicuous notation, the argument looks like this: Premise A: Premise B: Conclusion C:

Peter said that. —• Mary is pregnant. Mary is John's girlfriend. Peter said that. —• John's girlfriend is pregnant.

The arrows indicate here that the demonstrative "that", not the pronoun "that" is used. The demonstration points to the next utterance, and thus indirectly to the sentence which is used for reporting what Peter said. 9 The arrows also indicate that from a pragmatic point of view the demonstrative reference is part of the speech act performed by uttering the introductory clause. With the second sentence, however, no separate speech act is performed. That Mary is pregnant is not stated as a separate premise in (A). From a pragmatic point of view, the second utterance merely satisfies a conversational expectation raised by the demonstrative element in the saying-that locution. 10 Does this Davidsonian reading make the argument valid? The answer is: it depends. It depends on what the introductory clause "Peter said that" 8

9

10

The Carnap-intensionalist differs from the Frege-intensionalist by adopting a further principle which allows the substitution of co-intensional terms in intensional contexts. I ignore this complication here because the argument would not be valid on Carnap's view either, even if "Mary" and " J o h n ' s girlfriend" were co-intensional terms. The indirect reference to a sentence-type sacrifices nominalism for the sake of making the theory m o r e flexible. Several objections raised against it can be met by this move. See McFertridge (1975/76), pp. 135 ff., Bürge (1986), pp. 198 ff., and Boer & Lycan (1986), pp. 51 f. This point is elaborated in Platts (1979), pp. 120 f.

How Relational Are Davidson's Beliefs?

179

means. If it means that Peter exactly used the words indicated by the demonstration, the argument fails. Certainly Peter could have used the second sentence in (A) without having used the second sentence in (C). On the other hand, if the introductory clause means that Peter said more or less what the indicated sentence expresses, then the argument is valid. Premise (B) guarantees that, in this looser sense, Peter cannot help saying implicitly that John's girlfriend is pregnant if he says that of Mary. The final part of Davidson's solution consists in an intermediate reading of the saying-that locution. Suppose Peter speaks only German. Then he will not use the very words "Mary is pregnant", yet speaking German does not prevent him from saying explicitly that Mary is pregnant, 11 without saying explicitly that John's girlfriend is pregnant. This intermediate case Davidson handles by introducing the relation of samesaying. When we report Peter as saying explicitly this-or-that, we put ourselves in a semantic relation to him. We want to be faithful to what he said, but with the liberty of using our own words. The following redescription of (A) gives the idea: (A*) Peter said, in a way which makes me and him samesayers, that. —> Mary is pregnant. Davidson is reluctant to call (A*) an analysis of premise (A), because it merely gives the truth conditions of the analysandum. Clauses like (A*), he says, are no more than a "rephrasal designed to give a reader a feeling for the semantics; an expository and heuristic device". 12 But (A*) fixes the meaning of the saying-that locution in a specific way. In this sense it gives an analysis of what it means to say something explicitly. The move from (A) to (A*) is by isolating this reading of the saying-that locution, not simply by "definitional abbreviation" as Davidson suggests. 13 If we treat the saying-that locution in (A) and (C) on the model of (A*), then (C) does not follow from (A) and (B). It does not follow because the two utterances indicated in (A) and (C) do not samesay each other. 14 If Peter samesaid the utterance indicated in (A), he will not thereby 15 have samesaid the utterance indicated in (C). Therefore the argument fails. " We might also say 'literally' when Peter uses the literal German translation of "Mary is pregnant." 12 Davidson (1976), p. 177. 13 Davidson (1968), p. 105. See also Platts (1979), pp. 118 f. 14 Davidson prefers to treat 'samesaying' as a relation between speakers, not utterances. On occasion, however, he also speaks of "synonymy between utterances" (1968), T&I, p. 104. Nothing hinges on this difference here. What is noteworthy, however, is that even if the utterances indicated in (A) and (C) did samesay each other, this would not exclude that only one of them samesays Peter's utterance, because the samesaying relation need not be transitive. 15 Of course, Peter may have made a different utterance samesaying my second utterance in (C).

Johannes Brandl

180

In this way Davidson explains the logical failure of substituting coreferring terms in indirect discourse without invoking intensional entities. The explanation consists, to repeat, of the following parts: (I) a defence of the substitutivity of coextensional expressions with respect to the truth value of individual sentences, (II) a demonstrative reading of the saying-that locution, (III) an appeal to the samesaying relation for distinguishing what the speaker explicitly says from what he says implicitly and from the very words he uses. In this three-stage solution the samesaying relation plays the fundamental role. It explains which substitutions in the content-clause do not affect the truth-value of the introductory clause. Thereby the partition of "x says that p" into two sentences is justified and the substitutivity of co-extensional expressions is saved. III. The Troubles with Samesaying Giving that prominence to the samesaying relation, it is no surprise that the critics of the paratactic theory focus on this relation. 16 Not all of these critics, however, do justice to Davidson's overall project of a unified theory of meaning and action. It will be important for our later considertations to see how the paratactic theory fits into this larger project. The objections raised against the samesaying relation fall into three groups. Critics say, this relation: ( 0 1 ) is theoretically useless, because nobody knows exactly when it obtains; ( 0 2 ) makes it impossible to generalise the paratactic account to all forms of indirect discourse, in particular to ascriptions of tacit beliefs; ( 0 3 ) supervenes on the fact that something is said twice, and therefore cannot explain this fact. An objection of the first kind is made by Stephen Schiffer. 17 Schiffer notices that for Davidson fixing the samesaying relation is just the same task as giving an empirically testable theory of meaning for a particular language. Davidson thinks this can be achieved by turning a Tarskian truth-theory into a theory of Radical Interpretation. To make this work, Davidson assumes that the meaning-determining features of an utterance are specifiable only in a holistic manner. This holism carries over to the 16

17

The major critics are, in order of their publications: Haack (1971), Lycan (1973), Blackburn (1975), Arnaud (1976), McFertridge (1975/76), Cresswell (1980) and (1985), Baldwin (1982), Bürge (1986), Schiffer (1987), and Barcan Marcus (1990). See Schiffer (1987), pp. 117 ff.

How Relational Are Davidson's Beliefs?

181

samesaying relation. For Davidson there is no way to decide for just two utterances whether they samesay each other. To decide this one has to take into account the entire languages to which these utterances belong. Yet, nobody is able to overlook simultaneoulsy the interpretation of all possible utterances in a given language. So nobody is ever in a position to know when the samesaying relation obtains. If this is correct, then the following proposal by Ernest Lepore and Barry Loewer cannot work. They suggest that it is not the truth theory, no matter how it is constrained or what is known about it, which bridges the gap between heard utterances and ascribed assertions; ... one needs also to know the samesay relation, or some proper subset of this relation, for the language. 18

Contrary to what Davidson suggests, Lepore and Loewer think that a truth theory, even when construed holistically, permits a wide range of distinct interpretations. It is only knowledge of the samesaying relation which narrows down the acceptable interpretations to a managable set of alternatives. Hence they suggest that this additional knowledge must be exploited in Radical Interpretation. The problem for Davidson's semantic project raised by Schiffer is which way to go. Should we maintain the holistic character of Radical Interpretation, and thus admit that knowledge of the samesaying relation is 'not within the ken of plain folk'? Or should we use an implicit knowledge of the samesaying relation as a guide towards correct interpretations, and thus give up holism? I agree with Schiffer that the first strategy is not attractive. We do have the linguistic competence for understanding individual utterances of our fellow men. Thereby we have a grasp of particular instances of a samesaying relation. A theory of interpretation which is unable to account for this competence is not worth developing. To save the paratactic theory from Schiffer's objection, I think, the holistic assumptions in Davidson's theory should be dropped. The second type of objection to the paratactic theory is the standard objection to any sententialist theory of belief. How does such a theory account for 'tacit beliefs', i. e. beliefs that are never expressed? To handle such cases Davidson suggests the use of subjunctive conditionals. 19 We can describe a tacit belief as the mental state someone is in who could honestly assert such-and-such if he spoke our language. The difficulty is how to support such conditionals. Why is it, for instance, that Peter could assert "Mary is pregnant" if he spoke English? Again, there are two ways for Davidson to go: either he can say that the subject instantiates a physical belief-token which mentally samesays this English 18 19

Lepore & Loewer (1989b), p. 75. See, Davidson (1975), p. 167.

182

Johannes Brandl

utterance; 20 alternatively, he can of same-believing to a hypothetical explicitly, Davidson introduces following redescription of "Paul

say that the subject stands in a relation English speaker. Without mentioning it this latter relation when he offers the believes that":

Paul believes what I would believe if I were sincerely to assert what I say next. 21 This proposal, as we shall see, makes Davidson's account of belief vulnerable to the third type of objection. Here it is claimed that samesaying is a relation 'supervening' on the fact that something is said twice. As a consequence the samesaying relation cannot be used, without circularity, for explaining what it means to say something. Given the proposed analogy between 'saying' and 'believing', it follows that the same-believing relation can neither be used for explaining what it means to hold a certain belief. This objection I want to develop now in detail. IV. A Deflationist Theory of Meaning and Belief If somebody makes an utterance and somebody else reports his utterance, these are two independent facts: the utterance could be made without being reported and the report could be made without the utterance taking place; the report could simply be false. This independence causes trouble for Davidson's claim that (A') gives the logical form of (A): (A) (A')

Peter said that Mary is pregnant. Peter said that. —* Mary is pregnant.

The trouble is to show that the paratactic reading is logically equivalent to the original. 22 Various authors have doubted this. 23 Their arguments are twofold. On the one hand, they deny that (A) implies (A') because we can imagine a world without English in which the utterance demonstrated in (A') is meaningless. In that world Peter could still say in some other language that Mary is pregnant, thereby making (A) true. (A'), however, would be false because Peter cannot samesay a meaningless sequence of letters. On the other hand, it is argued, (A') does not imply (A) either. In 20

21 22

23

This solution is favoured by Lepore & Loewer in (1989a), pp. 352f., but rejected by Davidson: "I find no plausibility in the idea that thoughts can be nomologically identified with, or correlated with, phenomena characterized in physical or neurological terms." (1982), p. 322. Davidson (1989a), p. 14. Logical equivalence must hold here between two utterances, not sentences. However, one may define it as obtaining between a single sentence and two other sentences under the presumption that the latter two sentences are uttered together. See Lycan (1973), Blackburn (1975), Bigelow (1978).

How Relational Are Davidson's Beliefs?

183

this case we should imagine a world in which the utterance demonstrated in (A') means something different from what it actually does. Suppose it means that Mary is a widow. Then, if Peter utters in German "Maria ist eine Witwe", (A') would be true, but (A) false. As they stand, these arguments have little force. 24 They merely repeat an objection Arthur Pap once made against Tarski's definition of truth. 25 Pap argued that no instance of the scheme (Eq)

S is true in L iff p.

is necessarily true, and therefore no instance of (Eq) will follow from a truth-definition. Take the paradigm case: "Schnee ist weiß" is true in German iff snow is white. This might be false, Pap argues, either because no German is spoken in a world in which snow is white, or because snow is purple in a world in which the German predicate "weiß" happens to apply to purple things. What Pap misses is that Tarski's scheme is intended to hold only if the sentence "p" used on the right hand side is a translation of sentence S mentioned on the left-hand side. Making this explicit, we get: (Eq^

If "p" translates S, then S is true in L iff p.

This 'conditionalising' blocks the objection. 26 The price, however, is that Davidson's idea of using a Tarskian truth-definition as a theory of Radical Interpretation threatens to become circular. To avoid this, some weaker antecedent is needed. The following might do: (Eq 2 )

If S is an utterance in L at t, as this language is used at t, then J" is true iff p.21

Following this line we can also save the equivalence of (A) and (A'). We only have to make explicit under what condition those two utterances are 24

25 26

27

Davidson has been defended against these objections by Smith (1976), McDowell (1980), pp. 231 f., and most forcefully by Lepore & Loewer (1989a), pp. 347 f. and (1990), pp. 106 f. Though 1 think these defences are successful, the objections still have some weight if seen as the first step in a larger argument. See Pap (1955), pp. 64 ff. The objection is also blocked by the following expansion of (Eq) suggested by Lehrer (1974), p. 31: S is true in L and S means in L that p iff S is a sentence in L meaning that p and p. It is more difficult, however, to get this into line with Davidson's project than the equivalence (Eq,). (Eq2) makes explicit what is involved in replacing the absolute predicate "is true" by a language-specific predicate "true-in-Z,", as defined by Tarski. In the case of languages which change over time, or which contain indexicals, the truth-predicate would have to be "true-in-Z.-at-/". More about this language-dependence below.

184

Johannes Brandl

said to be logically equivalent. Fully expanded, the claim advanced by the para tactic theory becomes this: 28 (PA)

If (A) and (A') are both English sentences, as this language is used now, then the present utterance of (A) is logically equivalent to the utterance of (A').

This conditional statement neither implies nor presupposes the existence of English or of any utterances in English. If English were not available in our world, the antecedent in (PA) would simply be false. The utterances (A) and (A') might still be made in such a world, but they need not be equivalent for (PA) to be true. 29 What has to be admitted, however, is that unless the antecedent is satisfied, the paratactic theory tells nothing about the logical form of (A). The theory is simply silent about a world in which English does not exist. Naturally, then, it cannot be refuted by such a counterfactual situation. But a residual problem remains. The conditionalising suggested above is not a harmless logical manoeuvre. It weakens the analysis of what it means to say something explicitly by making the analysis depend on a particular language. Usually we expect that what is true of a given concept is true independently of any language in which this concept happens to be expressed. However, this may be too high an expectation. Perhaps we must content ourselves with a more modest view of conceptual analysis. Let us see. Tarski held that truth can be defined only if boundaries between languages are clearly observed. In view of the paradoxes, all we can expect is a definition of 'true-in-L' in the meta-language of L. Whether defining a list of such language-specific predicates 30 is all that can be said about truth is still an open question. Those who call themselves 'deflationists' 31 think that nothing beyond such lists is available. Davidson is not such a deflationist. He thinks that the language-specific predicates do not exhaust our notion of truth and that something more can, and must be said about

28

29

30

31 32

Reference to a language comes in here not as part of (A'), but as part of what the paratactic theory states about the relation between (A) and (A'). Therefore, a translation test will not cause trouble here. What may be problematic is the individuation of languages. But this trouble already arises for Tarski's truth-predicate when it is tied to a language. See Davidson (1968), T&I, 98. This point, I think, is overlooked by Blackburn in (1975), pp. 185 f. The point can be made without assuming that samesaying is a cross-world relation, as Lepore & Loewer suggest in (1989a), p. 348. It would be misleading to call "true-in-Z." a relativised predicate, because the "L" here is not a variable. Tarski does not define truth 'relative' to a language, in the sense in which model-theory defines truth relative to a model or relative to an interpretation. Sec e.g. Horwich (1990). See Davidson (1990), p. 292.

How Relational Are Davidson's Beliefs?

185

My concern here is not with truth, but with samesaying. So I ask: Is there more to say for Davidson about the samesaying relation than a deflationist can say? I think the answer is "No". Davidson conceives of the samesaying relation as fixed by a theory of Radical Interpretation. Such a theory provides for each utterance in the object-language an utterance samesaying it in the language of the theory. The interpreting utterance must be in the language of the theory because it is used, not mentioned, in stating a samesaying relation. This marks the big difference between "x means what is samesaid b y j " and " x is synonymous t o j " . In making a statement of the first form one must actually say what the utterance "x" means, whereas in stating a synonymy relation one can remain silent about what both "x" and " j " mean. 33 Now, if the Radical Interpreter speaks French, he fixes the relation 'x means what is samesaidby-the-French-utterance y\ if he speaks Spanish, he fixes the relation ' x means what is samesaid-by-the-Spanish-utterancejy'. 34 As long as we stick to Radical Interpretation, there is no hope of removing the reference to a language from these predicates. This point is essential, so let me repeat it. The theory of Radical Interpretation is a general theory about theories of meaning for particular languages. It provides us with a notion of means-in-A for variable languages. But it does not give us a general notion of samesaying for variable languages. The generality of 'means-in-Z,' concerns the object-language. The meta-language must be fixed. Therefore, we cannot treat the " L " in "is-samesaid-in-77' as a variable. We would thereby loose the language to state the samesay-relation. We would lapse into talking about synonymyrelations between utterances without actually making any of these utterances. How bad is this for the paratactic account? Does it show that Davidson misses something essential in the saying-that locution? One might think that he misses the very point of distinguishing between direct and indirect speech. Direct speech is introduced by the language-specific predicate "says-in-i". In quoting we must specify the language to which the quotation belongs. Indirect speech cancels this requirement. But Davidson re-introduces it by inventing artificial predicates like: "utters something which is samesaid by my next utterance in present-day-English". Thus, it seems, the whole point of indirect discourse is lost. But there is a reply for Davidson to make. Surely, he can say, the directspeech locutions "says-in-French", "says-in-Spanish", etc. have something in common, and this common element is expressed by the saying-that 33 34

See Davidson (1968), p. 104. Note that English is used here as the meta-meta-language. In fixing the relation "x means what is-samesaid-by-the-French-utterance y" French is the meta-language.

186

J o h a n n e s Brandl

locution. This common element, however, need not be the reference to a proposition expressed in different languages. What unifies the languagespecific predicates is the fact that a French and a Spanish speaker can be samesaid by a single utterance. Admittedly, to do this we need some language, but the particular language we use does not matter. We can imagine our own utterance equally as being samesaid in some other language. This process of samesaying can g o on indefinetely. That there must be some language in the end in which the samesaying is done, is as it should be. We cannot think and talk about what other people say from a viewpoint outside any language. Such a 'nowhere point of view', as it is called, is illusory. We are driven into this illusion when we imagine people saying or believing something which cannot be samesaid at all. 35 The only notions of'saying' and 'believing' we understand are notions which require the perspective of a potential samesayer. Why deny this? Davidson recommends this 'internalist' or 'deflationist' attitude in passages like the following: A s a matter o f principle, . . . , meaning, and by its connection with m e a n i n g , belief, also are open to public determination. I shall take a d v a n t a g e o f this fact ... and a d o p t the stance of a radical interpreter when asking a b o u t the nature o f belief. What a fully informed interpreter could learn about what a speaker means is all there is to learn; the same g o e s f o r what the speaker believes. 3 6

I have no quarrels with the view that what is said and what is believed is publically accessible. What I find problematic is the stronger claim advanced by Davidson that all that can be said about the nature of meaning and belief is contained in explaining how we recognise what is said and believed when adopting the 'stance of a radical interpreter'. This does not seem to be the case. When we talk about the 'nature' of meaning and the 'nature' of belief we are interested in questions of the following sort: why does this utterance mean that p, why does that behavior indicate that the agent believes that p, etc.? If these questions are treated in the way recommended by Davidson, we get answers of the following form: (M)

Utterance U means that p, because this is the best way to interpret it in L.

35

L e p o r e & L o e w e r use this thought-experiment as a counter-example to Schiffer w h o suggests a substitutional reading o f " G a l i l e o said s o m e t h i n g which w e cannot express in our o w n l a n g u a g e " . Unfortunately they d o not tell us h o w this sentence can be true on the paratactic account. See L e p o r e & L o e w e r (1990), p. 94.

36

D a v i d s o n (1986), p. 315. D a v i d s o n even g o e s further and claims that what goes for belief also g o e s for objectivity. T o get hold of an objective world, he thinks, w e equally need the interpreter to f o r m the 'base-line' on top o f which objectivity can be defined as the third point in a triangle. See D a v i d s o n (1982), p. 327, and (1989b), p. 199.

How Relational Are Davidson's Beliefs?

(N)

187

Behavior Β indicates that the agent believes that p, because attributing this belief to him in L is the best way to make sense of what he does.

The "L" in (M) and (N) is not a variable, but names the fixed metalanguage which we use as Radical Interpreters. This makes answers of type (M) and (N) flatly deflationist. What is explained are not the phrases 'means that' and 'believes that', but the more complex phrases 'can be samesaid-in-Z, by', and 'is expressible-in-Z. by'. This shift in the explanandum, we shall see, puts into question the explanatory power of the paratactic theory. V. Justification and Explanation Davidson wants to account for the fact that understanding the utterances of other people and attributing to them beliefs, desires and other attitudes, is a single business. There is no way to fix the content of what others say, independently of fixing the content of what they believe and desire. Therefore, he says, we need a unified theory of meaning and action. But why must this unified theory be given from the point of view of a Radical Interpreter? When we attribute to Peter the belief that Mary is pregnant we face two questions: (1) What justifies our attribution of this belief to Peter? (2) Why is it true that Peter holds this belief? The need to distinguish between these two questions arises because we may be justified in our claims about what somebody believes, though our claims are false. In discussing the rationality of animals, Davidson acknowledges this gap. He says that we may continue to explain the behavior of speechless creatures by attributing propositional attitudes to them while at the same time recognizing that such creatures do not actually have propositional attitudes. 37

Despite this acknowledgement, however, Davidson leaves no room for separating the question of having a certain attitude (or of being rational) from the problem of interpretation. He poses the question "What makes an animal rational?", but then immediately turns to the question "How do we tell when a creature has propositional attitudes?". This shift from explanation to justification is encouraged by phrases like 'determining what is said' and 'fixing the content of belief. These 37

Davidson (1982), p. 324.

188

Johannes Brandl

phrases are notoriously ambiguous. Taken in one sense they indicate that certain facts about the utterances and the mental states of a person give them the meaning and content they have. These may be facts about the language to which the utterances belong, facts about their causal history, facts about the behavioral dispositions of the person being in these mental states, etc. What we want to learn from such facts is what determines the meaning of an utterance and what fixes the content of a belief. A quite different issue is how it can be 'determined' or 'fixed' what somebody says and believes. What we are looking for in this case are the data on which a Radical Intepreter relies when he tries to find out what somebody says and believes. He needs facts which are easily accessible from his point of view and which j u s t i f y his interpretation. By contrast, the facts that explain how meaningful speech and contentful thinking arises may be quite remote from the interpreter's perspective. This is why, after solving the interpretation problem, we may still ask: why is it that utterances mean what they do according to our theory of interpretation, and why is it that people behaving in a certain way have the beliefs which we are justified in ascribing to them? Once this difference between 'justification' and 'explanation' is appreciated, the paratactic theory of indirect discourse appears in a new light. When we ask "What justifies us in attributing beliefs?" the theory tells us how to justify the samesaying relation which we establish between us and the subject to which we attribute a belief. 38 But when we ask "Why is our attribution true?" the samesaying relation is useless. No progress is made by transforming this latter question into the question: "Why is it true that my utterance samesays the mental state of the believer?" The samesaying relation obtains because my utterance and the ascribed mental state share certain semantic features. Asking for these features, we are referred back to where we started from, namely to the question, "What gives an utterance and a mental state the meaning and the content that we report in samesaying them?" In this new light, then, the correct view of the paratactic theory seems to be this: the theory contributes to justifying the ascription of beliefs and other mental attitudes. It does so by giving the samesaying relation a fundamental role in Radical Interpretation. This role is comparable to the role Davidson assigns to the attitude of 'holding a sentence true'. Davidson considers this attitude to be an observationally accessible starting point for Radical Interpretation, a starting point, as he puts it, which is "beyond 38

Strictly speaking this is only true for beliefs which are expressed by the subject or which are tokened by a mental sentence. The point, however, could also be made in terms of a relation of same-believing as noted in section III. I leave aside this complication for the moment.

How Relational Are Davidson's Beliefs?

189

and beneath" the basic trio of belief, desire and meaning. 39 This observational basis alone, however, will not take us to statements of the form "S says/believes that p". We need additional data that assure us of samesaying S by using the sentence "p". So the theorems to which Radical Interpretation leads are actually of the form "S says something that is samcsaid-in-L by that. —• p". Without invoking the samesaying relation in its theorems a systematic attribution of meaning and belief cannot be justified. In this way the paratactic theory becomes indespensible for the project of Radical Interpretation. By the same token, however, it looses the explanatory power expected from a relational theory of belief. I shall argue this next.

VI. The Ontological Commitment to Mental States Remember the dilemma Davidson posed for a relational account of beliefs and other propositional attitudes. The dilemma is to square a fundamental intuition with a theoretical plight. The intuition is that having a belief means to be related to an object of some sort. The theoretical plight is that metaphors like 'grasping a proposition' or 'being given an object in presentation' do not explain the required relation. Here the paratactic theory appears to offer a way out. It suggests that the objects of belief are not 'grasped' or 'entertained', but samesaid or same-believed. From an epistemological point of view, these relations are even looser than holding a sentence true. In holding a sentence true, one must be somehow aware of that sentence. By contrast, one need not have the slightest idea which sentences are samesaying one's beliefs. This lack of epistemological contact makes it impossible to say that believing is a relation to an entity which samesays the belief. We cannot say: X believes that p because X stands in the samesaying relation to an utterance of "p". It is just the other way round. The samesaying relation obtains because X believes that p, and because "p" means that p. For that reason, the samesaying relation cannot replace a psychological relation between a believer and a propositon. It cannot replace it because it does not have the same explanatory power. The conclusion I draw from this is that Davidson does not hold anything like a sententialist theory of belief. He does not claim that "X believes

59

Davidson (1980), p. 2.

190

Johannes Brandl

that p" is equivalent to "X holds a certain sentence true." 40 Instead, Davidson says: There should be no temptation to call the utterance to which reference is made according to the paratactic analysis the object of the attributed attitude. 41

This makes it plain, I think, that Davidson does not take sentences or utterances to be objects of propositional attitudes. Is there anything else that could be an object of belief for him? Davidson's advice is this: It cannot be an independent goal of a theory or analysis to identify the meanings of expressions or the objects of belief. 42

The advice here is not to view beliefs and propositional attitudes as relational states at all. Mental states are just as non-relational as the physical state of weighing 100 pounds. Davidson uses this analogy for illustrating that a relational statement like " X weighs 100 pounds" need not describe a relational state of X. A relational reading of "X weighs 100 pounds" is needed merely for enabling us to interpret a potential infinity of statements of this form. 43 Similarly, the paratactic reading of "S believes that p" helps us to interpret a potential infinity of belief-ascriptions. The analogy here is this: the relation we establish between a number and a physical object in saying that X weighs 100 pounds is supervenient on the seize and the physical structure of the object. Similarly, the samesaying relation supervenes on the semantic features of the entities standing in this relation. In neither case do we describe a real physical or psychological relation. What we introduce are just heuristic tools for keeping track of what things weigh and what people believe. But Davidson wants to go a step further. He also wants to use the analogy for arguing that we are not ontologically committed to mental states, just as we are not ontologically committed to entities like weights. What we need in our ontology of measuring are merely the objects which we compare as being more or less heavy (and perhaps the numbers which we use to describe these relations systematically). Can we achieve the same parsimony in the case of mental states? Does the paratactic theory show 40

41 42

43

Barcan Marcus reads Davidson as claiming that these two statements are equivalent "in a shared interpreted language", and that therefore "believing is a conscious relation of subjects to their utterances" (1990), p. 134. On my reading, it makes no difference to which language the sentence belongs that is used in attributing the belief. The belief itself is never an attitude towards a sentence or an utterance according to Davidson. Davidson (1975), p. 166. Davidson (1974), p. 149. This remark does not follow the strategy of Jeffrey, contrary to what Davidson suggests by quoting Jeffrey as saying that "for the purpose of the theory, the objects of these various attitudes could as well be taken to be sentences" {ibid., p. 148). See Davidson (1989a), pp. 9 f.

How Relational Are Davidson's Beliefs?

191

us how to compare what different people say and believe without thereby committing ourselves to the existence of mental states? This idea of using the paratactic theory for gaining ontological parsimony, I think, is mistaken. The trouble, once again, lies in the explanatory weakness of the theory. Suppose we ask in the measuring case: why does object X way 100 pounds? To this we can reply: X weighs 100 pounds because it weighs 100 times as much as the object we have chosen as our unit pound. In the case of a belief-ascription no such answer is available. It is not the case that Peter believes that Mary is pregnant because he stands in a samesaying or same-believing relation to my utterance of "Mary is pregnant". 44 Thus I would submit that Davidson goes one step too far when he claims that in thinking and talking about the beliefs of people we needn't suppose there are such entities as beliefs. Nor do we have to invent objects to serve as the 'objects of belief or what is before the mind, or in the brain. 45

If my interpretation of the paratactic theory is correct, this theory merely shows how we can read belief-ascriptions as relational statements without taking the mental states themselves to be relational states. It makes us free, as Davidson says, "to divorce the semantic need for content-specifying objects from the idea that there must be any objects at all with which someone who has an attitude is in psychic touch." 46 Yet, I do not see that this liberates us from referring to beliefs as entities whose nature we have to explain. On the contrary, I think that we are comitted to mental states just as we are committed to physical events. Ironically, it is an argument used by Davidson himself that supports this view. 47 This will be my final point. VII. Conclusion When Davidson discusses the ontology of events, he inquires about the validity of arguments like the following: Premise: The explosion was in the cellar. Conclusion: There was an explosion in the cellar. 44

45 40 47

As Blackburn notices, the ascription of mental states is in this respect more similar to predicating colours than to predicating weights. See (1975), pp. 1 8 7 f . Davidson (1989a), p. 11. Ibid., p. 9. This was already noticed by Jennifer Hornsby in (1976/77). Since to say something means to perform a speech act, and since actions are events for Davidson, she analyses statements of the form 'S says that p' by quantifying over utterances. By analogy, then, she analyses belief-ascriptions as quantifying over mental states.

192

Johannes Brandl

T h e v a l i d i t y o f this inference can be established by a f o r m a l rule. B u t this is n o t e n o u g h , D a v i d s o n admits: What is essential, ..., is not formality, but the justification of inference on the basis of conditions of truth. We need to say why what makes the premise true must make the conclusion true, and it is in accounting for this that we need an ontology of explosions. 48 D a v i d s o n distinguishes here b e t w e e n a superficial justification o f the i n f e r e n c e by c o n s i d e r i n g o n l y the t r u t h - c o n d i t i o n s o f p r e m i s e and conclusion, and an explanation o f w h y these conditions necessarily obtain tog e t h e r . 4 9 T h e justification is a matter o f ' f o r m a l i t y ' , w h e r e a s the explanation s h o u l d be ontological. It should tell us w h y w h a t makes the one c o n d i t i o n obtain, also makes the o t h e r obtain. T o account f o r this, an o n t o l o g y o f e v e n t s is needed. Similarly, I w o u l d argue, once w e ask w h a t makes o u r belief-ascriptions true, w e need an o n t o l o g y o f beliefs. A n d once w e h a v e made this o n t o l o g i c a l c o m m i t m e n t , w e h a v e t o explain the n a t u r e o f these states. T h e paratactic t h e o r y s h o w s us h o w t o a v o i d the standard v i e w that beliefs and desires are relational states. It does n o t o f f e r a n y p o s i t i v e t h e o r y instead, n o r does it eliminate the need f o r such a t h e o r y b y semantic ascent.

REFERENCES

Arnaud, Richard B. (1976), "Sentence, Utterance and Samesayer" Nous X, 283—304. Baldwin, Thomas (1982), "Prior and Davidson on Indirect Speech" Philosophical Studies 42, 255-82. Barcan Marcus, Ruth (1990), "Some Revisionary Proposals About Belief and Believing" Philosophy and Phenomenological Research, L Supplement, 133 — 153. Bigelow, John (1978), "Believing in Semantics" Linguistics and Philosophy 2, 101—44. Blackburn, Simon (1975), "The Identity of Propositions" in S. Blackburnn (ed.), Meaning, Reference and Necessity, Cambridge: Cambridge University Press, 182 — 205. Boer, Stephen & Lycan, William G. (1986), Knowing Who, Cambridge MA: The M.I.T. Press. Bürge, Tyler (1986), "On Davidson's 'Saying That'" in: E. Lepore (ed.), Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson. Oxford, Basil Blackwell, 190—208. Cresswell, Max (1980), "Quotational Theories for Propositional Attitudes "Journal of Philosophical Logic, 9, 17—40. Cresswell, Max (1985), Structured Meanings. Dordrecht, Reidel. 48 49

Davidson (1971), p. 190. The answer to both questions would be trivial if the premise and the conclusion had exactly the same truth condition, i. e. if we added a singularity condition in the conclusion. The interesting cases, however, are inferences in which the premise(s) and the conclusion differ in truth-conditions, for instance, when we move from "There was a heavy dynamite explosion in the cellar" to "Something exploded in the cellar". I owe this point to Peter Sullivan.

How Relational Are Davidson's Beliefs?

193

Davidson, Donald (1968), "On Saying That", T&I, 9 3 - 1 0 8 . - (1971), "Eternal vs. Ephemeral Events" in: Essays on Actions and Events, Oxford, Basil Blackwell 1980, 189-203. - (1974), "Belief and the Basis of Meaning", T&I, 141-154. - (1975), "Thought and Talk", T&I, 155-170. - (1976), "Reply to Foster", T&I, 1 7 1 - 7 9 . - (1979), "Quotation", T&I, 7 9 - 9 2 . - (1980), "Towards a Unified Theory of Meaning and Action" Grader Philosophische Studien, 11 (1980), 1 - 1 2 . - (1982), "Rational Animals" Dialectica 36, 3 1 7 - 2 7 . - (1984), Inquiries into Truth and Interpretation, Oxford, Basil Blackwell, ( = T&I). - (1986), "A Coherence Theory of Truth and Knowledge" in: E. Lepore (ed.), Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson. Oxford, Basil Blackwell, 307-19. - (1989a) "What is Present to the Mind?" in: J. Brandl & W. Gombocz (ed.), The Mind of Donald Davidson. Amsterdam, Rodopi, 3 — 18. - (1989b) "The Conditions of Thought" in: J. Brandl & W.Gombocz (ed.), The Mind of Donald Davidson. Amsterdam, Rodopi, 193 — 200. - (1990) "The Structure and Content of Truth" Journal of Philosophy, 87 (1990), 2 7 9 - 3 2 8 . Haack, R. J. (1971), "On Davidson's Paratactic Theory of Oblique Contexts" Nous, 5, 351-61. Hornsby, Jennifer (1976/77), "Saying O f ' Analysis 37, 1 7 7 - 8 5 . Horwich, Paul (1990), Truth, Oxford: Basil Blackwell. Lehrer, Keith (1974), Knowledge. Oxford, Clarendon Press. Lepore, Ernest, ed., (1986), Truth and Interpretation. Perspectives on the Philosophy of Donald Davidson. Oxford, Basil Blackwell. Lepore, Ernest & Loewer, Barry (1989a), "You Can Say That Again" in: P. French et. al. (ed.), Midwest Studies in Philosophy, XIV, Minneapolis: University of Minnesota Press, 338-356. - (1989b), "What Davidson Should Have Said" in J. Brandl & W. Gombocz: The Mind of Donald Davidson, Amsterdam 1989, 6 5 - 7 8 . - (1990), "A Study in Comparative Semantics" in: T. Anderson and J. Owens (ed.): Prepositional Attitudes: The Role of Content in Logic, Language, and Mind. Chicago: University of Chicago Press, 9 1 - 1 1 2 . Lycan, William G. (1973), "Davidson on Saying That" Analysis 33, 1 3 8 - 3 9 . McDowell, John (1980), "Quotation and Saying That" in M. Platts, ed., Reference, Truth and Reality, London, Routledge and Kegan Paul, 206 — 38. McFertridge, Ian (1975/1976), "Propositions and Davidson's Account of Indirect Discourse" Proceedings of the Aristotelian Society, 76, 131 — 145. Pap, Arthur (1955), Analytische Erkenntnistheorie. Springer, Wien. Platts, Mark (1979), Ways of Meaning. London. Routledge and Kegan Paul. Schiffer, Stephen (1987) Remnants of Meaning. Cambridge MA: The M.I.T. Press. Smith, Peter (1976), "Blackburn on Saying That" Philosophical Studies 30, 423—26. Soames, Scott (1987), "Substitutivity" in: J. J. Thomson (ed.), On Being and Saying: Essays for Richard Cartwright, Cambridge MA: The M.I.T. Press, 9 9 - 1 3 2 .

Donald Davidson: Reply to Johannes Brandl There is much in this paper that I applaud; it accurately reflects my position on a number of issues. I shall comment briefly on some of these issues. Are beliefs relational? My answer is: No, beliefs are not relational. There is no entity to which a person is related psychologically or epistemically, where standing in that relation explains or constitutes the belief. To have a belief is to have a certain property; I assume that even if there are such things as properties, the property of having a certain belief is not the "object" of that belief. From an ontological point of view, for a person to have a belief is just what it is like for a room to have a temperature. We don't suppose that if a room has a certain temperature then the room is in a relationship to another entity, namely, its temperature. Long ago, Carnap explained that we should construe sentences such as "The temperature of the room is 68° Fahrenheit", not as expressing an identity between a temperature and something called 68° Fahrenheit, but rather as identifying the temperature of the room in Fahrenheit with the number 68. This analysis allows us to use numbers to measure things without invoking an ontology of temperatures, weights, lengths and speeds. The response to this observation may well be that perhaps we don't need such entities as beliefs but this doesn't show that having a belief isn't being in some appropriate relation to an entity, perhaps a proposition. After all, Carnap's proposed analysis of sentences about temperatures does introduce a relation between things that have a temperature and numbers. This response is fair: the semantic analysis of sentences that attribute attitudes does make such sentences relational. When I denied that having a belief is being in some psychological or epistemic relation to an object, I did not imply that the semantics of attitude-attributing sentences were not relational. Just as sentences attributing temperatures to rooms relate rooms to numbers, so sentences attributing attitudes to people relate people to some entity; what I deny is that the person is in any way aware of, entertains, or grasps, that entity. There is a potential infinity of different beliefs that people can have, so we need some potential infinity of entities to relate people to if we are to be in a position to attribute an arbitrary belief to them.

Reply to Johannes Brandl

195

What are relational, then, are not states of belief, but the sentences we use to attribute them. Since it is the attributor, not the attributee, who needs this infinite supply of entities, we must ask what those entities could be. The candidates must be potentially infinite in number, entities to which the attributor can refer, and which mirror the structure of beliefs as the numbers can represent the structure of weights or temperatures. Some philosophers have chosen sentences for this job, others propositions. My paratactic account chooses utterances. Since I have spelled out my reasons for this choice elsewhere, I won't repeat them here. I note here a few respects in which I cannot agree with Brandl. My analysis proposes that in saying "Peter said (that) Mary is pregnant" one actually produces what are semantically two sentences: "Peter said that" and "Mary is pregnant". Brandl remarks, "With the second sentence,...no separate act is performed." Perhaps it doesn't matter how we count speech acts, but I count each utterance of a sentence as a speech act, and this is important for my analysis, since the act of uttering "Peter said that" refers to the utterance of "Mary is pregnant". Of course I agree that in uttering "Mary is pregnant" in this context, the attributor is not stating or asserting that Mary is pregnant; but many speech acts are not statements or assertions. Brandl also thinks to improve on my analysis by making the "that" refer to a sentence rather than an utterance. It may be that in the end this route solves more problems than it raises, but Brandl certainly doesn't say why. He mentions criticism of the utterance analysis in Boer and Lycan, but I find their criticism unintelligible. He also mentions a criticism of Burge's, which is that on my analysis, "Peter said that Mary is pregnant" doesn't entail, as it should, "Peter said something". Bürge apparently didn't notice that the logical form I took to be official was "(Ex) (Peter uttered x, and my next utterance makes Peter and me samesayers)" followed by my utterance of "Mary is pregnant". It is true that I was vague in "On Saying That" whether samesaying related utterances or people; and I can't see that it much matters. But I see that it would be simpler, and therefore a bit better, to take it to express a relation between utterances. Thus my preferred form will be, "(Ex) (Peter uttered x, and my next utterance samesays x ) " . Finally, I am not quite sure whether Brandl has kept in mind the essential distinction between giving the logical form of a class of sentences on the one hand and analyzing the individual words in such sentences on the other. Logical form doesn't attempt to give the meanings of individual words (or phrases that the semantics treats as single words). For this reason it is irrelevant to the cogency of my analysis of indirect discourse that I have not analysed my word "samesays". It is not intended as an improvement on what is already built into the "says" of idiomatic indirect

196

Donald Davidson

discourse. In my opinion the various "improvements" on samesaying essayed by Bürge, Brandl and others miss the point of the project of assigning logical forms to sentences that make the semantics of those sentences transparent. [In response to a question concerning the comparison with temperature:] Well, apparently my example about temperature didn't take. It seems to me evident that our sentences that attribute temperatures are relational, and it seems to me equally obvious that there isn't some physical object that we are relating something to when we say it has a certain temperature. The same for weights, lengths and speeds. We should take the same view of psychological states. If you ask what kind of properties we're attributing when we attribute beliefs, I think a theory about how we tell that beliefattributing sentences are true provides the best answer. This shows what kind of property it is: it's a property which you determine to apply to an individual in the following way...(and here you describe the method). Is there something more to say about it? I don't see why there has to be. [In response to a question concerning the notion of samesaying:] The notion of samesaying was never intended to provide an analysis of the concept of saying that is employed in indirect discourse. It was just meant to give an insight into the nature of the semantics, the logical form, that I was attributing to such sentences. No new weight was supposed to be put on the word "samesaying"; I invented the word, as far as I know, and it's not meant to improve our understanding of what is communicated by indirect discourse. In so far as I think such a thing is possible, my "analysis" of belief (not of hcMef-sentences), is contained in my theory of interpretation. That's what's supposed to answer the question, how do you tell when a belief-attribution is true?

First-Person Authority and Radical Interpretation E V A PICARDI

I. The issue of first-person authority is touched upon in several of Davidson's papers — e.g. "Empirical Content", "First-Person Authority", "What is Present to the Mind?", "The Conditions of Thought". Very crudely put, the issue of first-person authority centers around the question, What justifies the presumption that a speaker is not mistaken when he says he has a certain propositional attitude? In Davidson's own words: When a speaker avers that he has a belief, hope, desire or intention there is a presumption that he is not mistaken, a presumption that does not attach to his ascriptions of similar mental states to others ("First-Person Authority", p. 101). 1

The presumption of freedom from error in the first person case and the lack of such a presumption in the third person case seem to be the distinguishing mark of utterances issued with special authority. One may be tempted to put matters thus: of myself I know whether I am entertaining a belief or desire, whereas of others I can only believe that they have a belief or desire. This reformulation simply restates the problem: for why should it be more appropriate to speak of knowledge in my own case and of belief in the case of others? Besides, as Wittgenstein pointed out regarding sensations, in real circumstances it is often very hard to doubt the fear or pain of others (Philosophical Investigations § 303) and therefore to frame matters in terms of degree of certainty is not only unhelpful, but possibly misleading. It seems therefore more promising to confine our attention to the linguistic form of expression, to the actual use which we make of words indicating propositional attitudes, leaving the epistemological issue in the background. Now the distinguishing mark I have just mentioned (i. e. asymmetry between speaker and hearer regarding freedom from error) is far from being uncontroversial and it is perhaps worth remembering that selfascriptions of sensations such as those which crowd the literature about sense-data statements are rarely questioned not because the self-ascriber is ' D. Davidson, "First-Person Authority", Dialectica

38 (1984). pp. 1 0 1 - 1 1 1 .

198

Eva Picardi

bound to have the last word, but because such questioning would often be pointless. I agree with David Pears when he says that we rarely have an interest in the peculiar character of other people's sensa data and it is only politeness that makes us regard what people say as a sufficient condition of the truth of what they say. 2 The case of belief, however, is different from that of sensations, and it is only with belief that I concern myself here. Matters are further complicated by the circumstance that Davidson brings to bear on the phenomenon of first person authority certain general insights concerning the asymmetry between the first and the third person present indicative exhibited by what for brevity I shall call psychological verbs. Davidson claims that philosophers like Wittgenstein, Strawson and Austin, who laid great stress on certain asymmetries in the use of psychological predicates failed to give a satisfactory answer to the problem of first-person authority. This in itself is not very surprising, for the asymmetry with respect to freedom from error does not play a decisive role in Strawson's or Wittgenstein's account. Thus the question arises what the role of the presumption of freedom from error is supposed to play in the interpretation of speech, over and above the acknowledgement of other asymmetries we come across in the use of psychological predicates. I will return to this question in the second part of my talk. It is important to recall that the asymmetry between first and third person is a very widespread phenomenon which concerns all performative utterances; indeed, one of the reasons which prompted Austin to stress the asymmetry between say " I promise not to be late" and "You promise not to be late" was that in the first case we have an instance of doing something (promising) and in the second case an instance of describing something. Since certain uses of the verb "believe" to which Davidson draws attention have a lot in common with performative utterances it would be prejugding the issue as to the source of the authority if we took 2

At the close of his excellent discussion of Ayer's and Austin's disagreement over sensedata, Pears says: "In real life the character of his sense datum is usually unimportant, and it is politeness that leads us to speak as if his say-so were a sufficient condition of the truth of the statement about it. For example, a chef would be interested in a diner's sense datum only if he had been labouring to produce a sensory effect in him, or, more rarely, if his sense datum provided the only evidence of the taste of the sauce. ... Otherwise, the character of his sense data would not be sufficiently important for his statement to be worth challenging or testing, and so social reasons for letting him have the last word would begin to operate. This trivial fact is worth stating, because it is part of the basis of the common intuition that the speaker is the final authority on the character of his sense datum." (A Comparison between Ayer's Views about the Privileges of Sense-Datum Statements and the Views of Russell and Austin, in G.F. Macdonald (ed.) Perception and Identity. Essays presented to A.J. Ayer with his replies to them, 1979, London, Macmillan, pp. 61—81.)

First-Person Authority and Radical Interpretation

199

the terminology of "self-ascription" and "other-ascription" for granted. The legitimacy of this terminology rests on the fact that "I believe that it is snowing" cannot possibly be true in virtue of anything but my believing that it is snowing and that "You believe that it is snowing" cannot possibly be true in virtue of anything but your believing that it is snowing. Those who dislike the phrase "in virtue o f ' may replace it by "if and only i f ' . But such a statement of truth-conditions throws little light on the use of the verb "believe" in those two sentences; it possibly blurs the issue by suggesting that in both cases our only or main concern is the ascription of an attitude. Thus by using in what follows the terminology of selfascription and other-ascription I do not mean to commit myself to anything more than what is entailed by the trivial statement of truth-conditions I have just mentioned. But are not all utterances of a speaker issued with authority? What is the point of trying to single out those issued with special authority? The elusiveness of such a search is perhaps the reason which has led some philosophers to deny that there is such a thing as first-person authority, or, while conceding that there is such a thing, deny that it plays an important role in understanding speech. This is not Davidson's position, for he holds that our pretheoretical intuitions concerning first-person authority are sound. Philosophers, however, have failed to uncover its source and therefore failed to grasp what the linguistic and epistemological evidence underlying the phenomenon of first person authority is evidence of. In want of such an account we have, in Davidson's opinion, no answer to the skeptic who asks how we know that by using the same word in ascribing propositional attitudes to ourselves and to others we are really dealing with the same mental state or disposition. Davidson's main thesis is that first person authority turns out to be a necessary feature of radical interpretation. The acknowledgement of first-person authority goes hand in hand with the acknowledgement that not every instance of understanding speech is an instance of interpreting speech — in Davidson's technical sense of the word. I do not interpret my own speech, nor do I look at the jottings of my pupils in order to figure out what I mean while giving a lecture. 3 I do not rely in 3

At least two familiar parallels spring to mind and are worth mentioning. The first is Wittgenstein's remark in the Philosophical Investigations (§ 201) that there is a grasp of a rule which is not an interpretation of it. Following a given rule is a practice and, one may add, interpreting it is a different practice: in contrast with, say, talking and walking, one can't (generally) do both things at once. The second parallel is Ryle's discussion of higher order practices in chapter VI of The Concept of Mind, in particular his discussion of the systematic elusiveness of I and the partial lack of parallelism between it and the notion of "you" or my last year's self. Such elusiveness is Ryle's ersatz for the asymmetry between the first and the second person which he does not acknowledge.

200

Eva Picardi

my own case on the same type of evidence on which I usually rely in the case of other people. The fact that I do not rely on the same type of evidence does not mean that I rely on no evidence — a delicate point to which I shall return presently. However the crucial question is whether I could rely on the same type of evidence, even if in fact I do not. If this were the case, the skeptical challenge could easily be met. But in Davidson's opinion no such strategy is available. According to Davidson the assumption of first-person authority is, as it were, a condition of the possibility of the activity of interpreting another person's speech. The interpreter has no choice but to assume that the speaker is not mistaken when employing a psychological verb in the first person present indicative followed by a that-clause. In Davidson's own words: There are of course beliefs that carry a very high degree of certitude and in some cases their content creates a presumption in favour of their truth. These are the beliefs about our own present propositional attitudes. But the relative certitude of these beliefs ... springs from the nature of interpretation. As interpreters we have to treat self-ascriptions of belief, doubt, desire and the like as privileged: this is an essential step in interpreting the rest of what the person says and thinks. ("Empirical Content", p. 332) 4

I will reserve the expression "attitude-authority" — which is my main concern in this paper 5 — for referring to the phenomenon alluded to in the above quotation, in order to distinguish it from a more general phenomenon underlined by Davidson which, for the sake of symmetry, may be called content-authority; a phenomenon to which I now turn.

II. According to Davidson the phenomenon of first person authority is by no means confined to utterances in which we self-ascribe attitudes. Indeed, there is a sense in which also what I mean in uttering a sentence (the content of my utterance) and not only the propositional attitude is "present to the mind". Your knowledge of what my words mean has to be based on evidence: you probably assume you have it right, and you probably do. Nevertheless it is an hypothesis. Of course, I may not know what I mean by those words either. But there is a presumption that I do, since it does not make sense to suppose that I am generally mistaken about

4

D. Davidson, "Empirical Content" (1982), repr. in LePore (ed.) Truth and

Perspectives 5

Interpretation:

on the Philosophy of Donald Davidson, Blackwell, 1986, pp. 320 — 332.

Since attitude-authority is treated more fully in Davidson's paper "First-Person Authority" — the emphasis of his later papers being on what I here call content-authority — it is with the earlier paper that I shall mainly concern myself here.

First-Person Authority and Radical Interpretation

201

what my words mean; the presumption that I am not generally mistaken about what I mean is essential to my having a language — to my being interpretable at all. ("What is Present to the Mind?", p. 17). 6

The nature of the "knowledge" alluded to is peculiar for, in Davidson's opinion, it is to be construed neither as a grasp of propositions nor as an (intensional) acquaintance with the events involved in the causal history of my beliefs. In other words, the knowledge of what I mean by an utterance of mine — if knowledge is the word I want — is not to be accounted for in terms of a direct grasp, a grasp unmediated by language, of objects outside or inside my mind. Often the mythology of direct contact has been used by philosophers as a step in the attempt to secure a stock of beliefs free from error. As direct inspection was supposed to guarantee at least in part the freedom from error, such access has been called "privileged". Both adjectives, "direct" and "privileged", are of dubious value inasmuch as their mythological implicatures are difficult both to cancel and to detach. Davidson does not share this mythology. A chief merit of his proposal consists in his attempt to spell out the content of belief without postulating the possibility to grasp bare thoughts and without accepting the idea of a divorce of knowledge of meaning and knowledge of reference, implicit in so-called causal theories of reference. Readers of "The Inscrutability of Reference" may be puzzled by the statement I have just quoted; for does not the indeterminacy of reference threaten the presumption that the content of an averred belief is at all determinate? In that paper we were told: Perhaps someone (not Quine) will be tempted to say "But at least the speaker knows what he is referring to". One should stand firm against this thought. The semantic features of language are public features. What no one can, in the nature of the case, figure out from the totality of the relevant evidence cannot be part of the meaning. And since every speaker must, in a dim sense at least, know this, he cannot even intend to use his words with a unique reference, for he knows that there is no way for his words to convey this reference to another. ("The Inscrutability of Reference" p. 235) 7

This prima facie tension is dispelled by a division of labour among speaker and interpreter; it is the latter who has to figure out the reference of words: "It is not the speaker who must perform the impossible feat of comparing belief with reality; it is the interpreter who must take into 6

7

D. Davidson, "What is Present to the Mind?", in J. Brandl, W. L. Gombocz (eds.) The Mind of Donald Davidson, 1989, Amsterdam, Rodopi ( = Grazer Philosophische Studien vol. 36), pp. 3 - 1 8 . D. Davidson, "The Inscrutability of Reference" (1979) , repr. in Inquiries into Truth and Interpretation, Oxford, O.U.P. 1984, pp. 2 2 7 - 2 4 1 .

202

Eva Picardi

account the causal interaction between world and the speaker in order to find out what the speaker means and hence what he believes." 8 In Davidson's opinion the plurality of interpretations does not make the speaker's utterance vague in the mouth of the speaker. A desk can be measured in inches or in centimeters, but this does not render its length indeterminate; something similar — Davidson argues — applies also to reference. This is in my opinion only partially satisfactory. For, although I may no doubt fail to refer to and fail to have a clear conception of that to which I mean to refer, I cannot be meaning to refer to anything to which I explicitly do not mean to refer — though I can, as Professor Davidson points out in his reply, occasionally be referring to something to which I did not mean to refer9. This point is not new; it was made by Frege in 1892 in a context in which he tried to meet the skeptical challenge how we know that in asserting that the moon is smaller than the earth we are referring to the moon and not to our image of the moon. Frege observes — very reasonably, in my opinion — that "to do so would be flatly to misunderstand the sense. If that is what the speaker wanted he would use the phrase 'my idea of the moon'. [...] to justify mention of the reference of a sign it is enough, at first, to point to our intention, in speaking or thinking — even if with the proviso: if such reference exists."10 As I understand it, this is a very modest claim of first-person authority over the meaning of the words we employ: it only entitles us to make certain disclaimers, while leaving wide open the possibility that we may be speaking of nothing or of something of which we have only a vague idea. However I do not see how this modest claim can be formulated within Davidson's framework. I do not see how, according to Davidson's conception of radical interpretation, it would be possible to articulate the dismay of a speaker who emphatically denies — pace Frege — that by using number-words he meant to refer to extensions of second-level concepts. And yet what difference could be greater! It is hard do sums and multiplications with Grundgesetze^. Nonetheless, if the interpreter chooses to construe the content of the speaker's utterance by appealing to classes of classes or to extensions of concepts, he will have accounted for his dispositions of assent and dissent to various sentences in which number "Empirical Content", cit. p. 332 ' In the original version of this paper the text read "I cannot possibly be referring to anything to which I explicitly do not mean to refer". Not only is this formulation wrong as it stands, as Professor Davidson points out in his reply, but it actually misrepresents what I intended to say. 10 G. Frege, "Uber Sinn und Bedeutung", Zeitschrift für Philosophie und philosophische Kritik, 100 (1892), pp. 2 5 - 5 0 , p. 3 1 - 3 2 ; Translations from the Philosophical Writings of Gottlob Frege by Peter Geach and Max Black, Oxford, Blackwell, 1952, pp 61—2. 8

First-Person Authority and Radical Interpretation

203

words occur and thereby for the content of the speaker's beliefs. To borrow Davidson's simile, the interpreter measures in inches, while the speaker measures in centimeters. The trouble with this picture is not, as Searle for instance has claimed 11 , that the element of intentionality has been disregarded, for mere intentionality has no power to discriminate between useful and useless interpretations. The trouble is rather that the type of "evidence" to which both Quine's radical translator and Davidson's radical interpreter have access is inadequate. In their accounts no role is allotted to such features as, say, the speed and greater reliability of numerical calculations over logical calculations, for these features do not play a role in a statement of truth-conditions. But a conception of cognitive equivalence on the basis of which any two materially equivalent predicates are on the same footing is obviously inadequate to capture the content of a speaker's belief. So much for my reservations about content-authority. Let me turn to attitude authority.

III.

According to Davidson our beliefs about our own propositional attitudes are less liable to error than many of our other beliefs. Not only do we generally know what we believe but we also know whether we are at a given moment believing, doubting or wishing something. This is not, however, knowledge based on evidence or inference, for we do not have toward our own words the same attitude which we have toward other people's words. I do not normally infer my beliefs from my utterances, which is what we tend to do concerning the beliefs of others. We do not guess our future actions from listening to what we say, nor do we treat beliefs as sense-impressions. This thought was forcibly expressed by Wittgenstein in his Philosophical Investigations, in the course of his discussion of Moore's paradox ("It's raining, but I do not believe it"). After having tentatively put forward for consideration the view that belief may be conceived as a disposition of the believing person, Wittgenstein goes on to say: This is shown to me in the case of someone else by his behaviour and by his words. And under this head by the expression "I believe" as well as by the simple assertion. What about my own case: how do I myself recognize my own disposition? Here it will have been necessary to take notice of myself as others do, to listen to myself 11

J. R. Searle, "Indeterminacy, Empiricism and the First Person", The Journal of LXXXIX (1987), pp. 1 2 3 - 1 4 6 .

Philosophy,

204

Eva Picardi talking, to be able to draw conclusions from what I say! My attitude to my own words is wholly different from other people's" (II, X) 1 2

This attitude to our own words which is "wholly different from other people's" is what Davidson has in mind — if I understand him correctly — when he contrasts the speaker's attitude and the interpreter's attitude toward an utterance of the speaker. Davidson, in contrast with Wittgenstein, is not content with registering a difference, but interprets the difference by appealing to an asymmetry in the presumption of freedom from error. He therefore considers the speaker's position "privileged" and not merely different from that of the interpreter. Let us look more closely how and where the implication of privilege intervenes. According to Davidson, philosophers have brought to bear on the issue of the ascription of propositional attitudes Wittgenstein's remark that we do not recognize our own sensations from anything, we simply have them. Then, having drawn a parallel between self-ascriptions of beliefs and of sensations, they have declared that in both cases we are dealing with knowledge not based on evidence. The question then arises why a claim not based on evidence should carry more weight than one based on evidence. An obvious objection which might be raised against the proposed parallel between the case of sensations and the more general case of knowledge of meaning is this. Whereas if I am queried as to the criterion of the redness of my image I can answer "none" and still be using the word appropriately, in the case of someone who asks me which is my criterion for knowing that I am conveying a belief and that I understand which belief I am conveying, no sweeping reply along those lines would do. It would of course be ridicoulus if I answered that query by saying: "I am generally reliable, and when I say something I mean it: let us see how I am going to act on what I have said and we'll find out what I believe." In saying that I would be taking toward my own words the same attitude I occasionally take toward other people's words, and for reasons which we have not yet spelled out, this will not do. Luckily I can give a better answer and say: "I speak English". In offering this answer I would be adverting to the institution of a common language, to the accepted meanings of words and to the many rules and conventions to which I hold myself responsible when speaking to others. My training as a member of a linguistic community would be the sort of "evidence" I would be gesturing to. A murky sort of evidence, perhaps, but better than an ineffable intention or intuition. 12

L. Wittgenstein, Philosophical Anscombe and R. Rhees.

Investigations,

Oxford, Blackwell, 1953, edited by G. Ε. M.

First-Person Authority and Radical Interpretation

205

Such an objection to the parallel between ascriptions of sensations and ascriptions of beliefs is not open to Davidson and in fact he does not raise it, for he holds or held that speaking a language is a condition of having conventions, and not the other way round. What prevents Davidson's account of the speaker's "authority" over his own understanding of the language from collapsing into the authoritarian one given by Humpty Dumpty is the realization that speech and hence meaning presuppose partnership: "before anyone can have thoughts there must be another creature (one or more) interacting with the speaker". 13 Since Davidson makes use of the contrast between knowledge based on evidence and knowledge which is not based on evidence for describing the issue of first-person authority, the question arises how the presumption that the speaker is not mistaken is to be accounted for, since obviously "claims not based on evidence do not in general carry more authority than claims that are based on evidence, nor are they apt to be more correct." ("FirstPerson Authority", p. 103) As I said, in Davidson's opinion both Strawson and Wittgenstein merely registered certain asymmetries in the application of psychological predicates but failed to give an explanation of first person authority. And yet Davidson's account of it and Wittgenstein's discussion of the first and third person of the verb "to believe" in the (so-called) second part of the Philosophical Investigations exhibit a certain similarity which is worth exploring. Both ask themselves a question about sameness of content. Davidson asks: whence do we draw the assurance that when I say "I believe that Wagner died happy" and when Ernest says "Davidson believes that Wagner died happy" Ernest's utterance matches mine as regards content as well as attitude? When asking this question Davidson is not alluding, I suppose, to the remote possibility that under the effect of a strong drug he may not know that he is Davidson and thus fail to assent to Ernest's statement "Davidson believes that Wagner died happy" after having himself asserted "I believe that Wagner died happy". The question he is asking concerns the interpreter's possibility of rendering a speaker's utterance: The question then comes to this: what explains the difference in the sort of assurance you have that I am right when I say "I believe that Wagner died happy" and the sort of assurance I have? We know by now that it is no help to say I have access to a way of knowing about my own beliefs that you do not have nor that we use different concepts of belief (or the word "believes"). So let us simply consider a shorter utterance of mine: I utter the sentence 'Wagner died happy'. Clearly if you or I or anyone knows 13

"The Conditions of Thought", in J. Brandl, W. L, Gombocz (eds.), The Mind of Donald Davidson, 1989, Amsterdam, Rodopi ( = Grazer Philosophische Studien vol. 36), pp. 1 9 3 - 2 0 0 , p. 198.

206

Eva Picardi that I hold this sentence true on this occasion of utterance, then she knows what I believe — what belief is expressed. ("First-Person Authority", p. 109)

But this is a bit fast. For, what entitles the interpreter to choose Davidson's shorter utterance instead of the longer one to convey Davidsons' belief? Obviously the two sentences are not interchangeable in all contexts. One of the reasons why the interpreter can make this short-cut is that, in Davidson's opinion, he is supposed to be able to tell whether a speaker holds a sentence true without knowing the meaning of the sentence uttered and, a fortiori, without having identified the words indicating propositional attitudes. Now, quite independently of the perplexities — which I voiced elsewhere14 — as to the possiblity of identifying a speaker's truth-directed intention, it seems to me that in the context of the present discussion which focuses on utterances issued with special authority we cannot afford to let the verb "believe" disappear without a word of comment. Indeed we should pay closer attention to the truly remarkable move that allows the interpreter to use the shorter sentence instead of the longer one thereby using attitude-authority as a lead to content-authority. I am not here questioning the legitimacy of that short-cut; on the contrary, I am asking what makes it possible.15 In section X of the second part of his Philosophical Investigations Wittgenstein asks the question: How is it possible that "It is raining" and "I believe it is raining" can as isolated assertions fulfil the same function, namely that of a report on the weather, whereas when they feature as suppositions, for instance in the antecedents of conditionals, they cease to be interchangeable and mean different things? For "If Wagner died happy, there must be some evidence in Cosima's diaries" is in Davidson's mouth obviously not equivalent to "If I believe that Wagner died happy, there must be some evidence in Cosima's diaries" and Davidson himself knows that. How is the interpreter going to handle these different embeddings? The suggestion that "I believe" has two different meanings, one in the supposition and one in the assertion, is to be resisted: the verb "believe" works in different ways in different contexts and the question is whether and how these uses are related. Let us think of conditionals appropriate 14

15

Cfr. E. Picardi, Davidson on Assertion, Convention and Belief, in J. Brandl, W. L. Gombocz (eds.) The Mind of Donald Davidson, 1989, Amsterdam, Rodopi ( = Grazer Philosophische Studien, vol. 36), pp. 9 7 - 1 0 8 . In his reply Professor Davidson says that no such short-cut was intended; he had only chosen a simpler sentence. If this is the case, however, he has given us no account of the relationship between "I believe" and "You believe" and hence no hint how the skeptical challenge as to the sameness of mental attitude in self-ascriptions and otherascriptions could be met. In what follows I try to fill the gap by suggesting how, in the case of belief, a fuller account of the intermingling of attitude-authority and contentauthority might help to defuse the skeptical challenge.

First-Person Authority and Radical Interpretation

207

to these two different uses. For instance, to elaborate on Davidson's own example, we may try something like: "If I believe that Wagner died happy, I should think the same of Mozart". Conditionals of this kind can be viewed as a way of articulating a general policy of judgment, checking the consistency of a belief with respect to other beliefs, figuring out what consequences the acceptance of a certain statement would commit me to. In so doing I am looking at the content of my own belief from the point of view of another person, who, in order to figure out what I mean by "happy" in that context may look for other relevant information. Was I thinking of Parsifal or of Carrie Pringle, Wagner's last love? Was I expressing my admiration for Parsifal or disclosing my romantic cast of mind? Was I thinking of a happy composer or of a happy lover? The rest of the conversation is very likely to settle this point. Conditionals like the one just mentioned admit shortening and can without much violence be reformulated thus: "If Wagner died happy so did Mozart" — this is what I believe, "A man who has achieved something is likely to die happy, I am sure". However, when we look for an antecedent of a conditional that matches in content the assertion while at the same time preserving a trace of our own attitude toward the content we catch ourselves entangled in conditionals like "If I believe that Wagner died happy, I am likely to have good reasons — I always think before judging" or "If I believe that Wagner died happy, I am bound to disagree with Cosima", which sound at best awkward. Their third-person counterparts, on the other hand, "If he believes that Wagner died happy, he is likely to have good reasons", "If he believes Wagner died happy, he disagrees with Cosima" sound alright. In the antecedents of those awkward conditionals we are trying to depict a situation in which we put forward an assertion while in the same breath pausing to consider our attitude. The maximum of awkwardness is reached in the case of Moore's paradox: one makes an assertion and then one takes a distance from it, indeed, one goes so far as to deny that one believes what one has just said: "Wagner died happy. I don't believe it." 1 6 We have a self-defeating utterance, comparable perhaps to

16

In his Remarks on the Philosophy of Psychology (Oxford, Blackwell, 1980, edited by G . E . M . Anscombe and G . H . von Wright), vol. 1 (490—93), Wittgenstein describes the quandary thus: " T h e paradox is this: the supposition may be expressed as follows: 'Suppose this went on inside me and that ouside me' — but the assertion that this is going on inside me asserts: this is going on outside me. In the supposition the two propositions about the inside and the outside are quite independent, but not as assertions." (490); " S o it looks as if the assertion Ί believe...' were not the assertion of what the supposition supposes."(493). (I have modified the English translation slightly.). For a discussion see N. Malcolm,"Disentangling Moore's Paradox" (1989) (Forthcoming), and J . Schulte, " ' E s regnet, aber ich glaube es nicht'. Zu 'Philosophische Untersuchungen II.x'", Teoria, 1985, 187 — 204. ( = McGuinness, A. Gargani (eds.) Wittgenstein and Contemporary Philosophy).

208

Eva Picardi

"I can utter no word of English", but more intriguing, at least for those who hold that the notion of belief plays a crucial role in an account of meaning. But why is it possible to make something of conditionals like "If I believe that Wagner died happy, I should think the same of Mozart", whereas conditionals like "If I believe that Wagner died happy, I am likely to have good reasons" sound definitely awkward? Here is a suggestion: our understanding of conditionals of the first type rests on the fact that we are conversant not only with the use of the verb "to believe" in the first person present tense but with a good deal more, indeed with the whole grammar of the word "believe". We understand the assumption embedded in the antecedent of the conditional because we know what it is like to ascribe beliefs to our fellow creatures. This is an understanding which is not exhausted by the observation that to assert "p" and to assert "I believe that p" can often amount to the same thing — the assertibility contitions of the two sentences are the same. Our grasp of the words "I believe" as they occur in the antecedents of conditionals is derived from our practice of ascribing beliefs to other people, an understanding which we can then derivately — with more or less strain — apply to our own statements. As Strawson says, it is an essential feature of the use of Ρ [person]-predicates that they should be both self-ascribable and otherascribable for the same individual. "It is not" — Strawson argues — "as if these predicates have two kinds of meaning. Rather it is essential to the single kind of meaning that they do have, that both ways of ascribing them, should be perfectly in order." 17 . Wittgenstein's analysis, like Strawson's, rests on the insight that it is a necessary condition of one's ascribing states of consciousness to oneself that one should also ascribe them, or be prepared to ascribe them, to others. It is not as if such a perspective were foreign to Davidson; in fact in a different context but in not too dissimilar a vein he writes: The contrast between the grounds a self-ascriber has for his self-ascription, and the grounds an interpreter has for accepting that same ascription would be stark if we were to assume that no question can arise concerning the speaker's interpretation of his own words. But of course it can, since what his words mean depends in part on the clues to interpretation he has given the interpreter or that he justifiably believes the interpreter has. ... The asymmetry rests on the fact that the interpreter must, while the speaker doesn't, rely on what if it were made explicit would be a difficult inference in interpreting the speaker. ("First-Person Authority", p. I l l )

A speaker in order to know what he means by an utterance, need not assume the stance of the interpreter he is addressing, let alone that of the set of potential interpreters he may be addressing at different times on 17

P. F. Strawson, Individuals,

1959, London, Methuen, p. 110.

First-Person Authority and Radical Interpretation

209

different occasions. And the thought that there is a variety of interpretations of what he says should not shake his confidence in the univocity of what he says; yet this thought is at the back of his mind and, provided that he wants to be interpretable, he will try to convey what he wants to say in such a way as to make it as unambiguous as possible from the point of view of a potential interpreter. The point I have being trying to make is not that interpreting speech presupposes mastery of P-predicates and not the other way round; I am not chiefly interested here in the issue of conceptual priority; it is rather that once the peculiarity of P-predicates is appreciated, the emphasis on first person authority loses much of its point and, in the case of belief, may turn out to be misleading. It is the fact that the expression of first person authority in sentences like "I believe that Wagner died happy" can disappear which makes it possible that the speaker's words can serve as a vehicle of information not about his own inner state, but about the world. That which makes it possible for you to use my shorter sentence in order to reproduce what I want to say is precisely the fact that the ingredient of self-ascription has shrunk to nil. That by "It is raining" and "I believe that it is raining" I can say the same, namely that it is raining, is due to the fact that in neither case am I self-ascribing anything, except, possibly, the "I believe" which accompanies all my utterances.

Donald Davidson: Reply to Eva Picardi When I heard this paper read in Bielefeld I said I was uncertain whether there was anything in it I disagreed with; on careful reading I must now confess that I am uncertain whether there is much I agree with. Since I have found it difficult to be sure exactly where Eva Picardi stands on several crucial points, I shall set out some of my own views which others may (as I do) judge to be at odds with hers; but I shall refrain from trying to establish where and how we differ. I take it to be a fact beyond dispute that there is a basic difference in how (usually) a person knows what she believes and how others (always) know this when they do. (I shall speak of belief throughout, understanding that what I say goes for all the propositional attitudes.) Perhaps the best way to state this difference is to say that knowledge of one's own beliefs is not generally based on evidence or observation, while knowledge of the beliefs of others always is. I take this difference to be intimately tied to the fact that in general there is a presumption that people are not mistaken about their own attitudes, while there is no such presumption in the case of others. But whether or not one thinks there is such a presumption, there is an asymmetry to be explained due to the point about evidence. Perhaps I should add that the existence of a presumption that one is right is not the same thing as being more certain that one is right or being more certainly right. These remarks were epistemological. Now I turn to the sentences we use to attribute attitudes. I take it that it is a fact that people have beliefs, that we have linguistic resources for attributing beliefs, and that such attributions are sometimes true and sometimes false. Jones can attribute the belief that Samos is near Turkey to Jenkins by saying "Jenkins believes that Samos is near Turkey". It would be amazing if there were no way for Jenkins to attribute the same belief to himself, presumably by saying "I believe that Samos is near Turkey". If there were no way of attributing a belief to oneself, we surely would invent one. There also are good, simple reasons why sentences that begin "I believe that" can be used in this way. Given the obvious semantic rules for the pronoun "I" and predicates like "believes that Samos is near Turkey", an utterance of the sentence "I believe that Samos is near Turkey" must be true if and only if the speaker believes Samos is near Turkey unless there is a basic

Reply to Eva Picardi

211

ambiguity in the pronoun or in the predicate. As far as I know, no one has thought it useful to suggest this. But if the semantics of such sentences are regular, then there must be a straightforward use of the sentence to ascribe a belief to oneself. This is, of course, perfectly consistent with there being uses of such sentences to perform many other speech acts: for example, when Jenkins says "I believe that Samos is near Turkey" he may be doing no more than asserting that Samos is in Turkey. Explaining this does not require that the sentences "Samos is near Turkey" and "I believe that Samos is near Turkey" have the same truth conditions, or that sentences beginning "I believe that" are never used to make self-ascriptions. If the points made in the last two paragraphs are right, it follows that speakers stand in a different epistemic relation to the literal truth of sentences like "I believe that Samos is in Turkey" than they do to sentences like "Jenkins believes that Samos is in Turkey" (at least when the speaker is not Jenkins). This is why I claim that there is a presumption that sincere, literal, assertive uses of "I believe that"-sentences are true. It is such uses only that I had in mind in "First Person Authority". I did not intend the word "authority" to add any weight or substance to the idea of a presumption of truth. The "asymmetry" I made so much of in my paper can't, then, be explained simply by the fact that there exists a non-self-ascribing use of "I believe that"-sentences. Nor can simply describing belief-predicates as being ascribed on the basis of evidence to others and without evidence to oneself explain why this does not imply that such predicates are ambiguous. (This "solution" of Strawson's does, however, make clear that Strawson's view of the asymmetry, like mine, is based on an epistemic asymmetry, and not on the idea that there are no self-ascriptions.) My main aim in "First Person Authority" was to throw light on an epistemic problem, not a linguistic one. It seemed to me that in their concern with the problem of "other minds", philosophers had overlooked a more basic problem, namely the asymmetry between the kind of knowledge that we have of our own thoughts and the kind of knowledge that we have of other people's thoughts. I argued that even if we have a good answer to the question how we know what there is in other people's minds, something needs to be said to explain the asymmetry. I thought (and think) Strawson and Wittgenstein had described the asymmetry, but had done nothing to explain it. The explanation I offered did not depend on "slipping" from belief sentences to the sentences conveying the contents of such sentences, nor of slyly "substituting" one for the other: I was as clear then as I am now about the difference. What I claimed was that in any normal utterance there is a presumption that the speaker knows what he means, while there

212

Donald Davidson

is no such presumption involved in our understanding of others. (The word "knows" here is not essential in making the point; it would do as well to say the speaker is not misusing his own words, or that he means, in a literal sense, what he says.) This presumption in turn I contended was essential to communication, on the grounds that we must depend on accepting a speaker's actual linguistic behavior as the basis for interpreting his words. This can be seen from the fact that it makes no sense (in general) to ask how we interpret our own words, but it is a serious question how we understand others. ("My sentence 'Snow is white' is true if and only if snow is white" is a tautology; "Your sentence 'Snow is white' is true if and only if snow is white" is not.) The answer to the question how I know what I believe is (aside from special cases) as empty or trivial as the question how I know what I mean. Thus I attempt to explain the epistemic asymmetry in the case of thoughts ("the problem of other minds") by reference to the obvious, and necessary, asymmetry in the case of interpretation. Perhaps I should add that I am much more confident that there is a problem here than that I have solved it. [In response to a question:] It's true that I didn't respond to the question how I reconcile my acceptance of the doctrine of the inscrutability of reference (which I take, with modifications, from Quine) with my conviction that we mostly know what we think and mean. In my opinion there is no conflict. To say reference is inscrutable doesn't mean something is there which it is impossible to detect; it means there is nothing there to scrute. So in particular it doesn't mean that there is something I know about my words which you can't figure out; there is no such thing for anyone to know. In the case of reference, the claim may be put this way: if there is one right way of mapping "referring" words (singular terms, predicates) on to objects which gives the truth conditions of all sentences correctly, then there are endless other ways of mapping the same expressions onto the same entities which are equally correct. There is no point picking out one of these ways as the one that defines reference, since reference is only a theoretical concept whose job is to give the truth conditions of sentences. I no more can determine which is the right mapping for myself than I can for you, or than I can determine whether lengths come in inches or in centimeters. What may fool you into thinking I know that my word "rabbit" refers to rabbits and not to the members of some equinumerous class is that there is nothing else I can say without confusing rabbits with those others. No correct interpretation of my words should make me identify things I can keep separate. But inscrutability leaves all distinctions where they were, since it leaves truth conditions unchanged.

The Explanation of First Person Authority BERNHARD THÖLE

Philalethe: II n'est pas aise de concevoir qu'une chose puisse penset et ne pas sentir qu'elle pense. Theophile: Voilä sans doute le noeud de l'affaire et la difficulte qui a embarasse d'habile gens. Leibniz: Nouveaux Essais. II.1.11

I know my mind better than you do. You know your mind better than I do. Something like this seems to be true. What is the explanation? In his paper on "First Person Authority", Davidson criticises a number of attempted explanations and offers one (or two) of his own. According to Davidson, the Cartesian fact has a non-Cartesian explanation: first person authority is "built into the nature of interpretation" (FPA 111). I shall in this paper criticise Davidson's explanation and defend what I shall call the traditional account against his criticism. I begin by assembling Davidson's scattered remarks about what first person authority comes to, and make some critical remarks about them. Then, I shall propose a minimal conception of first person authority that any explanation should account for (I). In the second section, I shall try to defend the traditional account against Davidson's objections (II). In the last section, I turn to Davidson's own account (III). There I shall argue that his explanation applies only to a very limited aspect of the problem, and that it is difficult to see how it could be extended to account for other aspects and other cases of first person authority.

I. What is first person authority? What is first person authority? Like most philosophers today, Davidson rightly insists that the claim to first person authority should not be identified with the (supposedly) Cartesian claim that self-knowledge of our mental states is incorrigible and self-intimating: Not everything we believe about our own minds is true and not everything that is true of our mental states is known or even believed by us. This becomes evident

214

Bernhard Thöle

once we realize that some mental states are dispositional and that some mental states are unconscious. Furthermore, even if we restrict the claim to first person authority to present occurrent conscious mental events, self-knowledge may not be infallible. What then does first person authority amount to? According to Davidson there is a presumption that present tense self-ascriptions of mental states are not (in general) mistaken. 1 Although it is possible for others to discredit my self-ascriptions 2 this cannot happen generally: Usually the speaker is not mistaken when she sincerely self-ascribes mental states of a certain class. This is — as Davidson acknowledges — a rather vague description. Perhaps it is not possible — for reasons that will be mentioned later — to eliminate this vagueness altogether. But something more can and must be said. We should, at least, address the following questions: (a) a first preliminary question is: 'What are first person self-ascriptions?'; (b) the second question is 'In what does first person authority consist?'. Since first person authority does not extend to all mental states, and where it does it does so to various degrees, we have to address a third question (c): 'What is the range of mental states to which first person authority (primarily) applies?'. These questions are not altogether independent from each other. I shall begin with some comments on the first question·. It should be obvious that not all cases in which a person ascribes a mental state to himself are even prima facie candidates for first person authority, since they are not first person self-ascriptions in the intended sense. For sometimes the speaker self-ascribes states on the basis of publicly available evidence only — the very same sort of evidence that could be used by another person to ground his corresponding third person ascription. In such cases there is no reason why another person might not be in an even better position to determine the speaker's mental state. But there is a class of self-ascriptions to which first person authority does apply, although there may be no simple way to demarcate this class without presupposing a bit of controversial theory. One thing that is characteristic of these self-ascriptions is that they do not rely on publicly available evidence. 3 This in itself is certainly not sufficient to delineate the

1 2 3

Cf. FPA 101. Cf. FPA 103. Of course there are mixed cases: self-ascriptions that do rely on publicly available evidence, but do not exclusively rely on them. These are subject to first person authority, though only to a limited degree.

The Explanation of First Person Authority

215

intended class since it excludes neither random guesses nor lies. Some positive account of what is characteristic of self-ascriptions belonging to the intended class would be desirable. But familiar attempts to do this tend to smuggle controversial elements into the description. 4 Although there seems to exist no simple theory-neutral positive description, I think that the above remarks suggest a workable conception of the intended class of first person self-ascriptions: it should only contain sincere self-ascriptions that are not (exclusively) based on publicly available evidence and do not result from mere guesses and the like. In what follows, I shall restrict the term 'first person (self-)ascriptions' to those of the intended class. Correspondingly, I shall use the term 'third person ascription' for those ascriptions of mental states that depend on such publicly available evidence, on which other people typically rely. Let us now turn to our second question·. 'What is first person authority?'. I shall begin by listing Davidson's remarks on this question and shall make a few comments on them. I shall add further specification when I turn to the third question. Introducing first person authority, Davidson claims that when a person self-ascribes a mental state 5 "there is a presumption that he is not mistaken" (FPA 101). Let us call this the truth presumption thesis·. (TT)

There is a presumption that first person self-ascriptions are in general, though not universally, true.

Now, since Davidson has argued elsewhere 6 that a similar claim holds for beliefs in general (not just for first person second-order beliefs), (TT) needs further specification. We must add something to (TT) to bring out what is special about first person authority. 7 What we need is supplied by a second claim: the asymmetry thesis·. 4

5

6 7

To say that these self-ascriptions depend on "our 'inner' awareness of our mental states", whereas third person ascriptions rely on "their 'outer' manifestations" (cf. C. McGinn (1982) p. 6), seems to invoke the notion of introspection that many philosophers reject. The suggestion that first person self-ascriptions of mental states (causally) depend on the existence of the corresponding states seems to exclude (some types of) erroneous selfascriptions and does not exclude many cases of third person ascriptions. If we try to exclude guesswork by restricting the class to self-ascriptions that — although not based on publicly available evidence — neverless constitute knowledge, we would not only prejudge the issue against those Wittgensteinians who emphatically deny that title to any first person self-ascription, we would also make first person ascriptions true by definition. This needs qualification since not all sorts of mental states are ascribable with special authority. The necessary qualifications will be introduced in our discussion of the third question. Cf. Davidson (1977) and (1983). Cf. Shoemaker (1990) p. 194.

216 (AT)

Bernhard Thöle

First person self-ascriptions have (in general) more authority than the corresponding third person ascriptions. 8

(TT) and (AT) constitute the core of Davidson's characterization of first person authority. But there are some further claims that should be mentioned. The first of these consists in the conjunction of two subtheses: Davidson claimes — (TT) and (AT) notwithstanding — that first-person self-ascriptions may be erroneous (this qualifies (TT)) and that such selfascriptions are not incorrigible (this qualifies (AT)). Let us call these the error- and the corrigibilitj-theses·. (ET) (CT)

First person self-ascriptions might be mistaken. "it is possible for the evidence available to others to overthrow self-judgements" (FPA 103).

The second supplementary thesis is that a person cannot lose his special authority "even when his claim is challenged or overturned" (FPA 104; cf. 103). It is not altogether clear what this exactly means, but I suspect that at least the following two claims are intended: (RT)

(AT) and (TT) do not simply hold for first person ascriptions in general but also for first person ascriptions of every particular person.9

Let us call this the restriction-thesis. thesis·. (WT)

The second claim is the

special-weight-

Even in those cases where self-attributions are corrigible, the selfattributer "speaks with special weight" (FPA 103). 10

Two further theses about the status of first person authority play a chararcteristic role in Davidson's account. I have already mentioned one of these: I shall call it the interpretation thesis: (IT)

First person authority "is essential to" or "built into the nature of interpretation" (FPA 102,103, cf. KOOM 457).

Strictly speaking, (IT) does not belong to the characterisation of what first person authority is, rather it belongs to an account of why it holds. 8

5

10

Cf. FPA 101 f., 109. It should be noted that (TT) and (AT) are independent of each other, since claims that have (in general) more authority than others are not, therefore, in general true. (TT) and (AT) do not rule out the possibility that for a minority of people all of their self-ascriptions are false (or corrigible), if most self-ascriptions of the majority of people are true. (RT) excludes this possibility. This again is quite vague, but that need not be a disadvantage. In what follows, we shall encounter a more precise — but also more restrictive — variant of (WT) (and (AT)) proposed by Ayer: a person's self-ascriptions are corrigible, if the person himself immediately accepts the correction (cf. Ayer (1963) p. 73).

The Explanation of First Person Authority

217

But it has a consequence that (if true) would tell us something important about the status of first person authority and at the same time would show that any attempt to explain first person authority must meet a certain condition of adequacy. The consequence is that first person authority is not just a contingent fact but is, in a sense, necessary. Let us call this the non-contingency-thesis: (NT) First person authority is not just a contingent fact. According to Davidson, then, any explanation of first person authority must account for this special status. This comes out clearly in his criticism of Shoemaker's explanation 11 . Shoemaker's claim is, roughly, that we treat first person self-ascriptions as authoritative because of a convention of language. On the face of things, Davidson's objection that this 'explanation' is not really explanatory is surprising: Seen from the point of view of the interpreter this means the interpreter should interpret self-ascriptions in such a way as to make them true — or to assign a special priority to their truth. [...] our only reason for saying the speaker has special authority on occasion is that we are prepared to treat his utterance as a self-ascription. In other words, self-ascriptions have special authority: true, and that is where we began. (FPA 108)

This criticism seems unfair: why should the claim that it is a convention of language to treat first person ascriptions as authoritative not be explanatory? If we ask for an explanation why people in Britain drive on the left side of the road, we should be content when told that this is required by the traffic regulations. Of course we might ask further, why the regulations are as they are. But that is a different question. Similarly, referring to a convention of language would explain why we treat first person selfascriptions as authoritative. And to the further question why we have this convention, the only answer might be that this is just a brute fact. The reason why Davidson is not satisfied with Shoemaker's explanation then seems to be that he presupposes (NT) and requires of an explanation of first person authority that it should account for it. 12 But it is certainly not obvious that (NT) holds. 13 Thus, I think it would be rash to treat 11

12

13

Davidson refers to Shoemaker's position in Shoemaker (1963). Shoemaker has recently given a different explanation that is in some ways closer to Davidson's own account, although it also differs in important respects (cf. Shoemaker (1991) esp. sections IV-VI). Given the 'transcendental' character of Davidson's own explanation of first person authority (according to which first person authority is a necessary presupposition of interpretation), it is clear that he is commited to (NT). But this does not mean, that he takes (NT) as an adequacy condition that any successful explanation of first person authority must meet. M. Otto pointed out to me, that Davidson's criticism of Shoemaker's explanation commits him only to the more plausible claim that a successful explanation of first person authority should not only explain why we treat first person self-ascriptions as authoritative, but that we are justified in treating them thus. Whether Davidson's arguments for (IT) and thus for (NT) are convincing will be discussed in the last section of this paper.

218

Bernhard Thöle

(NT) as an adequacy condition for a successful explanation of first person authority.14 This completes my account of Davidson's characterization of firstperson authority. Before I discuss this characterization, I turn to our third question. Which mental states and events can be self-ascribed with first person authority? Not every sort of mental state can and, moreover, first person ascriptions of mental states exhibit first person authority in various degrees depending on the nature of the ascribed states (cf. FPA 102). A number of distinctions are relevant here: (a) the distinction between present and past (and future) states (b) the distinction between occurrent and dispositional states (c) the distinction between conscious and unconscious states (d) the distinction between states that are essentially (causally) connected to something about which we do not have special authority 15 and those states that are not so connected and (e) the distinction between sensory states and propositional attitudes. 14

Davidson is not alone in taking (NT) as an adequacy condition. J. Heil, for example, claims that the traditional explanation (according to which first person authority derives from the fact that the person, but nobody else, has direct knowledge of this person's mental states) cannot account for the privileged status of first person self-ascriptions, since the "access I enjoy to my own mental contents would be superiour to what is available to you, perhaps, but only accidentally so [my emphasis...] If I could look over your mind's shoulder, then my epistemological position would be no different from yours as you gazed inwardly" (Heil (1988) 248 f.). This relies on the — unargued — adoption of a thesis similar to (NT) (cf. Heil (1988) p. 240f). But it is not obvious that first person authority might not rest on some contingent, though real, limitation of our cognitive capacities. Thus, I think we should not rule out from the beginning a theory according to which first person authority might cease to hold when we develop cognitive capacities that would allow us to 'look over other minds' shoulders'. On the other hand, it is far from obvious that (as Heil assumes) an advocate of the traditional account must hold that you would lose first person authority, if I were to develop such cognitive capacities. For even if — to use Heil's example — I were "wired to you in such a way that I share your nervous system" (p. 240), this would not guarantee that I were then directly conscious ofjyour mental states. Thus first person authority might survive even in such a case. As Ayer puts it: "If someone else who claimed to be co-conscious with [the person who has a certain sensation], gave a different report, we might suspect that our subject was lying about his sensations, but not that he was mistaken: if we believed him to be honest, it is the other's claim to co-consciousness that would be discredited" (Ayer (1963) p. 69).

15

On most functionalist accounts of mental states there are no mental states of the former class. I cannot here explore the question whether the adoption of functionalism puts severe restrictions on the degree to which first person authority applies even to presently occurring conscious mental states. For a pessimistic assessment cf. Boghossian (1989) p. 14 f. Shoemaker, on the other hand, attempts to explain first person authority of beliefascriptions on the basis of "the functional nature of the contents believed, i. e., the mental states ascribed" (Shoemaker (1990) p. 210). But Shoemaker notes that this account of the authority of belief-ascriptions cannot be easily extended to cover ascriptions of other mental states.

The Explanation of First Person Authority

219

Special authority of the highest degree applies to first person ascriptions of present occurrent conscious mental states 16 . Let us call this the privileged class of mental states. Often, discussion of first person authority is restricted to a consideration of present occurrent conscious sensations and thoughts. This seems a reasonable starting point, since being equipped with a description and explanation of first person authority with respect to the privileged class we only need further information about (a) the reliability of memory, (b) the connection between occurrent and dispositional states, (c) the connection between conscious and unconscious states and (d) the functional dependence of mental states to account for those mental states that do not belong to the privileged class. Davidson's procedure, however, is different, for he is primarily interested in ascriptions of propositional attitudes. He claims that, although there are differences of degree and kind, "all propositional attitudes exhibit first person authority". But "special authority attaches directly to claims about [...] belief' (FPA 102). Thus we have the be lief-thesis: (BT)

First person authority (in the sense of (TT), (AT), (ET), (CT), (RT) and (WT)) applies to self-ascriptions of belief.

Nevertheless, although Davidson considers first person authority only as it applies to the propositional attitudes, he claims that "what holds for the propositional attitudes ought, it seems, be relevant to sensations and the rest" (FPA 102). This suggests that he thinks that his account can be extended to cover the elements of the privileged class. I shall now make some critical remarks on Davidson's account of what first person authority is. Let us first look at (TT), the claim that first person self-ascriptions are presumed to be generally true. This claim is ambiguous. We can distinguish a de facto and a de jure interpretation: 17 on the de facto interpretation it is only claimed that we in fact treat first person self-ascriptions as generally true. On the de jure interpretation the claim is that we not only treat them thus but are justified in doing so, since they are in general true. On the de

16

I am uncertain whether the degree of authority of ascriptions of present occurrent conscious mental states varies depending on whether they are propositional attitudes or sensations. Perhaps the complexity of propositional attitudes diminishes the degree of authority of their self-ascriptions. On the other hand, one might argue that the fact that the content of propositional attitudes is already conceptualized reduces the possibility of error in identifying these states, whereas we might misidentify our sensations. But even if this be true, we might analogously misidentify the nature (though not the content) of the attitude. Whether or not the distincion between sensations and propositional attitudes is mirrored in the degree of authority of their respective self-ascriptions, there might be a difference in the explanation of first person authority between both cases.

17

Cf. Alston (1976) p. 262.

220

Bernhard Thöle

facto interpretation (TT) seems safe18, but Davidson clearly subscribes to the de jure interpretation19. Since there are reasons to doubt that the de jure version of (TT) is true20, I propose a weaker claim. If it should turn out that Davidson's arguments support the stronger claim, then it is all the better, but we should not presuppose the truth of the de jure interpretation (TT) from the start.21 The weaker claim is just a variant of (AT), the thesis that first person ascriptions are in general more authoritative than their corresponding third person ascriptions. But let us first note that here too we have to distinguish between a de facto and a de jure interpretation. In its de facto interpretation, (AT) should be uncontroversial. But I think we can also — perhaps with some caution — assume that an explanation of first person authority should account for (AT) in its dejure interpretation. Otherwise the whole attempt at such an explanation would lose its interest. Only if all attempted explanations fail, would we have reason to doubt that we got the explanandum right. Thus, I think that an explanation of first person authority should at least account for: (A)

We are justified in treating first person self-ascriptions as (in general) more authoritative than their corresponding third person ascriptions. 9 9

An explanation of (A) is therefore the minimum we should expect from any acceptable account of first person authority. In what follows, I shall refer to (A), and not to (AT), as the asymmetry-thesis. 18

19

20 21

22

Even here, however, there is room for doubt. Mackie suggests that we should not claim that first person self-ascriptions (of the privileged class) are in general true but only that they cannot conclusively be shown to be false (Mackie (1963) p. 26 f.). Cf. Davidson (1989a) p. 194. It should be noted, however, that when he gives his own explanation of first person authority, it turns out that Davidson's argument addresses only the thesis that a speaker normally knows what he believes. This does not guarantee — as we shall see (cf. pp. 239 f.) — that most of his sei {-ascriptions are true. Cf. the discussion of Mackie's position below. The distinction between the de facto and the de jure interpretation should not be confused with the distinction between the claim that first person authority holds as a matter of brute fact on the one hand and the opposed claim that it is in some sense necessary. It might, e.g., be the case that there is no alternative to accepting first person self-ascriptions although there is no epistemic justification for this practise. Thus a realist about mental states might grant Davidson's thesis that interpretation requires general acceptance of first person self-ascriptions and, nevertheless, deny that this implies that such ascriptions are generally true. Only on the further assumption that beliefs are theoretical constructions of an ideal interpreter would it follow that they are in general true. In introducing his own explanation, Davidson distinguishes two related asymmetries. One of these is (A), the other is the asymmetry between the warrant of the self-ascriber that his self-ascription is true and the warrant of an interpreter that this self-ascription is true. Davidson adopts the second version because the first version faces the difficulty of deciding which third person ascriptions correspond to a given first person ascription (cf. FPA 109). As I shall explain in the third part of this paper (cf. Fn. 72), I think this option unnecessarily threatens his own explanation.

The Explanation of First Person Authority

221

What about (ET) and (CT)? Some people still claim that self-ascriptions of the privileged class are infallible or at least incorrigible. Ayer defends a slightly weaker condition: first person self-ascriptions are corrigible, if the subject himself is immediately willing to correct himself. But "if others were to contradict him we should not be entitled to say that they were right so long as he honestly maintained his stand against them"23. If the selfascriber is honest, and linguistic mistakes or inattention are ruled out, first person ascriptions of the privileged class cannot be overthrown. I shall call this the condition of weak incorrigibility. Although I think this is an attractive description,24 I shall not assume it here25. Whether or not we require something like weak incorrigibility for selfascriptions of the privileged class, (ET) and (CT) are certainly plausible when we consider self-ascriptions of beliefs. Since beliefs are — as Davidson acknowledges — "dispositions that manifest themselves in various ways, and over a span of time" (FPA 103), they do not belong to the privileged class.26 It might even be expected that some authors would deny that selfascriptions of beliefs (conceived of as dispositions) exhibit first person authority in any strong sense27. It is, thus, doubtful, whether (BT) really holds. Given the above remarks, it seems safer to assume at most: (B)

23

24

25

26 27

First person authority (in the sense of (A), (ET) (CT) and, perhaps, (WT)) applies to first person self-ascriptions of beliefs.

Ayer (1963) p. 73. There is another element in Ayer's account that I have deliberatly excluded. He claims that the special authority of such self-ascriptions belongs to the logic of such statements (cf. p. 73). Whatever that means, I think it is safer not to exclude what might be called "factual incorrigibility": the claim that we are in fact incorrigible though there might be conceivable circumstances in which we might lose special authority. Ayer denies that this could happen even if others were supplied with an ability to become co-conscious with a self-ascriber. Cf. Fn. 14. I am not sure whether Davidson would reject Ayer's weak incorrigibility-claim as applied to present occurrent conscious mental states. He says that the incorrigibility-condition "is perhaps reasonable" when we are concerned with "sensations such as pain" (FPA 108). In his earlier treatment in The Problem of Knowledge, Ayer himself did not accept even weak incorrigibility (Ayer 1956 p. 65 f.). In Ayer (1963), on the other hand, where he endorses weak incorrigibility, he sympathizes with the slightly stronger claim that under the conditions mentioned the self-ascription is not only weakly incorrigible but infallible. Mackie weakens even weak incorrigibility, when he says of first person self-ascriptions that they are not "incorrigible, but merely that we can never be forced to correct them. This is a guarantee not against error, but against the conclusive detection of error" (Mackie (1963) p. 26). Armstrong holds the even more radical view that there are conceivable cases where, although a person sincerely claims to feel severe pain, we might be justified in denying his claim (cf. Armstrong (1968) p. 108, Armstrong (1984) p. 135 f. and Churchland (1984) p. 76 ff.) This might be the reason why Davidson's characterization of first person authority is in some respects weaker than Ayer's weak incorrigibility. Cf. Fn. 24. Cf. Rorty (1970) p. 410 ff.

222

Bernhard Thöle

I shall say something about the plausibility of the claim (B) (and the stronger claim (BT)), when I consider the attempts to explain first person authority. In (Β), I have not only dropped (TT) and modified (AT), I have also dropped (RT). The reason for this is that it is not obvious that even the weaker version: (R)

(A) holds not only for first person self-ascriptions in general but also for every class of first person self-ascriptions of any particular person

holds. For suppose we have good reason to deny a particular self-ascription of a person. There must, in this case, be some reason why the person was mistaken in this particular case. But why should it be thought to be impossible that the very same sort of reason — whatever it was — is operative in all, or most, of this person's self-ascriptions. Without further arguments in favor of (R) (or (RT)) I see no reason why we should exclude the possibility that some people are extremely bad self-ascribers. In what follows, I shall assume that an explanation of first person authority should at least explain (A) as it applies to mental states of the priviledged class. 28 If stronger results are forthcoming, these will be welcome, but in the absence of a convincing stronger account we should not be dissatisfied with a weaker one. I shall begin with a defence of the traditional explanation against Davidson's objections. Since the traditional explanation is primarily concerned with mental states belonging to the privileged class, I shall, in that section, concentrate on these. But some remarks will be made about how beliefs fare within the traditional account. I shall then turn to Davidson's explanation of first person authority concerning beliefs. Since I will attempt to show that his explanation fails, I conclude that we can rest content with the weaker version of first person authority and its traditional explanation.

II. The Traditional Account If we conceive of first person authority in the way I have indicated, an explanation suggests itself. I shall call it the traditional account. If there is an epistemic asymmetry between knowledge expressed in first and in third person ascriptions, it would seem reasonable to trace this asymmetry back to the different ways in which the respective knowledge is established. If there were no difference between my way and your way of finding out 28

Only those who accept Ryle's absurd rejection of first person authority would dismiss (A) altogether.

The Explanation of First Person Authority

223

about my mind, why should there be a resultant difference in the epistemic status of our knowledge claims? I shall call the way which leads to first person ascriptions the first person way. Correspondingly, I will speak of the third person way. Central to the traditional account is the claim that the asymmetry rests on the fact that the first person way is direct whereas the third person way is indirect (or perhaps we should say: the first person way is more direct than the third person way). This is, of course, at best a preliminary description of the traditional view. More needs to be said about the relevant notion of directness, especially, since the traditional view is often associated with a certain picture of the mind that is now commonly viewed with suspicion. We might call this picture the theatre-view of the mind. The mind is pictured as a "kind of theatre, where several perceptions successively make their appearance; pass, re-pass, glide away, and mingle in an infinite variety of postures and situations." 29 It includes a number of claims that philosophers have found problematic, e. g., that first person ascriptions are the result of directing the mind's eye towards our mental states; that introspecting our minds brings our self into direct contact with our mental states, a contact nobody else can have with them; and that the objects of such inner awareness are "so directly before the mind that it is impossible to misidentify them" (Davidson (1989b) p. 3). I shall try to arrive at a clearer conception of 'directness' that would enable us to explain the special authority of first person self-ascriptions in the course of answering Davidson's objections against the traditional account. This will also help us to decide to what extent an advocate of the traditional view must be committed to the theatre view of the mind. Davidson's first objection Davidson rightly points out that invoking different ways of arriving at knowledge does not by itself explain why the first person way should be more reliable than the third person way. Against the suggestion that the directness of self-attributions consists in the fact that they are not based on evidence, he rightly insists that "it is hard to see, failing an explanation, why knowledge that is not based on evidence should be more certain than knowledge based on evidence" 30 ; on the contrary, claims that are based on nothing share the status of mere guesses. But those who claim that direct knowledge is not based on 29

30

Hume: A Treatise of Human Nature. I. IV. VI. p. 253. As is well known, Hume does not subscribe to the theatre view in its familiar version, since he rejects the idea of a distinct self watching its perceptions. Davidson (1989a) p. 194.

224

Bernhard Thöle

evidence do not mean that it has no basis at all 31 ; they typically mean that direct knowledge does not depend on an awareness of something else. So if knowledge of your mental state depends on knowledge about your behaviour, it does not count as direct in the intended sense. This still leaves us with at most a rough and negative idea, and it does nothing to distingish direct knowledge from guessing. So we need some positive conception of direct knowledge. A natural reaction would be to claim that the positive element is provided by saying that direct knowledge depends on awareness of the states ascribed. The special authority of knowledge claims that are based on awareness is then claimed to derive from the fact that "because there is no difference here between what is claimed and the evidence on which the claim is based, there is no room for error". 32 But this response does not suffice. Let us grant that it makes sense to say that self-ascriptions depend (evidentially) on a direct awareness of the ascribed state; and let us grant further that such ascriptions are not liable to the sort of mistakes that might typically occur when we transcend the available evidence. Still, the fact that certain sorts of mistakes are excluded when we rely on direct awareness does not show that there might not be other sources of error. As Mackie puts it, "though a gap between evidence and what it supports is one thing that makes error possible, it need not be the only one. Since there could be error in direct as opposed to indirect cognition, we cannot assume that our immediate awareness is infallible for that reason alone." 33 Mackie's objection is directed against the suggested explanation as applied to the infallibility thesis, but it can easily be applied to the asymmetrythesis as well. For suppose there are two ways of detecting certain sorts of things (say electrons); and suppose that only one sort of error is possible in the first way, whereas there are several different sources of possible error in the second way. This does not show that the second way is less reliable than the first, since there might be a very high probability of the kind of error possible in the first way, whereas the probability of error

31

32 33

Davidson cites a remark by Wittgenstein: "What is the criterion for the redness of an image? [...] For myself, when it is my image: nothing." (Wittgenstein (1953) § 377. Now Wittgenstein, in this passage, does not explicitly say that self-attributions are based on nothing; he only denies the existence of a criterion in this case. Nevertheless, he claims that first person self-ascriptions are (like) expressions and rejects the claim that they are based on an inner awareness. I think this is an absurd position, but here I need only the fairly safe claim that it is an extreme position that is certainly not representative. For a criticism of Wittgenstein's 'expression-thesis' cf. Ayer (1963) p. 59 ff. Mackie (1963) p. 17. Mackie — as will be seen — does not accept this argument. Mackie (1963) p. 17 f. Cf. Davidson (1987) p. 442.

The Explanation of First Person Authority

225

possible in the second way may be much smaller even, if we take all possible sorts of error together. So, even if we rely on direct awareness we might nevertheless press the advocate of the traditional account to explain why direct awareness is thought to guarantee truth, weak incorrigibility, or at least epistemic superiority over indirect ways. Davidson seems to make a similar point when he claims that the supposed explanation of self-knowledge on the basis of introspection "only leads to the question why we should see any better when we inspect our own minds than when we inspect the minds of others" (FPA 103). This is his first objection to the traditional explanation. To answer this objection, I shall first claim that since awareness is a source of knowledge, claims based on awareness, though they are not infallible, have prima facie credibility. Their epistemic status is, thus, distinct from guessing. Secondly, since claims that have initial credibility could be discredited only if there are reasonable grounds for doubt, we can complete our explanation of first person authority by showing that reasonable grounds for doubting a sincere first person self-ascription are not easily available. More specifically, I shall argue that (a) it is very unlikely that there are reasonable grounds for doubt, (b) that it is impossible that third person evidence could generally invalidate first person ascriptions and (c) that third person evidence is in general too weak to seriously call into doubt first person ascriptions. As to the first part, I think it is evident that we are (or can easily become) directly aware 34 of our present conscious occurrent mental states and events. Although it might be a difficult task to give a satisfactory account of such inner awareness 35 , we cannot seriously doubt its existence. Similarly we cannot seriously doubt that awareness is a source of knowledge: if there were no awareness, we would have no (empirical) knowledge at all. This in itself does not show that either awareness in general or inner awareness in particular provides us with epistemologically privileged 34

35

We are directly aware of X if we are aware of X and there is no Υ φ X such that we are aware of X in virtue of being aware of Y. For a detailed elaboration of this cf. Jackson (1977) p. 15 ff. Here the distinction between sensations on the one hand, and states and events with content on the other hand, is relevant. To have a conscious sensation is to be aware of it. But to have, e.g., a conscious thought does not consist in being aware of that thought; it consists in being aware of the content of the thought. Thus, whereas it seems plausible to claim that the occurrence of a conscious sensation constitutes its awareness, this is not true of states with content. One might still claim that even in the latter case the awareness of the content is — as Brentano claimed — necessarily interwoven with an awareness of the mental act itself. Although I sympathize with Brentano's view, I do not presuppose it here. All we need is the weaker claim that the occurrence of a mental state does in fact put the possessor of that state in a position to become aware of this state.

226

Bernhard Thöle

knowledge of the objects of awareness. At this stage of the argument, I am merely maintaining that awareness is a source of knowledge and that the epistemic status of claims based on awareness is thus superior to guesses and the like. Let us, then, turn to the second part of our explanation. Given that there is direct awareness of our mental states which provides first person self-ascription with prima facie credibility, we have to explain why such self-ascriptions have special authority. A first step is suggested by Mackie's claim that self-ascriptions are "practically incorrigible" 36 . Although he admits that there is no guarantee against error, there is a guarantee against "the conclusive detection of error". 37 The reason for this is that every conceivable evidence for doubting a self-ascription could always be accounted for in an alternative way that leaves the self-ascription untouched. A self-ascriber can, therefore, always insist on his claim. 38 It might be argued that this consideration could only explain why we treat first person self-ascriptions as authoritative, not that it is rational to do so. For suppose there is strong evidence that a self-ascription is wrong: we might grant practical incorrigibility, i. e. we might grant that there is no, and cannot be, conclusive evidence against the self-ascription. It remains to be explained, however, why we should prefer the self-ascription to the contravening evidence. A possible response to this would be to claim that the basic reason why we do this is that we can do it. One might perhaps add to this the observation that people are so constituted that they obstinately insist on their self-ascriptions and that cases in which there is very strong contravening evidence are rare. But let us see whether we cannot achieve something more. In The Problem of Knowledge, Ayer gives an account that, if acceptable, would provide us with an explanation of a stronger account of first person authority. One possible weakness of Mackie's account was that it is limited to an explanation of the fact that we treat first person self-ascriptions as authoritative, while doing little to explain why we are justified in treating them thus. Now, Ayer suggests an explanation that is relevant also to the de jure question: though self-ascriptions are not incorrigible, "statements which do no more than describe the content of a momentary, private 36 37

38

Cf. Mackie (1963) p. 26 f. Not only others but even the subject itself cannot conclusively detect errors in first person self-ascriptions of present occurrent mental states. One might ask whether first person self-ascriptions are practically incorrigible in principle or only given our actual cognitive capacities. This depends on what we would say if it became possible to introspect other minds. I shall leave that question open. Cf. the detained discussion in Broad (1925) p. 300 ff.

The Explanation of First Person Authority

227

experience achieve the greatest security because they run the smallest risk." 39 The point is not merely that others cannot conclusively show that a self-ascription is mistaken; the point is that whatever evidence they happen to adduce, there is a greater chance that they (and not the self-ascriber) made some mistake. And the reason for this is that it is hard to give a convincing and credible account of why and how the subject might have made the mistake. The only conceivable (factual) mistake the self-ascriber might make is that he misidentifies his present conscious experience. And although this might happen, it is unlikely if the speaker is not in a confused state of mind. It is unlikely since the task is (in general) fairly simple. 40 Not only is it highly unlikely that we fall into error when we selfascribe mental states on the basis of inner awareness, it is also unlikely that strong contravening third person evidence against a sincere first person self-ascription is available. First, it is always possible — as Davidson acknowledges 41 — to question either the truth or the relevance of the third person evidence. If a self-ascriber might misidentify his present conscious experience, an interpreter might equally misidentify the evidence on which he bases his third person ascription. Furthermore, he might misinterpret the evidence. Second, in many cases third person ascriptions depend on utterances of first person self-ascriptions. In some cases — notably in belief-attributions — first person self-ascriptions constitute the sole evidence available to an interpreter. And in such cases, third person evidence can hardly conflict with first person ascriptions. Third, pure third person evidence 42 is in most cases far too unspecific to provide strong grounds for detailed third person ascriptions. Taken together, this shows that we are, in general, quite unlikely to have strong third person evidence that contradicts a sincere first person self-ascription. If we combine this with our previous result according to which it is unlikely that an attentive self-ascriber makes mistakes, we have an explanation of the asymmetry-theses also in its de jure version. 43 39 40

41 42

43

Ayer (1956) p. 66. Of course, there are differences of degree. Self-ascriptions of more complex mental states are more liable to mistakes than ascriptions of less complex states. This means that we have to qualify (TT). But even in those more complex cases, the asymmetry thesis still holds good. For even in these cases, the only generally reliable method to detect and correct a possible mistake is through more attentive inspection. Davidson (1987) p. 442. I shall say that third person evidence is pure, if it does not include utterances of first person ascriptions. Still, Davidson might object that it does not give a full account of first person authority. For it seems to be compatible with the falsity of (RT) since some people might often make mistakes even in simple tasks. But I have already argued that it is uncertain whether (RT) really holds.

228

Bernhard Thöle

In his British Academy Lecture on Privacy, Ayer takes a further step44. He claims that the reliability of third person ascriptions depends on the (general) reliablity of first person self-ascriptions. Ayer illustrates this point by comparing first person authority with the authority of an eyewitness. I shall explain and defend Ayer's suggestion by considering the criticism Davidson levels against it. Davidson finds this analogy unsatisfactory for two reasons. First, it fails to tell us why a person is like an eyewitness with respect to his own mental states and events while others are not. And second, it does not suggest an accurate description of what first person authority is like. For first person attributions are not based on better evidence, but often at least on no evidence at all. The authority of the eyewitness is at best based on inductive probabilities easily overridden in particular cases: An eyewitness is discredited and his evidence discounted if he is a notoriously unreliable observer, prejudiced, or myopic. But a person never loses his special claim to be right about his own attitudes, even when his claim is challenged or overturned (FPA 104).

I think that both objections miss the target. Ayer does explain why the eyewitness and the first person case are analogous and he does not claim that the analogy is complete. I begin with the first objection. According to Ayer, the reason why we should be bound to give the eyewitnesses preference is that it is only through [...] agreement with their evidence that [second-hand reports 45 ] could come to merit any credence at all. If they triumph over one set of eye-witnesses it is only because they are supported by another. [...This] provides us with a satisfactory model for the logic of the statements that a person may make about his present thoughts and feelings. [...] If he is not infallible, others may be right where he is wrong. Even so their testimony is subordinate to his in the same way and for the same reason as [second-hand reports are] subordinate to that of the eye-witness. If his reports are corrigible it must be that he himself is ready to correct them [...] in the same breath. 46

Let me try to bring out the general thought behind this reasoning. There is a contrast between the eyewitness way and those ways on which second-hand reports rely. And there is a contrast between the first person way and the third person way. The eyewitness way and the first person way can be said to be direct compared with their contrasting indirect ways in the sense that the reliability of the indirect ways depends upon agreement with claims resulting from the direct way. 44 45

46

There are indications of this further step already in his earlier treatment in Ayer (1956) esp. p. 64 f. Ayer develops the analogy by considering the contrast between reports by eyewitnesses and those of a group of clairvoyants. This contrast is motivated by the aim to account for the special authority of first person self-ascriptions even in case others would become telepathically aware of the persons mental states. I assume — with Davidson — that Ayer would claim that the analogy holds not just in this special case. Ayer (1963) p. 73.

The Explanation of First Person Authority

229

Let us say that a way W is justificationally indirect (or j-indirect) relative to another way W*, if its reliability depends on its general agreement with the outcome of W*. W* is then called justificationally direct (or j-direct) relative to W. Claims based on j-indirect ways cannot, therefore, in general invalidate claims based on corresponding j-direct ways without seriously impairing their own reliability. Thus j-direct ways are epistemically privileged. The suggestion then is, that the first person way is j-direct relative to the third person way. All the evidence on which the third person way relies consists of indirect signs (outer effects) of those inner states to which no one but its possessor has direct access 47 . In establishing in a third person way (for example on the basis of behavioral evidence) that some mental state obtains we must, thus, rely on well-established correlations between the occurrence of the mental state (originally detectable in the first person way only) and its behavioral manifestation. But does this help to explain the special character of first person authority? This brings us back to Davidson's second objection. Davidson objects that the case of the eyewitness is different from that of the selfascriber since (a) the authority of an eyewitness is based on inductive probabilities whereas first person attributions are not based on better evidence (they are normally not based on evidence at all) and (b) an eyewitness might be discredited whereas a person never loses his special authority. Objection (a) rests on a confusion: 1. The assessment of the authority of a particular eyewitness might rest on evidence, but this does not mean that hisjudgements rest on evidence. Thus, as far as the particular judgements are concerned, there is so far no disanalogy between the eyewitness and the self-ascriber. 48 And both cases are analogous in that both the selfascriber and the eyewitness rely on awareness 49 which provides them with 47 48

49

In the sense suggested in Fn. 34. Furthermore, it is not evident that the authority of a particular self-ascriber might not equally rest on evidence. From the fact that first person ascriptions are not based on evidence it certainly does not follow that the authority of a person as a reliable selfattributer does not rest on evidence. And I see no reason why it should be impossible that a particular self-ascriber loses first person authority. This might, e.g., happen if a person after serious damage to his brain shows great uncertainty and makes conflicting self-ascriptions. It should be noted that Ayer — in an earlier passage of his article — does reject the claim that direct knowledge of our occurrent conscious mental states is analogous to 'outer' perception of physical objects on the ground that "our thoughts and feelings do not normally 'pose' for us in the way physical objects do" (Ayer (1963) p. 62). This is a rather weak argument, since it seems to rely too heavily on taking visual perception as the paradigm of 'outer' perception. I cannot see that in e.g. tactually feeling the hardness of the table in front of me the table 'poses' for me whereas a severe pain does not. Of

230

Bernhard Thöle

j-direct knowledge. 2. If an eyewitness is discredited 50 , it must ultimately be on the basis of disagreement with other eyewitnesses. In other words: discounting particular j-direct reports or particular persons as j-direct reporters depends ultimately on checks against other j-direct reports (together with some account of the causes of error). Now it is easy to see why first person self-ascriptions cannot be easily discounted in this way. there are no corresponding j-direct reports to check them. The self-ascriber is — so to speak — the only eyewitness of his own mental states.51 There is thus a difference between the eyewitnesses knowledge and selfknowledge: In the case of mental self-ascriptions there is one possible "observer" only, namely myself. But it is precisely this difference that helps to explain why first person authority surpasses that of an eyewitness. And there is a further difference that accounts for the special character of first person authority: we do have a rough theory about the possible sources of error in perceptual judgements. Thus, we can reasonably conjecture under which conditions a person is likely to make erroneous perceptual mistakes. But we do not have a similarly respectable theory about the sources of introspective error. Taken together, these facts explain first person authority. First person self-ascriptions enjoy prima facie credibility because they are based on inner awareness. Consequently, they can be called into doubt only if there is strong positive reason for thinking that they are erroneous. There can be strong positive reason for doubt only if (a) we are equipped with a convincing account of the sources of error or (b) if we have strong evidence which contradicts the first person self-ascription. But if the self-ascriber is sincere, attentive, not confused or drunk, and if he makes no linguistic mistakes, it is highly unlikely that his selfascription is erroneous (at least in those cases where the relevant state is

50

51

course there are differences of degree, but these are present in inner as well as in outer awareness. Furthermore, his denial of inner awareness seems to be inconsistent with Ayer's later comparison with the case of the eyewitness. It is sometimes claimed (cf. Shoemaker (1963) Chap. II; Alston (1976) p. 266f) that although we might be aware of the objects or contents of our mental states in a sense analogous to the awareness of physical objects, the same is not true of the mental states themselves: the acts of sensing, imagining and thinking. In the sense in which this is true it is irrelevant. It is true in the sense that being aware of a mental act is an awareness of a relational fact, whereas being aware of a tree or a pain is not. But this is irrelevant since there is also awarenes of relational facts in outer perception; and an advocate of inner awarenes of mental acts should claim that such awareness is analogous to outer awareness of relational facts. Cf. Broad (1925) p. 308 f. It should be noted that even in this case the presumption that reports of eyewitnesses are in general more reliable than inferences based on second-hand evidence is not affected. Given our present cognitive capacities. Cf. Fn. 14.

The Explanation of First Person Authority

231

not too complex 52 ). The task of identifying our conscious mental states is, in general, fairly simple if we do no more than describe the contents of our present occurrent conscious mental states. 53 Even if there were other sources of error, we do not (at present) have reliable means for their detection. Furthermore, we have seen that it is unlikely that there is strong contravening evidence either of a first person or a third person sort. Since first person authority is restricted to present conscious states, we can discount contravening first person memory-statements. And pure third person evidence is often too unspecific to constitute strong evidence against firmly held first person ascriptions. Furthermore, since the first person way is j -direct relative to the third person way, the relevance of third person evidence relies on the general reliability of first person ascriptions. We can, thus, account for the asymmetry-thesis and (perhaps) for weak incorrigibility. But the proposed account provides no guarantee that first person self-ascriptions are generally true. This means that the traditional account does little to explain first person authority in the strong sense (i. e. including (TT)) But as already indicated — it is not obvious that (TT) is true. Davidson's second objection We can now turn to Davidson's second objection. This objection is not only directed against the traditional explanation. He claims that "any attempt to explain the asymmetry [...] by reference to a special way of knowing or a special kind of knowledge must lead to a skeptical result" (FPA 104). By this I take him to mean that there is no guarantee that the different ways (the first person way and the third person way) lead to the same thing. If the third person way is radically different from the first person way, what reason is there that first- and third person ascriptions concern the same states?54 The answer is that there is no absolutely secure guarantee against radical skepticism anyway. But this does not mean that we are not justified at all in our belief that both ways lead to the same thing. This justification relies on the possibility, which each person can realize in her own case, of establishing correlations between directly inspected mental states on the 52

53

54

If the state is very complex we should expect that a sincere self-ascriber would himself show some uncertainty in his claim. Of course, we might transcend the bounds of what is given to consciousness. But such self-ascriptions do not belong to the privileged class and, thus, first person authority does apply to them only to a limited degree. Cf. FPA 106 and Davidson (1987) p. 442.

232

Bernhard T h ö l e

one hand and what is taken as their outer manifestations on the other hand. 55 There is no better way of validating non-direct attributions. And since the third person way is j-indirect relative to the first person way, it must lead to the very same states. Davidson's third objection Davidson's third objection is of a more general nature, and I cannot deal with it adequately here. It expresses a familiar uneasiness with a certain picture of the mind as "a theatre in which the conscious self watches a passing show [...]. The show consists of appearances, sense data, qualia, what is given in experience." (KOOM 453) We can agree with Davidson that these metaphors can mislead in several respects. But I do not think that the traditional account needs make use of the more questionable elements of the theatre view of the mind. Nevertheless, the traditional explanation does of course presuppose that there is inner awareness of our mental states and that such inner awareness confers prima facie justification on those first person self-ascriptions that depend on it. We might expect Davidson to reject this account on three grounds: (a) Invoking an inner awareness seems to buy too much of the supposedly forbidding theatre view of the mind, (b) Specifically, if awareness is thought to explain first person authority, it seems that we have to postulate that the objects of such an awareness must — as Hume put it — "necessarily appear in every particular what they are, and be what they appear" 56 . But according to Davidson, there are not, and there cannot be, such objects, (c) The attractiveness of the proposed account seems to exploit an ambiguity of the word "depends". 57 There is causal and there is evidential dependence. Now, Davidson would grant that self-ascriptions normally depend on the existence of the states ascribed. But the sense of 'dependence' in which they so depend is causal.58 The traditional account, however, obviously requires evidential dependence59. And according to Davidson, this is unacceptable since there is no logical relation — between a given 55

56 57

58

59

These sketchy remarks should, of course, n o t be taken as a complete account of the present problem. F o r a convincing defence o f the 'Argument f r o m A n a l o g y ' against familiar objections cf. Hyslop/Jackson (1972). Hume: A Treatise of Human Nature. I.IV.II. p. 190. C o m p a r e his critical remarks about the relation between sensations and beliefs in D a v i d s o n (1983). Since he rejects introspection, I am not sure w h e t h e r he w o u l d grant that self-ascriptions might depend on awareness of those states, even in the sense of causal dependence. A relation is evidential, if one of the relata contributes to the justification of the other.

The Explanation of First Person Authority

233

mental state and the belief that is expressed by the self-ascription of that state — that would j u s t i f y the self-ascription60. I begin with the last point. I do not see why the dependence could not, in that case, be both causal and evidential. I find it quite natural — even compelling — to say, e.g., that my (consciously) feeling pain does something to j u s t i f y my belief that I feel pain. Of course there is no logical relation between my feeling pain and my belief that I feel pain. But this only shows that not all relations that confer justification are logical (in a strict sense). Only reliance on the dogma that only beliefs can justify beliefs would prevent us from saying this.61 Let us now turn to (b). First, I do not think that we have to postulate objects that are what they seem and seem what they are. We did allow for mistakes even in claims based on direct awareness. And we did not assume that first person self-ascriptions give exhaustive descriptions of the mental 60

He certainly holds this view concerning sensations — on the ground that "the relation between a sensation and a belief cannot be logical, since sensations are not beliefs or other propositional attitudes" (Davidson (1983) p. 428). But what, if the state is (an awareness of) a belief? I am not sure what Davidson would say, but I suspect that he would point out that, although the belief is the right kind of thing to enter into logical relations (whereas sensations by their very nature cannot), a belief does not (normally) stand in the right logical relation to justify the respective second-order belief since a belief that ρ does not logically imply the second-order belief. The issue, however, is complicated, since one might argue that although "S believes that p" does not imply "S believes that he believes that p", there is a close connection between the utterance "I believe that p" and the utterance "/ believe that I believe that p" (or between my consciously believing that ρ and my belief that I believe that p), because it would be paradoxical to say "p, but I do not believe that p". An even stronger case could be made, if we take the canonical form of the second-order belief to be "I believe this", where 'this' is supposed to refer to a given belief. One might think that self-ascribing a belief by uttering "I believe this" logically cannot be wrong, since the attribution must be true if "this" refers but would be empty (and, consequently, could not count as a self-ascription) if "this" would not refer. But this is not true, since "this" might refer, although it does not refer to a belief. For suppose I consciously entertain a proposition which in fact I disbelieve (although in this situation I just think about it). In this case, "I believe this" would refer but be wrong. Davidson makes a related objection against a similar proposal made by Bürge (Davidson (1988) p. 664). I think, however, that his objection misses Burge's account, since Burge's proposal is even more restricted than the one just considered. According to Bürge, the truth of what he calls basic self-ascriptions is guaranteed by their self-reflexive character. They are of the form "I hereby judge that ρ". I agree with Bürge — against Davidson — that a self-ascription of this kind cannot be wrong. But I agree with Boghossian that to claim that this could serve as the basis of a general account of the special character of our self-knowledge would be to grossly exaggerate the significance of a very special case (cf. Boghossian (1989) esp. p. 20 ff.). Bürge is cautiously optimistic when he says that "basic self-knowledge is at most an illuminating paradigm for understanding a significant range of phenomena that count as self-knowledge" (Bürge (1988) p. 663 Fn. 11). The caution is appropriate, but I see no reason for the optimism.

61

For a critical discussion of the dogma cf. Sosa (1991) esp. Chap. VII.

234

Bernhard Thöle

state in question. T h e a r g u m e n t that direct k n o w l e d g e - c l a i m s w h i c h are restricted to w h a t w e are a w a r e of, are less liable t o be mistaken than o t h e r k n o w l e d g e - c l a i m s , does n o t rest o n d u b i o u s a s s u m p t i o n s about the nature o f the objects o f direct awareness. T h e central reasons w e r e that (1) there are o n l y f e w possible sources o f mistakes in direct k n o w l e d g e and (2) that it is h i g h l y i m p r o b a b l e that a p e r s o n makes such mistakes (at least u n d e r f a v o u r a b l e conditions). W h a t a b o u t (a)? I h a v e tried t o s h o w that there is n o t h i n g problematic a b o u t o u r reliance o n t h e existence o f i n n e r awareness. T h e traditional account makes n o use o f f u r t h e r — p o s s i b l y unacceptable — elements o f the theatre v i e w . W e did neither assume that inner awareness is a special, m y s t e r i o u s sort o f awareness, n o r did w e postulate that the objects o f inner awareness h a v e a special, m y s t e r i o u s nature. W e m i g h t justifiably be suspicious o f the theatre v i e w if it w e r e t o i n c o r p o r a t e s such assumptions. But this s h o u l d n o t lead us t o d e n y t h e o b v i o u s : Being in a conscious state p r o v i d e s the subject o f that v e r y state w i t h a direct awareness o f this state — an awareness n o b o d y else has o f it. T h u s , I think, D a v i d s o n ' s objections against the traditional account can be a n s w e r e d . 6 2 T h e tradition explains w h y w e are justified in treating first 62

I add that there are two further points which would require more extensive discussion: (1) Davidson finds "another conspicuous, though perhaps less appreciated, difficulty" in the theatre view. It concerns the location of the self in this picture. On the one hand the self seems to include theatre, stage, actors and audience; on the other hand the self seems to be the subject watching the show and thus seems to be distinct from what it perceives (cf. Davidson 1987 p. 454) I think this is a pseudo-problem since the self might certainly perceive some of its states. Similarly, in order to perceive my nose I need not cut it off. Needless to say, these remarks are not intended to solve the huge problems surrounding self-knowledge. Cf. the careful discussion in Broad (1925) Chap. VI and XIII. (2) There is a special problem about the objects of thought. If we assume that in having a conscious thought we are aware of a certain content, we seem to be forced to conceive of thinking as the grasping of propositions. But there are at least two objections against such a view: (a) there seems to be a difficulty in explaining how the mind could relate to an abstract entity, and (b) it is claimed that the identity of a proposition is determined by relations (to other propositions and to objects outside the mind); and this seems to threaten first person authority about thought-content (cf. Davidson (1989b) p. 3f.) As to (a), I am prepared to accept the claim that conscious thought involves the grasping of an abstract entity, if this is a consequence of — what I take to be an obvious fact — that we are aware of the contents of our conscious thought. We should not be surprised that such a grasping is difficult to explain, if we recall that it took a long time for evolution to produce thinking animals. As to (b), I restrict myself to two sketchy remarks: (i) first person authority in the full sense applies only to narrow content (here I take narrow content to be very narrow indeed: its identity conditions are determinded by what the subject of the state is (or can easily become) aware of).; (ii) if I use the same proposition that was grasped in a given thought in order to self-ascribe that thought, I thereby ascribe a thought that has the same broad content as the given thought. (I shall say a little more about (i) and (ii) when I discuss Davidson's comments on whether externalism threatens first person authority, cf. Fn 82).

The Explanation of First Person Authority

235

person self-ascriptions as in general more reliable than their corresponding third person ascriptions. The main reasons are the following: (a) inner awareness is a generally reliable source of knowledge, (b) strong contravening third person evidence is rare and (c) the reliability of third person evidence depends on the general reliability of first person self-ascriptions. A note on belief-ascriptions It remains to say something about how the traditional account is suited to //«/"-ascriptions. In Part I, I already expressed some doubts whether first person belief-ascriptions enjoy first person authority to a high degree. But of course we do normally know what we believe and we think that we know this better than others. Again, we think we know about some of our (conscious) beliefs in a way that is similarly direct as our knowledge of our present conscious feelings and thoughts. Furthermore, thoughts resemble beliefs in that they too have propositional content. All of this suggests that insofar as first person belief-ascriptions are privileged, they are so for a reason that is related to the source of first person authority of the privileged class. But there are nevertheless important differences. Most of our beliefs are most of the time unconscious, and beliefs are dispositions that are essentially connected (together with other mental states (notably beliefs and desires)) with behaviour. This suggests that self-ascriptions of beliefs that are exclusively based on the first person way are much more susceptible to correction by third person evidence than self-ascriptions of the privileged class. If the traditional account can be applied to belief-ascriptions, there must be some direct way of finding out about our beliefs. Perhaps it is wrong to think of just one single first person way in this case, and it would certainly be wrong to claim that we detect our beliefs in exactly the same way as we detect the occurrence of our current conscious feelings. If I inquire whether a few seconds ago I had the belief that ρ, I do not look in the darker corners of my past total mental state to determine whether they contained some faint proposition combined with a belief-feeling. Nevertheless, the process of detecting my past or present beliefs involves direct awareness. When I ask myself whether or not I believe that p, this question initiates a certain process which results in a conscious judgement about p. The process leading to the judgement might, but need not be, itself (partially) conscious. But if it should give us knowledge about my belief-state, it must result in some conscious assent, dissent or agnostic attitude towards p. My conscious assent to p, then, provides me with knowledge about my belief that p. But this knowledge is a product of both (a) my direct introspective knowledge of my assenting to ρ (as a result

236

Bernhard Thöle

of my asking myself whether I believe p) and (b) my tacit assumption that most of my beliefs are such that they dispose me to (immediately or after reflection) assent to the corresponding proposition. If there is first person authority about beliefs, it must depend on both the immediate awareness of our conscious judgements and the epistemic status of the tacit assumption just mentioned. If there is something special about />e//f/-ascriptions (as compared to first person authority about occurrent conscious states) this special character must derive from the status of this assumption. I am inclined to think that the assumption reflects a contingent fact: it is just a contingent fact that beliefs are normally functionally connected with a disposition to assent to the proposition believed 63 . There might well be exceptions. If we allow for unconscious beliefs in a deeper sense 64 , not all beliefs will fall under this assumption — even if we consider only creatures that are (in principle) capable of self-consciousness. If we conceive of beliefs as those states that (in combination with desires) rationalize behaviour, there can only be a contingent connection between beliefs and their corresponding conscious judgements. According to the traditional account, we have first person authority about our conscious judgements. Whether this leads to first person authority about our beliefs depends on the strength of the assumed causal connection between beliefs and their corresponding conscious judgements. Since in creatures like ourselves this connection does normally hold good, first person self-ascriptions of beliefs that are based on our inner awareness of conscious judgements are normally reliable. Thus something like (TT) holds for self-ascriptions of beliefs. But it is not clear that the given explanation also accounts for the asjmmetrjthesis about belief-ascriptions. For others might have strong behavioural evidence that I believe p, although I am not disposed to assent to p. But this does not necessarily indicate a defect of the tradional account because it is doubtful whether the asymmetry thesis is true of belief-ascriptions — given that we conceive of beliefs as states that, in combination with other mental states, rationalize behaviour. It has, however been argued that there must be a closer, non-contingent, connection between beliefs and a disposition to self-ascribe these beliefs. I shall briefly consider two arguments to this effect. 63

This is not to deny that these functional connections are useful. As Locke observed: "To think often, and never to retain it so much as one moment, is a very useless sort of thinking" (Essay II.1.15). I am here speaking only about beings with roughly the same cognitive capacities as those of normal human beings. I am inclined to think that there are lower animals that have (genuine) beliefs but do not have the capacity to assent to propositions; on the other hand, there might be creatures without language that nevertheless have such a capacity.

64

I.e., not just beliefs that are at a certain time not consciously entertained, but beliefs that can become conscious only after years of psychoanalytic therapy, or perhaps never.

The Explanation of First Person Authority

237

1. Davidson has argued that belief-attribution requires linguistic beliefexpression. The non-linguistic evidence available to an interpreter is said to be much too unspecific to allow for attributing specific belief-contents to the subject. Like many others, I do not find this radical view plausible, since it precludes belief-attribution to animals who lack language but nonetheless display highly complex behaviour. 65 2. Shoemaker has recently endorsed a more moderate position 66 . He claims that beliefs of creatures that have approximately the same cognitive capacities as normal human beings must normally be connected with knowledge of these beliefs. The reason for this is, roughly, that beliefs could not play the role they characteristically play in human beings if they did not enter into theoretical and practical deliberations. But this implies — according to Shoemaker — that the creature be capable of revising and adjusting its prior beliefs in the light of new information, which, in turn, presupposes that it knows what its beliefs are. Thus, beliefs that are not essentially connected with (possible) self-knowledge could not play the normal role of beliefs — and thus would not be beliefs. I cannot here do justice to the complexity of Shoemaker's arguments. I shall only make two critical remarks: (a) I am not convinced that Shoemaker has refuted the objection, that a creature might be "simply 'hard-wired' to make, in the light of new experience, the adjustments in its belief-desire system that are required to preserve rationality, without there being any second-order beliefs and desires that rationalize these adjustments'"57, (b) Even if we grant that, in rational creatures with the conceptual capacities of human beings, the relevant second-order beliefs (and desires) are necessarily connected with (most) first-order-beliefs, it does not follow that these second-order beliefs must be conscious states. Only on that further assumption, would they contribute to an explanation of first person authority. The upshot of our discussion of belief-ascriptions then is this: First person self-ascriptions of beliefs depend on awareness of our conscious judgements. We have first person authority about our conscious judgements and the explanation of this is essentially along the lines of the 65

One of Davidson's arguments seems to be that we cannot attribute a belief to a dog that, e.g., the cat is on the tree, since we have no means of deciding which of the following propositions he believes: (a) that the cat is on the oak tree, (b) that the cat is on the oldest tree in sight. If this were true, it would seem to preclude belief attribution even to creatures who happen to have a less discriminating language than we do. What belief should an eskimo, who is supposed to have a much more fine grained vocabulary for describing 'white' objects, ascribe to me if I say that the bear is white?

66

Cf. Shoemaker (1988) and (1990). Shoemaker himself raises this objection in (1988) p. 193. But as far as I can see he does not explicitly answer it.

67

238

Bernhard Thöle

traditional account: consciously judging that ρ is connected with a direct awareness of that judgemental state which puts us in a privileged position to self-ascribe it. Furthermore, since being in such a state already includes a conceptualization of the proposition believed, it is even less likely 68 that we make mistakes through misidentification of such states than in cases of self-ascriptions of, e.g., sensations or perceptions. Nevertheles, this does not guarantee first person authority about beliefs. But to the extent to which there is first person authority about beliefs, it is easy to supplement the traditional explanation to account for this. For there is a strong, though contingent, causal connection between beliefs and corresponding conscious judgements, and we have direct knowledge of our conscious judgements. This completes the defense of the traditional explanation against Davidson's objections. I shall now turn to Davidson's own explanation to investigate whether it might be better suited as an explanation of first person authority. III. Davidson's explanation As already indicated, Davidson is primarily concerned with an explanation of the belief-asymmetry: he intends to explain why we presume that a speaker knows what he believes whereas there is no corresponding presumption on the side of the interpreter. In order to explain first person authority, Davidson confronts us with a situation in which the speaker possesses, but the hearer may not possess knowledge of a belief the speaker expresses on that occasion. In a first part, Davidson traces this asymmetry back to another underlying asymmetry: the reason why the speaker does, but the hearer may not, know what the speaker believes lies in a "presumption that speakers, but not their interpreters, are not wrong about what their words mean" (FPA 110). I call this underlying asymmetry the 'meaning-asymmetry'. In the second part of his argument Davidson tries to explain the meaning-asymmetry by claiming that this presumption "is essential to the nature of interpretation" (FPA 110): "There is a presumption — an unavoidable presumption built into the nature of interpretation — that the speaker usually knows what he means" (FPA 111). Let us first look at the situation Davidson describes: A speaker utters a certain (declarative) sentence, e.g. "Wagner died happy". Given that in this situation the speaker (a) means something by this sentence and (b) holds it to be true, he thereby expresses a belief. The question then is, under what conditions does the speaker, and under what conditions does an interpreter, know that a belief was expressed on that occasion, and 68

Of course we might make mistakes if we use a different conceptualization.

The Explanation of First Person Authority

239

what belief it was. Now Davidson claims that if — in this situation — both the speaker and interpreter know (i) that the speaker holds the sentence uttered to be a true sentence and (ii) what the speaker means by the sentence, both know what belief was expressed. Davidson goes on to claim that there is an asymmetry in the assurance of the speaker and the interpreter concerning the question whether (i) and (ii) are true. Let us call the asymmetry concerning (i) the attitude-asymmetry and the asymmetry concerning (ii) the meaning-asymmetry. In order to give a complete explanation of the belief-asymmetry, it seems, then, we have to explain the attitude-asymmetry as well as the meaning-asymmetry. As already said, the second part of Davidson's explanation contains an attempt to explain the meaning-asymmetry. One would expect, therefore, that the first part gives an explanation of the attitude-asymmetry. But, strangely, Davidson does not even attempt to explain the attitude-asymmetry 69 . Instead he presents — in the first part of his explanation — a confusing argument which seems to be designed to show that we need not explain the attitude-asymmetry in order to explain the belief-asymmetry. I shall consider that argument in a minute. But before I shall turn to a critical examination of the two parts of Davidson's explanation, some general remarks are in order. First, Davidson considers only a very special type of case — a case in which belief-attribution concerns a belief that is verbally expressed. But, of course, first person authority is not restricted to those beliefs a speaker happens to express verbally. Second, Davidson seems to be committed to the strange view that in the type of situation under consideration a speaker's knowledge of what belief he expressed derives from his (prior) knowledge of (i) and (ii). 70 But a speaker normally knows what beliefs he has before he is prepared to express them. 71 Third, Davidson altogether ignores all non-verbal evidence the interpreter might want to make use of in order to justify a belief-attribution. In concluding that the evidence he explicitly grants to the hearer does not suffice to justify a third person-attribution, Davidson simply begs the very question at issue — at least against those who claim that third personevidence is logically tied to what it is evidence for. The problem was to explain why the third person way is inferior to the first person way. But in the situation described, the greater part of the evidence the hearer might 69

70

71

This is especially surprising, since he criticized Bürge for his failure to account for the attitude-component (cf. Davidson (1988) p. 664). This is so since his explanation depends on the claim that the speaker knows what he believes in virtue of knowing (i) and (ii). Davidson stresses this obvious point in his criticism of Ryle (KOOM 441), but seems to ignore it when he presents his own explanation.

240

Bernhard Thöle

want to rely on is not even mentioned. Davidson is certainly right when he claims that knowledge of (i) and (ii) is sufficient to know what belief the speaker expresses; but he seems to assume — without argument — that knowing (i) and (ii) is the only way to arrive at knowledge of what belief a speaker has; and this is certainly false. Fourth, by a strange manoeuvre, Davidson changes the original problem. The original problem was to explain the asymmetry between firstand third person belief-ascriptions. But when he starts his explanation the topic is no longer belief-ascription but belief-expression. 72 Although one might argue that belief-ascriptions and belief-expressions are closely connected, they are certainly different. And what is especially relevant here, is that belief-ascriptions are subject to the possibility of a kind of error that does not (could not) occur when we consider belief-expressions only. To see this let us consider the following situation. Suppose you ask me to compute the sum of two numbers. I follow your instruction and conclude that the sum is, say, 13. I thus believe that the sum of the given numbers is 13. You now ask me to tell you my result and I answer — because of my wrong belief about the number of the apostles —: "I believe that the sum of your numbers equals the number of the apostles". Now 72

The move from belief-ascriptions to belief-expressions occurs in a short paragraph: "The question then comes to this: what explains the difference in the sort of assurance you have that I am right when I say '/ believe Wagner died happy' and the sort of assurance I have? We know by now that it is no help to say I have access to a way of knowing about my own beliefs that you do not have; nor that we use different criteria in applying the concept of belief (or the word 'believes'). So let us simply consider a shorter utterance of mine: I utter the sentence 'Wagner died happy'" (FPA p. 109 my emphasis). I am at a loss to see what justifies the 'so' in the last sentence. The way in which Davidson arrives at his question as formulated in the first sentence of the above quotation is also odd. Davidson introduces his own account by distinguishing "two related but different asymmetries. On the one hand, there is the familiar difference between self- and other-attributions of the same attitude to the same person: my claim that I believe Wagner died happy and your claim that I believe Wagner died happy. [...] On the other hand, we may consider my utterance of the sentence Ί believe Wagner died happy', and then contrast my warrant for thinking I said something true, and your warrant for thinking I have said something true". Davidson adds that "for reasons that will soon be evident, I shall deal with the second version of the asymmetry" (FPA 109). But what soon becomes evident is that Davidson should have chosen the first version, since what he then says is true only on the first and false on the second version: if I and you know that I held the sentence 'Wagner died happy' to be a true sentence, and we both know that I knew what that sentence meant on the occasion of its utterance, then we both know that I would have said something true if I had uttered the sentence Ί believe that Wagner died happy'. There would thus be no asymmetry of the second type, though there would still be an asymmetry of the first type since, although we both know — under the given assumptions — that my belief-ascription would have been true, you might not know that your corresponding belief-attribution ('You believe that Wagner died happy') would be correct, since you might not know what I meant by my original utterance.

The Explanation of First Person Authority

241

given the above assumptions, I certainly said something true about my belief-state. But in a sense, I did not give the right answer to your question: since before you asked me to report the result of my computation, I did not even think about the number of the apostles, I just thought that the sum is 13. Although I said something true about my belief-state, I did not correctly report the belief I was asked to report. The possibility of this kind of error is blurred, if we restrict our attention to belief-expression. For although I might misidentify my belief, this does not necessarily lead me to express a belief I do not have: if I express a belief through a process of misidentification of a given belief, I normally express a belief I have (as a result of my misidentification), although this is not the belief I intended to express. After these general remarks, let us now turn to the first part of Davidson's argument. It is not clear, what is going on in this part of Davidson's argument. As already noted, it does not contain what one would naturally expect, namely: an explanation of the attitude-asymmetry. Since Davidson does not explicitly say what he intends to show in the first part of his argument, let us look at what he presupposes in the next part. The aim of the second part is clearly stated: "It remains to show why there must be a presumption that the speakers, but not their interpreters, are not wrong about what their words mean" (FPA 110). In the rest of his paper Davidson tries to explain why this presumption is a necessary presupposition of interpretation. Since nothing in this explanation seems to depend on the first part of the argument, the only possible function of that part must be to establish that nothing more has to be explained. Therefore, I conclude that the aim of the first part is to show that the only thing that remains to be done is to explain the meaningasymmetry. Thus, the first part must contain either of two things, (a) it contains an explanation of the belief-asymmetry, on the provisional assumption that the meaning asymmetry holds, or (b) it contains a proof that in order to explain the belief-asymmetry we need only to explain the meaning-asymmetry. But as far as I can see, it neither contains (a) nor (b). Instead, Davidson claims that if a speaker utters the sentence "Wagner died happy", and some further assumptions are fullfilled, the speaker does, while the hearer need not, know what belief the speaker expressed by that utterance. The assumptions are that both the speaker and the hearer know that the speaker held the sentence to be a true sentence and that the speaker knew what the sentence meant. 73 It is certainly true that under these assumptions the speaker knows what belief he expressed, whereas the hearer might not 73

Cf. FPA 109 f.

242

Bernhard Thöle

know this since he might not know what the speaker meant by his utterance. But it does not follow that the only thing that remains to be shown is to explain why we are entitled to assume that a speaker generally knows what he means (whereas there is no corresponding presumption on the side of the interpreter). This was only one of the two assumptions: the other was that the speaker knows that he held the sentence to be a true sentence. If we want a complete explanation of why the speaker generally knows what belief he expressed, we need an explanation of both assumtions. 74 And since Davidson rightly points out that there is an asymmetry "in the assurrance you and I have that I hold the sentence I have just uttered to be a true sentence", we need some explanation of this asymmetry too. But Davidson nowhere even attempts to do this. Instead he simply claims that "we can assume without prejudice that we both [speaker and hearer] know, whatever the source or nature of our knowledge, that on this occasion I do hold the sentence uttered to be true" (FPA 109).75 To put the point in another way: if the belief-asymmetry is claimed to be a consequence of both the meaning- and the attitude-asymmetry, we have to explain both latter asymmetries in order to give a complete explanation of the former asymmetry. Since Davidson's first part does not contain an explanation of the attitude-asymmetry and the second part is exclusively concerned with an account of the meaning-asymmetry, we have to conclude that his explanation is at best a partial explanation of the belief-asymmetry. Let us now turn to the second part. The aim of this part is to show that "there is a presumption — an unavoidable presumption built into the nature of interpretation — that the speaker usually knows what he means" (FPA 111). Here — I think — Davidson offers us two different considerations. As far as I can see only the first of these depends on the nature of interpretation. Suppose again that I utter 'p\ Suppose you and I both know that I hold ρ to be true. Now Davidson claims that under these conditions it 74

75

It is irrelevant to point out that even if we assume that both speaker and hearer know about the attitude we still have an asymmetry. For this only shows that the attitudeasymmetry is not the only factor that accounts for the belief-asymmetry. It does not show that the attitude-asymmetry does not contribute to the belief-asymmetry. And since it does, it needs to be explained. He even goes so far as to claim that in making these assumptions "we have not postulated or assumed any asymmetry at all" (FPA 110). One is tempted to ask why we could not likewise assume 'without prejudice' that both speaker and hearer know what the speaker meant by his utterance. If we added this assumption, the asymmetry would disappear. But I see no reason why assuming knowledge about the attitude would not amount to assuming an asymmetry, while assuming knowledge about the meaning would.

The Explanation of First Person Authority

243

cannot be generally true that I do not know what I believe. The argument for this claim is this: (a) it is a necessary presupposition of interpretation that a speaker does not just utter words at random; he must "use a finite supply of distinguishable sounds applied consistently to objects and situations he believes are apparent to his hearer". In short, he must 'get his language right', "for otherwise there would be nothing for the interpreter to interpret" (KOOM 456). No doubt, this is true. It is a condition of interpretation that a speaker must be interpretable, and a speaker would not be interpretable if he would not coherently use his language. From this incontestable fact Davidson quickly moves to the — certainly stronger — conclusion that a speaker, in order to be interpretable, must know what he means. Given this, Davidson can infer that our speaker must generally know what he believes76. For given that he knows that he made a sincere assertion and given that he knows the meaning of that assertion, he knows the belief that was expressed by this assertion. But Davidson gives us no reason why it is a precondition for interpretation that as speaker knows what he means. All he says supports only the weaker and trivial point that interpretation requires interpretability, i. e., a coherent use of language. Davidson seems to arrive at his stronger conclusion by sliding from: (a) A speaker's utterances are (in general) interpretable through (b) A speaker's utterances are (in general) meaningful and (c) A speaker (in general) means something by his utterances to (d) A speaker knows (in general) what he means by his utterances. The step from (b) to (c) is certainly not self-evident. But the most problematic step is from (c) to (d). The following passage is especially revealing: unless there is a presumption that the speaker knows what she means, i . e., is getting her language right, there would be nothing for the interpreter to interpret. To put the matter another way, nothing could count as someone regularly misapplying her own words. (KOOM 456)

In order to reach his intended conclusion, Davidson — in this passage — goes so far as to identify 'knowing what she means' with 'getting her 76

This, of course, applies only to those beliefs that he expresses in words (or can express in words).

244

Bernhard Thöle

language right'. 77 This identification is certainly unacceptable: we certainly would not say that a speaker knows the meaning of two of her utterances if she were agnostic about whether or not they mean the same. But I see no reason why a speaker who is agnostic about sameness of meaning should be uninterpretable.78 Furthermore, a speaker might be interpretable, without being able to explain what he means. In saying this, I do not mean to deny that a speaker who means something by an assertion, will also (normally) know what he means.79 But this is just a special case of first person authority — in fact, it is the meaning asymmetry: If I mean something by ρ, I know what I mean but you may not. True, but that is what Davidson professed to explain. Instead he just assumes it 80 . Thus, what Davidson needs is the stronger principle — namely, that it is a necessary presupposition of interpretation that the interpreter presumes that the speaker does in general know what he means if he means something. Although Davidson says this, he does not say why we should think that it is true. And although I agree that a speaker does in general know what he means (if he means something), I see no reason to think that this is true because it is a presupposition of interpretation. But if it is not shown to be such a presupposition, and shown to hold because it is such a 77

78

79

80

Perhaps Davidson would reply that a speaker must know the meaning of his words in order to mean something. There certainly is a sense in which 'knowing the meaning of x' and 'meaning something by x' are nearly equivalent. But this is clearly not the intended sense when we claim to have first person authority about our beliefs. (It would be satisfied e.g. when I tell stories while asleep.) And it certainly amounts to nothing more than the disposition that is manifested in meaningful (interpretable) utterances. Nothing follows about the truth or reliability of first-person-second-order ascriptions. And nothing follows — without further argument — about the ability to explain the meaning, or about the ability to answer questions about sameness of meaning. This would perhaps hold on a crude behavioristic account of 'knowing the meaning', but endorsing a behaviouristic conception would threaten first person authority rather than explain it. I think this is true of occurrent concious belief-expression. But it need not be true in general, e.g., of speaking while sleeping etc. In another passage Davidson suggests that a speaker must generally know what his words mean, since he "cannot wonder whether he generally means what he says" (FPA 110). To back the latter claim, Davidson points out that a speaker "is not in a position to wonder whether she is generally using her words to apply to the right objects and events, since whatever she regularly does apply them to gives her words the meaning they have" (KOOM 456). Let us concede the last claim. Then the speaker cannot wonder whether she is generally misapplying her words — if she knows that she does use her words consistently. But how does she know this? And why does she know this better than an interpreter? But even if we can satisfactorily answer these questions, it does not follow that she knows what she means; it only follows that she knows that she means something. Again, I agree with Davidson that under the conditions stated, the speaker does in fact normally know what she means. But to say this is not to explain it.

The Explanation of First Person Authority

245

presupposition, it is just a special case of first person authority — the very phenomenon that Davidson promised to explain. The first argument, thus, assumes what it intends to explain. But Davidson appeals to another "familiar, though often misunderstood, point: I can do no better, in stating the truth conditions for my utterance of the sentence 'The Koh-i-noor diamond is a crown jewel' than to say it is true if and only if the Koh-i-noor diamond is a crown jewel. But for you this is an empirical claim, though probably a true one" (WPM 17 f.) or: The speaker, after bending whatever knowledge and craft he can to the task of saying what his words mean, cannot improve on the following sort of statement: "My utterance of 'Wagner died happy' is true if and only if Wagner died happy." An interpreter has no reason to assume this will be his best way of stating the truth conditions of the speakers utterance. (FPA llOf.)

Now one might reasonably doubt the claim that this simple way of saying what my words mean cannot be improved. But we should certainly agree with Davidson that this way of telling the meaning of a sentence provides the speaker, but not the interpreter, with a secure method for saying what he means. 81 Thus Davidson's appeal to his dear "familiar, though often misunderstood, point" does something to explain the meaning-asymmetry. But it can hardly account for much. It certainly does not explain why a speaker who utters a meaningful (declarative) sentence knows the meaning of this sentence. For we would certainly not credit a speaker with knowledge of what he means by a certain sentence, if he could do no more to explain his meaning than stating its truth-condition by using that very sentence. If a speaker knows the meaning of two of his sentences, he should be in a position to say whether or not they mean the same. 82 Being equipped with Davidson's simple method does not put a speaker in this position. Davidson's second argument, therefore, can hardly count as a satisfactory explanation of the meaning-asymmetry. This concludes my discussion of Davidson's account of the beliefasymmetry. I add only one more general comment: Davidson claims that "what holds for the propositional attitudes ought, it seems, to be relevant to sensations and the rest" (FPA 102). But even if we can fill the gaps in 81

82

Given that he means something by ρ (and given that ρ is a declarative sentence) this method would provide the speaker with an infallible way of saying what he means. On the other hand, it would not be a very helpful way to inform an interpreter about the meaning of the utterance in question. This, by the way, shows that Davidson is wrong when he claims that externalism about meaning and mental content does not threaten first person authority (cf. KOOM). For suppose the external conditions are such that two of a speaker's sentences mean the same. Since — given externalism — the speaker need not know this fact about the external conditions, he might not know that the two sentences mean the same. The same reasoning applies to belief-content.

246

Bernhard Thöle

Davidson's explanation of the meaning-asymmetry, it is very doubtful that his explanation can be extended to cover other cases of first person authority. Since his account is exclusively concerned with first person authority concerning the prepositional content beliefs, it is hard to see how it could be relevant to mental states (like sensations) that lack propositional content. Thus, I think Davidson's explanation fails. It is restricted to a very special type of situation; it does not give a satisfactory explanation of the meaning-asymmetry; it does not even address the attitude-asymmetry and it is hard to see how it could be extended to first person self-attribution of sensations. Having found these faults the principle of charity strongly suggests that I have radically misinterpreted Davidson's meaning. But on this question only he has first person authority.83

REFERENCES

Alston, W.P.: 'Varieties of Privileged Access.' American Philosophical Quaterly Vol. 8 (1971) — 'Self-Warrant: A Neglected Form of Privileged Access.' American Philosophical Quaterly Vol. 13 (1976) — Epistemic Justification. Ithaca 1989 Armstrong, D.M. A Materialist Theory of the Mind. London 1968 — 'Consciousness and Causality.' in: Armstrong, D.M., Malcolm N.: Consciousness and Causality. Oxford 1984 — Belief, Truth and Knowledge. Cambridge 1973 Ayer, A. J.: The Problem of Knowledge. Harmondsworth 1956 — 'Privacy.' in: The Concept of a Person and Other Essays. London 1963 Boghossian, P.A.: 'Content and Self-Knowledge.' Philosophical Topics Vol. XVII (1989) Broad, C.D.: The Mind and its Place in Nature. London 1925 Bürge, Τ.: 'Individualism and Self-Knowledge' The Journal of Philosophy. Vol. LXXXV (1988) Churchland, P.: Matter and Consciousness. Cambridge 1984 Davidson, D.: Ά Coherence Theory of Truth and Knowledge, in: D. Henrich (ed.): Kant oder Hegel. München 1983 — 'First Person Authority' Dialectica Vol XXXVIII (1984) (cited as FPA) — 'Knowing One's Own Mind' Proceedings and Addresses of the American Philosophical Association (1987) (cited as KOOM) — 'Reply to Bürge' The Journal of Philosophy Vol. L X X X V (1988) — 'The Conditions of Thought' Grader Philosophische Studien Vol. XXXVI. (1989a) — 'What is Present to the Mind? Grader Philosophische Studien Vol. XXXVI. (1989b) — 'Meaning, Truth and Evidence' in: R. Barrett/R.Gibson (eds.):Perspectives on Quine. Oxford 1990 Heil, J.: 'Privileged Access.' Mind Vol. XCVII (1988) 83

I am grateful to Marshall Farrier, Stefan Gosepath, Michael Hardimon, Mark Hetterich, Markus Otto, Beate Rössler, Gottfried Seebaß, Alexander Staudacher and Ursula Wolf for helpful comments and criticisms.

The Explanation of First Person Authority

247

Hume, D.: A Treatise of Human Nature. Ed. by L. A. Selby-Bigge. Oxford 1975 Hyslop, A./Jackson, EC.: 'The Analogical Inference to Other Minds'. American Philosophical Quarterly Vol. IX (1972) Jackson, F.C.: Perception. Cambridge 1977 Leibniz, G.W.: Nouveaux Essais sur l'Entendement Humaine. in: G.W. Leibniz: Die Philosophischen Schriften V. Ed. C.I. Gerhardt. Berlin 1882 Locke, J.: An Essay concerning Human Understanding. Ed. by P. H. Nidditch. Oxford 1975 Mackie, J.L.: 'Are there any Incorrigible Empirical Statements?' Australesian Journal of Philosophy Vol. 41 (1963) McGinn, C. The Character of Mind. Oxford 1982 Rorty, R.: 'Incorrigibility as the Mark of the Mental' The Journal of Philosophy. Vol. L X V I I (1970) Shoemaker, S.: Self-Knowledge and Self-Awareness. Ithaca 1963 — 'On Knowing One's Own Mind.' Philosophical Perspectives Vol. II. (1988) — 'First person Access.' Philosophical Perspectives Vol. IV. (1990) Sosa, E.: Knowledge in Perspective. Cambridge 1991 Strawson, P.F.: Individuals. London (1959) Wittgenstein: Philosophical Investigations. Oxford 1953

Donald Davidson: Reply to Bernhard Thöle Like Bernhard Thöle, I am by no means satisfied with my explanation of first person authority. Much more remains to be said in defense of my position, and indeed it may be that there are aspects of first person authority that cannot even be approached by following the line I have suggested. It is clear, too, that my treatment of the topic is incomplete, because I have said nothing about knowledge of our own sensations. We share our dissatisfaction with my work, but apparently that is about all we share. I still think the "traditional" explanation of first person authority, which Thöle attempts to defend, is no explanation at all, and I do not believe my own explanation is as riddled with confusion and bad reasoning as he holds. Since I shall be able to take up only a small number of Thöle's criticisms, I would urge anyone who is interested in my position to read my paper rather than accept Thöle's tendentious and inaccurate account of it. The question I set out to answer is: what explains the difference between my knowledge of my own thoughts and the thoughts of others, or the closely related difference in the justification we have when we attribute thoughts to ourselves and to others? I used a variety of phrases to express this difference: I spoke of a "presumption" that we are right in the first sort of case which is lacking in the second, and of the "authority" with which we (usually) make sincere self-attributions, an authority we do not have for our other-attributions. I invested nothing in these phrases. What I did invest in was the idea (hardly unique to me) that knowledge of our own thoughts, unlike our knowledge of the thoughts of others, is not generally based on evidence or observation. It is normally pointless to ask someone why she believes she has the thoughts she claims to have; the request for reasons or justification is out of place. I rejected a number of purported explanations on the grounds that they merely redescribed the problem. I took this line with respect to some well known proposals of Strawson and Shoemaker. Strawson, concentrating on the linguistic form of the asymmetry, but obviously believing this to throw direct light on the epistemological version, observed that we correctly apply predicates like "believes that Saturn is smaller than Jupiter" on the basis of (behavioral) evidence when we apply them to others, but without evidence when we apply them to ourselves. I agreed that this was

Reply to Bernhard Thöle

249

true, but that it merely restated the asymmetry without explaining it. I thought Shoemaker's claim that there is a linguistic convention to the effect that we generally treat self-attributions of attitudes as true was not much better, since it gave no hint as to why we should have such a convention. We need a convention about what side of the road to drive on; but why should we arbitrarily elect to treat the same predicates in two very different ways? The "solution" Thöle defends seems to me equally empty. He puts great weight on the notion of awareness, a notion he says almost nothing to explicate. I suppose that if I am aware of my thoughts, this means, or at least implies, that I am aware that I have those thoughts, and this certainly implies both that I do have those thoughts and that I believe I do. So far, this is no help. For if I am aware that you have some thought, you must have it, and I must believe that you have it. The difference, Thöle tells us, is that my awareness of my own thoughts is "direct", of yours, indirect. This difference is in turn made to turn on whether or not the awareness depends on (behavioral) evidence. Good; that's the problem: we know our own thoughts without need of the kind of evidence that others need. But why is that? In my opinion, what Thöle calls the "traditional" solution does not even attempt to answer my question. What it assumes — direct, inner awareness — is, under the somewhat obscure wording, the very thing that sets the problem. It seems to me this was obvious to those who first formulated the problem. That is how the traditional problem of other minds is often stated: we cannot know other minds in the "direct" way we know our own, so why should we think we can know other minds at all? I simply generalized the skeptical problem by changing the challenge to: if we have two very different ways of applying predicates like "believes Saturn is smaller than Jupiter", why should we think such predicates have the same meaning in both applications? The advantage of putting the question this way is that it makes clear why it is no answer to describe how we in fact do justifiably attribute thoughts to others. My strategy in thinking about the problem was this: to state the asymmetry in terms of directness of awareness, or the way our words are used is merely to restate the problem, and leave us with no answer to the skeptic. But the asymmetry to be explained must depend on some other asymmetry, so we should seek an asymmetry that can be accepted without inviting skepticism. I thought I found this asymmetry in how we understand what we and others say: I argued that our knowledge of what others mean by what they say must depend on observation, while the question how we know what we mean by what we say cannot, in general, even be raised. For the sake of making this point, one might decline to use the word "knowledge" for what accounts for our correct use of words to express our thoughts; but it seems to me enough to see why it is that

250

Donald Davidson

justification is simply out of place here. The reason is (to put the central point without the necessary precautions) that a person cannot generally misuse his own words, because it is that use which gives his words their meaning. What a speaker says can be misinterpreted by others, but it cannot be misinterpreted by the speaker because no content can be given to the idea of interpreting one's own words. Here I endorse a fairly extreme form of individualism about meaning, and it is a view that is not accepted by those who think languages are defined by shared practices. But I have defended my approach to language elsewhere; I mention it here because my article "First Person Authority" perhaps did not sufficiently emphasize that my "solution" to the problem about self-attributions of attitudes depended on my theory of meaning. There is, then, or so I argue, a presumption that a person is correctly expressing his thoughts whenever he intends to. This does not, of course, mean that the beliefs a speaker expresses are true. But it does mean that with respect to an essential ingredient in self-attributions of beliefs there is an asymmetry between speaker and hearer with respect to truth: there is a justified presumption that the speaker has not misidentified the content of his belief, while there is no such presumption with respect to the hearer's knowledge of the content of the speaker's belief. Thöle complains that I simply assume that the speaker knows that the content-sentence he utters is one he believes is true, and that I give no reasons for assuming this. This is true: but I assume the same for the hearer. Thöle has missed the thrust of the argument. As I said in my essay, if there is asymmetry with respect to the content of the attitude, there is also asymmetry with respect to the claim that the attitude exists. But all I set out to show was that there was a natural explanation, which did not in itself invite skepticism, of the asymmetry between first and third person attributions of attitudes, and to show this all I had to do was point out some asymmetry that was not mysterious and did not invite skepticism. In order to do this, I assumed (contrary to fact) that there was no asymmetry with respect to knowledge of the existence of the attitude in order to prove that nevertheless one could demonstrate an asymmetry. It does not damage this argument if there is a further asymmetry.

The Architecture and Evidential Base of the Unified Theory LORENZ LORENZ-MEYER The attempt to make sense of our fellow human beings' behaviour requires interpretation of their verbal and nonverbal actions as well as interpretation of the words they use in their utterances. Professor Donald Davidson holds that neither of the two tasks can be accomplished independently. His claim that the interpretation of words and the interpretation of deeds are mutually dependent motivates the search for a theory whose application would make it possible for an interpreter to solve both problems at the same time. In this paper I shall try to recapitulate some of the reasons Davidson gives for his thesis of the mutual dependency as it becomes relevant in the context of his own research. I shall follow him on his way toward a Unified Theory of Meaning and Action and on the go I will give some considerations about the special role Bayesian decision theory is supposed to play in this enterprise. I shall give a very short sketch of the architecture of the Unified Theory as it can be reconstructed out of Davidson's sparse indications. And I shall end up making some critical remarks about the attitude of 'preferringto-be-true', which is supposed to be the successor of the attitude of'holdingtrue', providing the evidential base for radical interpretation.

I. A Davidsonian theory of meaning is a finitely axiomatized empirically verifiable semantical theory based upon Tarski's recursive characterization of the truth-predicate for certain formalized languages. Empirical verification of such a theory under the conditions of radical interpretation, i. e. without any previous understanding of the object language, rests upon the identification of cognitive attitudes of the speakers towards their sentences, prominently the attitude of holding a sentence true. The point is that this attitude is more open to observation than the beliefs or thoughts which the speakers intend to express by their utterances. It might, for example, be relatively easy to identify acts of sincere assertion or public assent as manifestations of holding a sentence true.

252 A. DAVIDSON'S

Lorenz Lorenz-Meyer THEORY

OF

MEANING:

But even this attitude is a kind of belief, and if we are to read off beliefs from a speaker's behaviour we have to make assumptions about his desires, for the actions of a person depend on her or his desires as well as beliefs. For example: if we are to identify an utterance as an act of sincere assertion we have to know whether the speaker has the desire to forward information. In radical interpretation we have to make assumptions about the desires, intentions, etc. that constitute the motivational background of actions which in turn could be taken as manifestations of the speakers' attitudes towards their sentences; and those assumptions are in the need of independent evidential support. This fact suggests that the theory of meaning requires a supplementary theory of the speakers' actions as seen in the light of their beliefs and desires. II. Before I turn toward theories that explain actions by means of belief and desire ascription, let me cite another deficiency of the narrow picture of radical interpretation based simply upon the attitude of 'holding-true'. If the interpreter wants to specify the role of single expressions of the speakers' language she needs to know a lot about the inferential relations of sentences that contain these expressions. And this pertains not only to deductive but also to inductive relations. Davidson argues: "It is obvious that a correct interpretation of a speaker's words will depend heavily on knowing to what extent the speaker counts the truth of one sentence in support of the truth of another. For the content of sentences and predicates more or less remote from

The Architecture and Evidential Base of the Unified Theory

253

what is immediately observed depends on what is taken to favor their truth or application, while the meaning of a sentence tied more directly to what is observed is partly determined by the theoretical truths its truth is taken to augment." 1 In a later version of the quoted text Davidson calls this phenomenon the "interanimation" of sentences and stresses the probabilistic and holistic character of the relevant relations. 2 What is required, then, is more information about the speakers' attitudes towards their sentences. It would be advantageous for the prospects of radical interpretation if the interpreter could use information not only about whether a speaker holds a certain sentence true but also about the 'strength' of the speaker's holding true, i. e. the probability he ascribes to the truth of his sentences. So much for the demands of a theory of meaning. Now I want to turn to the perspective of action theory.

III.

If we want to conceive ourselves as thinking, deliberating, acting beings, in short: if we want to treat ourselves as persons and not as biological automata, we are forced to see and explain our behaviour against the background of a richly developed mental life. In the variegated multiplicity of psychological factors that contribute to the picture we make of a person in its surroundings, belief and desire play a prominent role. This is due to the part they play in ordinary reason explanations. We are used to seeing reason explanations as the main kind of everyday action explanations and the most important contribution to the rationalisation of a person's behaviour. The shortcomings of this kind of explanation have brought psychological theory into disrepute, but I think that there is still some need for clarification about what exactly we do when we explain an agent's deed by giving his reason for doing it. One important point is that reason explanation seems to be not strong enough even within the limits of rationality. Davidson, in his famous article "Actions, Reasons, and Causes" speaks about an "ex postfacto atmosphere of explanation and justification, as the reason frequently was, to the agent at the time of action, one consideration among many, a reason." An agent, when deciding what to do, very commonly has to weigh different, possibly competing, desires and must consider the probabilities of circumstances that might influence the outcome of his doings. This fact brings Davidson to the 1

2

D. Davidson: "Toward a Unified Theory of Meaning and Action", in Grader Philosophische Studien 11, 1980, p. 7. D. Davidson: "The Structure and Content of Truth", in The Journal of Philosophy 87, 1990, p. 321.

254

Lorenz Lorenz-Meyer

conclusion that "any serious theory for predicting action on the basis of reasons must find a way of evaluating the relative force of various desires and beliefs in the matrix of decision". 3 I think that the relevant aspect is not that reason explanations have an irreducible ex post facto character — we very often do predict behaviour by giving a reason for it — but rather that the explanation or prediction of an action by giving a single reason is elliptical in a way that is very difficult to specify. A belief and desire constitute the reason for a piece of behaviour only as seen in the context of a whole intentional profile that is deliberately left out of account because its relevant features are taken for granted. And if we give a reason explanation, it is not the pair of belief and desire alone but the whole intentional profile that we suppose to be causally relevant for the action.

IV. The quest for a "quantitative calculus that brings all relevant beliefs and desires into the picture" 4 leads Davidson back to the roots of his career, to the consideration of Bayesian decision theories. Those theories make it possible to give a precise quantified characterization of the beliefs and desires of a person on the basis of her or his preferences. According to the different types of theory this might mean preference between betting options, possible worlds, the obtaining of states of affairs, or truth of propositions. The implicit assumption that the preferences of a person are manifested more directly in her or his actions than the relevant beliefs and desires provides decision theory with greatest import as part of a general explanatory theory of behaviour. The strategy which is at the core of decision theories is backed by everyday conceptions in a similar fashion as the philosophical elaborations of reason explanations. Decision theories model in a precise way our ordinary concept of deliberation and its function in action explanation.

V. It is very enticing to let those elegant and powerful theories do their duty in the frame of a general interpretative theory of behaviour. But there are several problematic aspects worth mentioning. Let me begin with some general points: 3

4

D. Davidson. "Actions, Reasons, and Causes", in Essays on Actions and Events, Oxford 1980, p. 16. D. Davidson: "Psychology as Philosophy" in Essays on Actions and Events, p. 233.

The Architecture and Evidential Base of the Unified Theory

255

First, with the application of Bayesian theories one departs from ordinary usage of the words "belief" and "desire". Unlike our ordinary concept of belief the Bayesian concept allows for continuous gradation between belief in the truth of a thought content and belief in the truth of its negation. And unlike our ordinary concept of desire Bayesian desire includes negative values. On the other hand, Bayesian negative desire with regard to a thought content resolves into ordinary positive desire with regard to its negation. And degrees of certainty or subjective probability with respect to thought contents play an important role in ordinary action explanation. This fact calls in question the central role that traditional action theory assigns to the notion of (unquantifled) belief. Second, Bayesian theories don't distinguish between different degrees of certainty and different degrees of subjective probability with respect to thought contents. Both factors are viewed from the perspective of their causal influence on our behaviour and bundled as one action-determining dimension. I think that this simplifying identification, inevitable as it might perhaps be for the first steps of interpretation, is in the long run untenable. Third, Bayesian theories are applicable only to preferences that are rational in a sense characterized by the theory (but inspired by our ordinary conception of rationality). Those theories are prescriptive rather than descriptive. This is an often-discussed fact and its consequences are in dispute 5 . Davidson emphasizes that it is not a clear-cut empirical question whether an agent acts rationally. To set the interpretation of actions going we have to find rational patterns in the preferences of the agent. A 'descripitive' empirical use of normative decision theories has to be supported by a generous application of some kind of'principle of charity'. The strategy is sometimes called by Davidson the " p o l i c y of rational accomodation" .6 Maintaining the hypothesis of rationality, we could perhaps infinitely accomodate our interpretative means to integrate recalcitrant behaviour. But this is not recommendable, because there certainly are limits to rationality, as Davidson is the last to deny. Having — via rationalization — obtained some primary understanding of an agent's mental life, we will have to provide interpretatory techniques which make allowance for the temporary occurrence of irrational behaviour. But these techniques presuppose the ascription of a wealth of propositional attitudes, and this in turn depends on the rationalizability of a majority of the agent's actions. 5

6

Compare Carl G. Hempel's article "Rational Action", in Proceedings and Adresses of the American Philosophical Association, Yellow Springs (Ohio) 1962, pp. 5 — 24, and Davidson's reply: "Hempel on Explaining Action", Essays on Actions and Events, pp. 261—275. D. Davidson: Expressing Evaluations, The Lindley Lectures of the University of Kansas, 1984, p. 18.

256

Lorenz Lorenz-Meyer

Fourth, it is more than probable that a Bayesian ascription of propositional attitudes will conflict with the psychological self-description of the agent. 7 In some way, decision theory contradicts the paradigm of first person authority with respect to the mental. But for the ascription of propositional attitudes of an agent, we can do just as little without his first person knowledge mediated by his speech as we can do without third person perspective rationalization. 8 The former is an indispensible corrective to the latter, if interpretation is not to become arbitrary to a large degree. So there exists a certain tension that should not be left out of account if one wants to allocate an important role in action explanation to decision theory. These four points suggest that Bayesian theories could provide at most a core element to a suitable and developed scientific theory of action which could keep up with the conceptual richness and flexibility of our everyday action explanation. VI. The decision calculus Davidson employs for his purposes was designed in the 60s by the logician Richard Jeffrey and the mathematician Ethan Bolker. 9 Essentially it is a development and elaboration of Frank Ramseys pathbreaking study "Truth and Probability" from the year 1926.10 Ramsey solved the problem of measuring the intensities of beliefs and desires which are manifested in the behaviour of a person by considering preferences with respect to betting options. He described a procedure which allows to isolate — on the basis of betting behaviour — the subjective prob7

For an illustration, let me quote Carl G. Hempel reporting the results of Davidson's research in the 50s about the empirical verifiability of decision theory: "Though the subjects make their choices in clearly structured decision situations, with full opportunity for antecedent deliberation and even calculation, they act rationally (in a precisely refined quantitative sense) relative to subjective probabilities and utilities which they do not know, and which, therefore, they cannot take into account in their deliberations; they act rationally in the sense of acting as if they were trying to maximize expected utilities. We seem to have here a type of conscious choice which is non-consciously rational with quantitative precision. What might Freud have thought about this?" "Rational Action", p. 22.

8

For Davidson, the thesis of 'first person authority' boils down to the fact that for interpretation of speech it is a necessary condition that a speaker can and often does non-inferentially know what he believes or desires. Cf. D. Davidson: "First Person Authority", Dialectica 38,1984, pp. 1 0 1 - 1 1 1 . Richard Jeffrey: The Logic of Decision, New York 1965. The revised second edition (Chicago 1983) does not contain the 9th chapter of the first edition, which is very important for Davidson's concerns, because it contains the description of the measurement procedure, which determines an agent's desirability and probability functions. Frank P. Ramsey: "Truth and Probability", in D. H. Mellor (ed.): Foundations, London 1978, pp. 5 8 - 1 0 0 .

9

10

The Architecture and Evidential Base of the Unified Theory

B. RAMSEY'S DECISION

257

THEORY:

abilities of the states of affairs that constitute the objects of the bets on the one hand and the utilities of the winnings and losses on the other. Richard Jeffrey succeeded in generalizing this procedure: his calculus does not consult preferences with respect to betting options as a starting point, but preferences with respect to the truth of propositions. The basic relation which establishes the preference ordering is the following: Person S prefers the truth of proposition Ρ over the truth of proposition Q. If a person's preference ranking fulfills certain nonquantitative conditions 11 , then Jeffrey's calculus makes it possible to determine a probability function and a desirability function for the person, which assign to each proposition in the preference ranking numerical values for subjective probability and desirability. For the gain in generality and the more transparent ontology in comparison with standard theories, Jeffrey accepts two drawbacks: First, the probability and desirability functions are to a larger degree underdetermined by their preference rankings. A greater number of equivalent functions 'explain' the same preferences. Second, the techniques for calibration of the scales and for measurement are far more complicated. Apparently these two disadvantages don't bother Davidson very much. As for the second, I think it should. Even though his considerations about human behaviour interpretation are not meant to give a realistic picture of 11

These conditions are stated in an informal way in the new 9th chapter of the second edition

of Jeffrey's Logic of Decision (see Fn. 9).

258 C. JEFFREY'S

Lorenz Lorenz-Meyer DECISION

THEORY:

our actual tricks and means, the plausibility of his enterprise depends on whether he succeeds in describing an interpretatory procedure that is practicable at least in principle. I'm not a specialist in questions of complexity, but I think that anyone who shares with me the experience of reading chapter 9 of the first edition of Jeffrey's Logic of Decision will agree that a lot of complicated work has to be done before even a static picture of an agent's psychology emerges in the process of Jeffreyan behaviour interpretation. Davidson is worried by some other feature of the Jeffrey calculus, a feature also to be found with the more traditional theories. They don't consider the question of how the propositions (or betting options) which are the objects of an agent's preferences are identified. The determination of the propositional content of these preferences normally rests upon interpretation of the agent's utterances. In a normal psychological application of decision theories, the experimental setting provides for a lively verbal exchange between agent and interpreter. And one quite naturally presupposes that both parties share the same language or at least understand each other. If one wants to abandon this assumption to get a faithful model of the 'radical interpretation' scenario, it becomes clear that decision theory is in the need of a supplementary theory that allows to determine the meaning of the agent's sentences. So this is where the circle closes. VII. In the previous paragraphs I tried to elucidate how in the context of his own studies, Davidson's general thesis about the mutual dependency of action theory and meaning theory is confirmed. There might be other approaches

The Architecture and Evidential Base of the Unified Theory D. DAVIDSON'S

259

MODIFICATION:

to language and behaviour, but if Davidson is right they all at some time or other will have to meet the challenge and take into account the interrelation between the understanding of words and the understanding of deeds. Now I will try to show how Davidson himself envisions a solution to the problem. The key to a solution lies in the fact that Jeffrey's measuring procedure presupposes only very little knowledge about the interior structure of the propositions: it is only their truth-functional structure (in the sense of the propositional calculus) that must be known to the interpreter. If it would be possible to identify the sentential connectives in the language of the agent, one could take the easily accessible (still uninterpreted) sentences of the agent's language instead of the hidden propositions as the objects of preference. Davidson solves this exercise by describing a procedure for locating the Sheffer stroke, the one sentential connective which defines every other. 12 So it becomes possible to apply a modification of the Jeffrey calculus, which does not consider the agent's preferences with respect to the truth of propositions but preferences with respect to the truth of his sentences. And for those sentences it is no longer assumed that the interpreter knows their meaning in advance. As results of an application of this theory the interpreter gets two special psychological attitudes of the agents towards their sentences: one is a quantified variant of the familiar attitude of holding-true, the other might be called quantified desiring-to-be-true. The correlation between changes in these attitudes of an agent on the one hand and alterations in his environment on the other hand provides an excellent data base for empirical semantics of his language. This fact makes it pos12

D. Davidson: "A New Basis for Decision Theory", Theory and Decision 18, 1985, p. 95.

260 Ε. THE ARCHITECTURE

Lorenz Lorenz-Meyer

OF THE UNIFIED

THEORY:

sible to fuse the modified Jeffrey calculus and the theory of meaning and get one big Unified Theory of Meaning and Action. An interpreter who puts this theory into practice first finds out the extent to which an agent wants which sentences to be true and how probable he holds those sentences to be. And he secondly finds out what those sentences mean. Bringing both results together he is thirdly in the position to use the

The Architecture and Evidential Base of the Unified Theory

261

sentence meanings to determine the propositional content of those beliefs and desires which are mirrored in the attitudes of holding-true and desiringto-be-true. In this way decision theory is supplemented by an interpretative meaning theory as required, and both challenges concerning the isolated theory of meaning have been met: the interpreter is furnished with detailed information about the subjective probabilities the speaker assigns to his sentences; this allows him to take the inductive relations between the sentences into account. And for the identification of those acts, in which the relevant psychological attitudes of the speaker towards his sentences are manifested, it is no longer necessary to make assumptions about his desires which are in the need of independent evidential support. But other difficulties arise.

VIII. In the 'narrow picture' of radical interpretation, the relevant psychological attitude of the speaker towards his sentences was the attitude of holding true. This is a very simple two-place relation which holds or does not hold between the speaker and each of his sentences. The attitude of holding true shows itself in frequent acts of assertion or assent. And the advantage of those acts is that they directly present the relevant sentences. There is nothing comparable in the case of the Unified Theory. The attitude of preferring true is not a two-place but a three-place relation between the speaker and every two sentences of his. And no speech act of the speaker directly supplies the interpreter with enough information about it to construe the preference ranking that is needed to set Bayesian interpretation going. In a footnote to the third part of his Dewey Lectures Davidson argues: "[...] every utterance that can be treated as a sincere request or demand may be taken to express the utterer's preference that a certain sentence be true rather than its negation." 13 This observation is certainly correct, but it doesn't get us much further. Common sense tells us, and Jeffrey's theory faithfully reflects the fact, that if a proposition is valued positively then its negation is not valued positively (and vice versa). 14 So everything that can be expected of a close observation of directive speech acts alone is an arrangement of the agent's sentences into two classes: one class of all sentences valued positively, another of those not valued positively. No further detailed information about the preference arrangement can be drawn from this basis. In his footnote, Davidson makes another proposal: "Most experimental work in decision theory takes as data the choices subjects make between 13 14

D. Davidson: "The Structure and Content of Truth", p. 325; fn. 67. Richard Jeffrey: The Logic of Decision (2nd ed.), p. 82.

262

Lorenz Lorenz-Meyer

alternatives as described in writing or speech. It is normally assumed that subjects understand these ascriptions as the experimenter do. Dropping this assumption yields data of exactly the sort required by the approach presented here." 15 But does this describe a practicable procedure for radical interpretation? Wasn't it exactly this picture of a quasi-omnipotent scientist limitlessly offering his experimental subjects options about the further courses of the world that gave Ramsey's ingenious scheme such an unrealistic flavour? It might and perhaps will be considered unsporting to confront a project which is conceived merely as a "conceptual exercise" 16 with questions of practicability. But as I already said when talking about problems of complexity, I think that the relevance and plausibility of Davidson's considerations hinge upon his describing a procedure that is realisable at least in principle. And in my eyes this means more than mere logical possibility. Professor Davidson's design of a Unified Theory as it stands is impressive and illuminating indeed. But with respect to its practical dimension I conclude that some questions are still open.

15 16

ibid., p. 325. D. Davidson: "Toward a Unified Theory of Meaning and Action", p. 12.

Donald Davidson: Reply to Lorenz Lorenz-Meyer I'm in agreement with the last sentence of this paper: there is much work to be done. But while sympathizing with some of Lorenz-Meyer's misgivings, I would emphasize different problems than he does as those that need work. Lorenz-Meyer worries that the Bayesian concepts of subjective probability and preferential strength have little to do with our ordinary concepts of belief and desire. This seems to me quite wrong, and I think the criticisms are based on a common misunderstanding of Bayesian theories. A central point to bear in mind is that such theories say nothing at all about the numerical strength of desires and beliefs. These quantitative concepts do not appear in the statement of the theories, nor do they play any role in testing or applying the theories. It is therefore pointless to criticize such theories on the ground that they try to quantify our ordinary concepts of belief and desire; by the same token it is meaningless to say of these theories that they "include negative values". And in what sense can a Bayesian ascription of a propositional attitude "conflict" with a selfascription? If this means that agents are not good at judging their own subjective probabilities in terms of numbers, this is true, but irrelevant, since neither Jeffrey's nor Ramsey's theory says anything about numbers. The theories we are discussing consist of sets of axioms which specify relations among simple, unquantified, preferences. The numbers come in only as, so to speak, a comment on the axioms: they are useful in proving that the axioms have certain properties, and in simplifying calculations. The numbers are not supposed to have any direct psychological significance. Lorenz-Meyer calls the fact that Jeffrey's system makes the probability and desirability functions less determinate than they are, say, in Ramsey's system a "drawback" and a "disadvantage"; however, he does not give any reason for this judgement, and I can think of none. He also raises the old question whether Bayesian theories are normative or descriptive, but it is unclear whether he appreciates the power of the view that any correct description of an agent's propositional attitudes and actions will have an important normative aspect. Is it a failing of the theory I propose that it is dubiously practicable? Well first, is its application so difficult? It depends on what you mean.

264

Donald Davidson

The complicated sequence o f steps I outline for determining beliefs, desires and meanings is certainly impractical (so for that matter is the series o f steps Ramsey outlines to persuade us of the adequacy o f his axioms). I have made clear on a number of occasions, most recently in my Dewey Lectures, that I do not imagine for one moment that anyone could in practice carry out all the steps in this sequence. The point of the story about how it could (very much "in theory") be done is rather to provide an informal proof that the postulated structure has a certain interesting property: the observable behavior o f an agent whose behavior exhibits that structure fixes that agent's beliefs, desires and meanings. Knowing this, are there ways other than by following the "route o f the theorem" that an observer can use the available evidence to learn things about an agent? O f course there are, as one can see from the many ways o f using the structure postulated by Ramsey. Lorenz-Meyer points out that if all you know is that an agent prefers a certain sentence true rather than its negation, Jeffrey's theory doesn't allow you to draw many conclusions. This is true but misses the point I was making: I wanted an example to show that the kind o f data on which the application o f the theory rests is not remote from easily observed behavior, in this case the use of a sentence to make a request. I also suggested ways in which one can determine more generally that an agent prefers the truth o f one sentence to that o f another. Data o f this sort is all that the application o f the theory requires. It is obviously grotesque to suppose that in our everyday encounters we interpret the words and thoughts o f others by working through the Byzantine sequence I have outlined. As I said, my aim was not to do armchair psychology. I was interested in the theoretical question what sort o f structure o f beliefs, desires, and meanings would explain the possibility o f one person coming to understand another. My attempt to describe such a structure is certainly too crude to be right. Nevertheless, I was able, I think, to show that the structure I described is rich enough to demonstrate the theoretical possibility of working one's way into it from the outside. I f there is one way to do this, there will be other ways. In practice we justifiably assume we know, concerning a person we want to understand, much o f what the Byzantine sequence is designed to prove we could learn without the assumptions. It would be a mistake to think this observation destroys the interest of my enterprise; for what, in the beginning, justifies the assumptions we all make about the thoughts and motives of others? It would be wonderful if we could specify in some simpler way the norms that constrain how our attitudes can be described. But any serious attempt to modify Tarskian and Bayesian theories in the direction o f greater realism will surely result in far more complicated, not simpler, theories. Nevertheless, this is the direction in which I believe fruitful and valuable progress will be made.

Reasons, Actions, and their Relationship R A L F STOECKER

The following paper is devoted to the discussion of three important and closely interlocked topics in the philosophy of Donald Davidson, the questions: What are reasons? — What are actions? — And: What is the relation between a reason and an action, when the reason explains the action by giving the agent's reason for doing what he did? The last question is actually a quotation; it is the first sentence of Davidson's famous article Actions, Reasons, and Causes. Although subsequently modified in various important respects, it still provides the metaphysical groundwork for Davidson's later views on action theory and the philosophy of mind. It is this metaphysical basis that I want to concentrate on. Offhand, Davidson's answers to all the three questions are common in philosophical action theory: Actions are events, reasons are causes of those events, and giving the agent's reason explains an action because it explains the action causally. In what follows, I shall first recapitulate Davidson's reasoning toward the third claim, which I take to be true and to be what Davidson thinks. In discussing a possible objection against this claim I shall then try to show that the second thesis, the thesis that reasons are causes, is neither true nor is it Davidson's opinion. And finally, I want to argue with regard to the first thesis that although Davidson actually holds that actions are events, he should better abandon this position.

I. What is the explanatory relation between what I do and my reasons for doing it? — The question already presupposes that there is an explanatory relationship. In Actions, Reasons, and Causes Davidson calls the explanations in question rationalisations (p. 3). He distinguishes three kinds of rationalizations: 1. Explanations by pointing out a primary reason for the action. 2. Explanations by pointing out a non-primary reason for the action. 3. Explanations by a redescription of the action.

266

Ralf Stoecker

According to Davidson, explanations of the second and third kind depend on explanations of the first kind, in the sense that they obtain their explanatory strength from explanations based on primary reasons. 1 I shall, therefore, start with this first type of explanation. "Primary reason" is Davidson's term of art. A primary reason for a particular action A is a pair of attitudes, consisting of a pro attitude toward actions of a certain kind Κ and a belief about A that it is of the required kind K. — To take his well-known example: Davidson entered a room, he wanted to turn on the light and believed that his flipping the switch would be a turning on of the light. This pair of attitudes was his primary reason for his act of flipping the switch — or at least: one of his primary reasons, since he must have had a variety of primary reasons: He wanted to illuminate the room and believed that his flipping the switch would be an illuminating of the room; he wanted to prevent himself from running into a piece of furniture and believed that his flipping the switch would prevent him from running into something, etc. It is obvious that primary reasons play an important part in everyday action explanations. Davidson flipped the switch because he wanted to turn on the light and he regarded the flipping of the switch as a way of doing this. But in what sense do the primary reasons explain an action; why are they so valuable as explanations of what someone does? — Part of Davidson's answer corresponds to what was the standard account in philosophical action theory at the time he wrote Actions, Reasons, and Causes. He says: "In the light of a primary reason, an action is revealed as coherent with certain traits [...] of the agent, and the agent is shown in his role of Rational Animal." (ibid. p. 8). It is always possible to transform the components of a primary reason into the premises of a practical syllogism leading to the conclusion that the particular action was in certain respects attractive. In this very meager sense, what the agent does is reasonable — from his or her own perspective it is seen as justified. Insofar the agent is a Rational Animal. It is a difficult question, exactly how the relation between primary reasons and actions on one side and the premises and conclusion of a practical syllogism on the other should be construed. Davidson himself changed his mind from Actions, Reasons, and Causes to How Is Weakness of the Will Possible? and again in Intending. But I do not want to pursue this, instead I shall concentrate on the other part of Davidson's account of the explanatory value of primary reasons which was almost heretical in 1963 and became the new orthodoxy immediately after. Why should one not be content with the answer already given? — Davidson states two reasons: First, although there are several primary 1

Cf. Davidson's thesis 1 on page 4 of Actions,

Reasons, and

Causes.

Reasons, Actions, and their Relationship

267

reasons for each particular action, not any pair of attitudes that satisfies the above-mentioned condition with regard to a particular action is a primary reason for the action. Take again Davidson's example of flipping the switch. Imagine Davidson entering the room and taking off his jacket. Accidentally he thereby flips the switch. It may very well be that he wants to turn on the light, and he certainly knows that flipping the switch would be a turning on of the light — still, those attitudes do not explain his flipping the switch. And therefore they do not constitute a primary reason for what he does. Something more is required. About his second reason not to be content with the account of rationalizing explanations given above, Davidson writes: "[...] it is an error to think that, because placing the action in a large pattern explains it, therefore we now understand the sort of explanation involved." (Actions, Reasons, and Causes, p. 10). It is not obvious why fitting an event into a pattern should, in itself, give us any clue as to why the event had to happen. So, more has to be said about the interrelation between primary reasons and actions to account for their explanatory power. And this is the well-known second half of Davidson's proposal: Rationalizations by way of primary reasons have to be construed as causal explanations. This second part of Davidson's proposal has the advantage of compensating the two short-comings of the first part. The first is met, because one would not expect Davidson's desire to turn on the light and his well entrenched knowledge about the functioning of the switch to be causally involved in his act of inadvertently flipping the switch by taking off his jacket. Hence, the desire and belief do not constitute a primary reason. 2 And the second short-coming is met at least insofar as being a causal explanation is regarded as a sufficient account of the explanatory power of an explanation. The fact that Davidson's proposal gets around these objections is strong evidence for its truth. Still, there are a number of other, new objections to his proposal, five of which he discusses at length in Actions, Reasons, and Causes. The most prominent ones are the second and third one. The 2

This reaction may seem to be unsatisfactory, since it is easy to show that a causal relationship to the action still does not provide a sufficient condition for being a primary reason. In examples of so-called deviant or wayward causal chains pairs of attitudes fulfill the syllogistic requirement as well as the requirement of a causal relationship; none the less they do not give a rationalizing explanation of what is done. (Cf. Davidson's discussion in Freedom to Act, pp. 79 — 80, and Problems in the Explanation of Action, p. 39.) But the hint at the phenomenon of wayward causal chains cuts both ways: It is quite obvious that in these cases the propositional attitudes do not explain the occurrence of the action because they do not cause it in the right way (whatever that may be) — and this in turn confirms the importance of causality considerations with regard to action explanations.

268

Ralf Stoecker

second objection is the so called logical connection argument, which states that the syllogistic relationship between primary reasons and actions contradicts a causal relationship. According to Davidson this argument is due to a logical mistake: the failure to distinguish relations between sentences (or parts of sentences) from relations between events. An inferential connection between two declarative sentences should not, in itself, imply anything regarding the relationship between the entities those sentences talk about; at best it could be a challenge for the idea of an explanatory relationship between those sentences. (If for example you asked someone why her car exploded, you would not expect to be told that it happened because a cause of the explosion occurred. Though it is certainly true that the car exploded because a cause of the explosion took place, it is perfectly pointless and uninformative.) But obviously rationalizations have their point. You cannot infer from the fact that Davidson flipped the switch to the conclusion that he had the desire to turn on the light; he might as well have intended to demonstrate that the fuse was blown or might have wanted to switch on the heater. Hence, whatever logical connection there is between reason sentences and action sentences, it does not preclude rationalizations from being causal explanations. The logical connection argument has not been very popular during the last twenty years. Much more attention was devoted to arguments along the lines of the third objection Davidson considers in Actions, Reasons, and Causes. It states that causal explanations are based on laws, rationalizations are not; hence rationalizations cannot be causal explanations. Davidson repeatedly repudiated this objection. Therefore, I want to bracket it and instead remind you of the first objection against the thesis that rationalizations are a species of causal explanations, which to my mind has not provoked the attention it deserves. Davidson formulates it as follows: The first line of attack is this. Primary reasons consist of attitudes and beliefs, which are states or dispositions, not events; therefore they cannot be causes. {Actions, Reasons, and Causes, p. 12)

The objection (which in the following I shall call Objection One) reads: If rationalizations of the first type listed above, i. e. rationalizations by primary reasons, were causal explanations, primary reasons would have to be causes of the actions to be explained; and since causes are events, primary reasons would have to be events; but they are not events — consequently they could not be causes, rationalizations by primary reasons could not be causal explanations.

269

Reasons, Actions, and their Relationship

II. What is Davidson's reaction to Objection One? — There are two possible reactions that suggest themselves, corresponding to the two assumptions on which Objection One is based: the first assumption that since primary reasons are states they cannot be events, and the second one that causes are always events. Both assumptions can be challenged, but interestingly enough Davidson resists taking one of those ways out. Before I turn to Davidson's own treatment of Objection One, I shall say something about Davidson's reasons why he accepts those premises. Perhaps the most obvious way to meet Objection One would be to attack the first assumption, and to maintain that, although primary reasons are states, they are, none the less, events. This reaction can be based on different considerations. It can result from a particular understanding of events (like Jaegwon Kim's view that events are property exemplifications). It can also be the consequence of a more general scepticism with regard to the common sense categorization that distinguishes states from events. And it can, finally, find its motivation inside Davidson's philosophy, in the idea that the author of Mental Events and proponent of the psychophysical token-identity thesis would have no choice but to subsume primary reasons under the category of events. None the less, Davidson resists taking this line against Objection One. In Actions, Reasons, and Causes he accepts the first premise and agrees that primary reasons are not events.3 Davidson is not very explicit about why he does so; but I think the main reason is to be found in his view on what events are. What are events according to Davidson? — In 1969 he published The Individuation of Events, which put forward a new proposal for identity conditions for events, saying that events are identical if they share all their causes and e f f e c t s (cf. p. 179). In the discussion that ensued from Davidson's paper, most critics accepted the proposed condition as being true, but only a very few regarded it as also informative. The majority was of the opinion that conditions of individuation should throw more light on what events are than Davidson's proposal did. And the question now under consideration, as to whether states in general or reasons in particular are events, gives a good illustration of what they had in mind. Both claims — that states are and that they aren't events — are compatible with the causal identity condition for events. Insofar Davidson was free to hold that reasons were events and hence causes; but at the same time it was not very clear at all what events were according to him. 3

Cf. also Davidson's Reply

to Bruce

Vermaßen,

p. 221.

270

Ralf Stoecker

This became much clearer when, in 1985, Davidson wrote his Reply to Quine on Events, in which he withdrew the causal event individuation, and replaced it with Quine's proposal that events are identical if they share their spatiotemporal locations. Davidson's new opinion about the individuation of events is far less liberal than its predecessor, it is not compatible with the view that states are events. In order to be events in the sense of Quine's proposal, states would have to satisfy two conditions: First, they would have to occupy a position in space and time, and, secondly, this location would have to be peculiar to the particular state. The first condition can be met; states normally have temporal boundaries, and since they are always states of someone or something they can be spatially localized at the person's or object's place. But this way of providing states with a spatiotemporal location renders hopeless any prospects of meeting the second condition. Everything and everybody is in a huge variety of states that begin and end simultaneously without being identical. Take me as an example: From the time I woke up this morning until now I have been tired and I have been worrying about a friend of mine. Both states coincide spatiotemporally, but they are doubtlessly different. The same is true for a lake that was poisoned by some evil polluter exactly when it started to freeze; its state of being frozen occupies the same space-time area as its state of being poisoned, still they are certainly not the same state. I take these examples to be sufficient to demonstrate that states in general are not individuated spatiotemporally and therefore, as long as one sticks to Quine's proposal for the individuation of events, should not be counted as events. One could, in principle, still maintain that, contrary to other states, primary reasons are individuated spatiotemporally, but I doubt that this is an attractive way out. If events are spatiotemporally individuated space-time worms, reasons are not events. And the fact that as early as 1963 Davidson refused to regard primary reasons as events shows that by then he already had such a coarse-grained event conception. Before I turn to the second premise of Objection One let me say something about the relation between the first premise and Davidson's token-identity thesis. If primary reasons are not events, then, certainly, they are neither physical events, i. e. primary reasons are not (token-) identical with physical events. It seems as if this is just the opposite to what Davidson said in his famous paper on Mental Events and would therefore constitute a strong reason to deny the first premise. But a closer look at Mental Events shows: There is at least no overt conflict with the token-identity thesis. Davidson is very careful in Mental Events not to apply his thesis to primary reasons. He gives quite a number of examples for mental events (perceiving, remembering, deciding, noting, calculating,

Reasons, Actions, and their Relationship

271

judging 4 ), yet nowhere in Mental Events does he explicitly subsume beliefs, desires, pro-attitudes, hopes, preferences, holding-trues etc. under the class of mental events. The paradigmatic mental phenomena of Davidson's action theory and philosophy of language simply do not appear as instances of his token-identity thesis. To my mind this is more than an accident, it is evidence that he never regarded them to be events at all. 5 I hope to have shown that Davidson does not and, in the light of his latest writings on events, should not regard primary reasons as events. He accepts the first premise of Objection One. What about the second premise, according to which only events are causes? Prima facie, this premise is not much stronger than the first one. As Davidson immediately points out in Actions, Reasons, and Causes, it is usual to say that a bridge collapsed because of a structural defect or a plane crashed on takeoff because the air temperature was abnormally high. 6 So, why should one not concede that other entities instead of events could be causes as well? Davidson's writings do not answer this question. But I think that, at least, there is a consideration that shows that the philosophical price one would have to pay for giving up the second premise is higher than it may seem at first glance. For states to be causes, first of all, they have to be existing. One has to regard them as part of the world's entities. And this in turn has consequences that might not be desirable in the case of states. Offhand, it may seem strange to doubt the existence of states. I enjoy a particular state of being relaxed and sleepy, hence there is a state I enjoy. We simply claim very often that there are states. In this respect states are not different from material objects or events. But neither are they different from properties, voices, habits, fates, and many other more or less dubious creatures of the linguistic means that Quine called "reification". Any serious ontology has to select among this variety, between what are merely ways of talking and what is actually there, it cannot take language at face value. What are the criteria for this choice? — To my knowledge Davidson nowhere explicitly says whether he thinks that states really exist or not, but it is obvious that he takes ontological questions like this one seriously. 4 5

6

Cf. Mental Events, pp. 2 0 7 - 2 0 8 . There is one remark in Mental Events that makes obvious that Davidson is well aware that propositional attitudes like beliefs and desires are not mental events: "Thus intentional actions are clearly included in the realm of the mental along with thoughts, hopes, and regrets (or the events tied to these)" (p. 211, my emphasis). Another evidence for my supposition that Davidson intentionally did not apply the token-identity thesis to propositional attitudes is to be found in his argument for the anomalous character of the mentalistic vocabulary (ibid. ch. II). As soon as he shifts from talking about entities to talking about a vocabulary, he uses propositional attitudes as examples. Cf. Actions, Reasons, and Causes, p. 12.

272

Ralf Stoecker

In a number of articles he argues very carefully for the existence of events, and he certainly would not have done this if simply our everyday talk about 'things that happened' had shown that there are events. Now the same would have to be accomplished in support of the claim that there are states. In his discussion of events Davidson gives two criteria that lead him to the conclusion that there are events: First, the logical form of certain obviously true sentences reveals that they claim the existence of particular events; and secondly, it is possible to find identity conditions for events. If there were states as well as events they, too, would have to satisfy those conditions. I have no argument to offer that shows that in the case of states it is impossible to meet these criteria, but at least with regard to the first one I am sceptical. And I am even more sceptical that Davidson would accept that states satisfy this criterion. Let me explain why I think so. In the case of events we have to assume that there really are events because, according to Davidson, otherwise we could not account for the truth of a multitude of basic sentences like "Doris sings" or "Vesuvius erupted". Davidson's philosophy of language is based on the presumption that for any speaker it is in principle possible to construct a truth theory that gives the truth conditions for all the sentences of his or her language. To achieve this with respect to sentences like "Vesuvius erupted" and "Vesuvius erupted in 1906" and "Vesuvius erupted in May 1906 burying a fortune of wineyards", those sentences have to be taken as not only saying something about a mountain (and the wineyards), but also about an event. More precisely: Their logical form has to include an existential quantification over events. And existential quantifications over events can only be true, according to Davidson's realistic reading of quantifications, if there really are events. So, either one is willing to accept the existence of events or one has to deny the truth of sentences like "Vesuvius erupted". And Davidson takes it for granted that only the first alternative is tenable. 7 The fact that Davidson emphasizes the hidden quantificational structure of certain sentences is a strong indication that to his mind this is not a common feature of all sentences. I do not know an example of Davidson's that proves my interpretation, since he explicitly states the logical form of a sentence only when there is a problem conjoined with it. Perhaps the nearest candidate is "Socrates is wise" in The Method of Truth in Metaphysics (p. 206 — 210), which I take to have the logical form according to Davidson that there is someone named Socrates who is wise. This sentence has no other ontological consequence than the existence of a certain person. In 7

For Davidson's argumentation cf. The Logical form of Action Sentences pp. 118 — 120, The Individuation of Events p. 166 — 167, The Method of Truth in Metaphysics, pp. 210—214.

273

Reasons, Actions, and their Relationship

particular, it has no consequence with regard to the existence of properties, classes, or — states. If being wise is a state, a state in which Socrates is, this is still not expressed in "Socrates is wise"; and this would be a strange consequence. One should expect this sentence to be the most natural way to say that Socrates is in this state, if such a state exists at all. Hence, the logical form of this sentence is strong evidence against the existence of such a state (despite the normality of saying, for instance, that Socrates reached the state of wisdom very early in his life). This leaves open the question of whether other supposed state sentences display a hidden quantification or not. Wisdom may be a weak candidate for a state, but what about being tired or frozen? In order to decide whether the sentences "Doris was tired" or "The lake was frozen" assume the existence of more than Doris and a lake, one would have to consider the possibility of developing a truth theory for them that accounts for their inferential relations to other sentences (like "Doris was tired for three months"). I do not know, whether or not this will, in the end, lead us to the conclusion that those sentences quantify over states. But it seems to me that, should those sentences turn out to hide such a quantification, there will hardly be any sentence left that does not. And this would, in turn, shed a strange light on Davidson's entire semantical project. Therefore, I doubt that Davidson takes those sentences to quantify over states. And if they do not, states are, in a crucial respect, ontologically worse off than events — their most important linguistic witnesses disavow their existence. To my mind this is a strong reason for Davidson not to counter Objection One by giving up the second premise, i. e. taking states to be a second type of causes besides events. And likewise it speaks against meddling with the first premise that states are not events — since there are no states, they should hardly be regarded as events. But if both premises of Objections One are true then the conclusion is inevitable: Primary reasons are not causes. And this straightforwardly contradicts one of the two theses to be defended in Actions, Reasons, and Causes, namely the thesis that "The primary reason for an action is its cause." 8 — My only idea about how to resolve the contradiction is that the quoted thesis is misleadingly shorthand. What Davidson wanted to say, as I see it, was not that primary reasons are causes, but that explanations by primary reasons are causal explanations. Seen from this perspective the antagonism disappears, it is replaced by the claim that, according to Davidson, there are causal explanations that do not mention causes, and that rationalizations of the first kind belong to these explanations. And this is no longer logically incompatible with the two premisses. 8

Ibid. p. 4. Similar passages can be found in Psychology in the Explanation of Action, p. 41.

as Philosophy,

p. 233, and in

Problems

274

Ralf Stoecker

III. The proposed reading of Davidson's harmonizes with the two premisses of Objection One, but in itself it seems to be patently implausible. How could a rationalization that is based on primary reasons — i. e., a rationalization of the first kind mentioned above — be a causal explanation if primary reasons were no causes?! The answer is to be found in Davidson's account of the explanatory value of other rationalizing explanations — explanations of kind two that rely on non-primary reasons, and explanations of the third kind that redescribe the action. There he makes clear why causal explanations need not mention causes. Those explanations need not mention causes because, according to Davidson, they gain their explanatory value by indicating other explanations (namely, the rationalizations of the first kind). What does that mean? As has been said, rationalizations of the first kind are characterized by pairs of attitudes (the pro-attitudes and the corresponding beliefs). But when we actually give such an explanation, we rarely mention both of them. I can explain why Davidson flipped the switch by saying: "He wanted to turn on the light", and I add little to the explanatory force if I go on: "And he thought that flipping the switch would be a turning on of the light". Every hearer already knows without being told how to complete the pair of primary reason. The explanation rendered by just mentioning the pro-attitude was elliptical. But it is not always the case that a hearer exactly knows or should know how to supplement the missing attitude. If I explain Davidson's act by saying: "He thought that flipping the switch would be a turning on of the light", I leave it open as to whether he wanted to turn on the light, felt obliged to turn on the light, was crazy about turning on the light, or whatever pro-attitude he might have had toward turning on the light. 9 Still, this is a correct rationalizing explanation of the first kind. So, there are rationalizations of the first kind that do not provide the hearer with a complete knowledge of an action's primary reason. What they do is this: they strongly restrict the range of pro-attitudes that come into question as part of the primary reason. This fact alone is, according to Davidson, sufficient as an account of the correctness of a rationalizing explanation. This leads to the question of the explanatory value of the rationalizations of the other two kinds. In those cases the explanations do not mention a 9

Davidson notes that one may take wanting to be a genus for all species of pro-attitudes (.Actions, Reasons, and Causes, p. 6), but this does not imply that the pro-attitude is a wanting that goes along with a craving, feeling obliged or the like. What it means is that to say that someone wants something leaves open what kind of attitude she or he actually has.

Reasons, Actions, and their Relationship

275

primary reason at all. I may for example explain Davidson's behavior by saying that he wanted a nice and cosy apartment (which is an explanation of the second kind). Neither a pro-attitude toward a kind of action nor a belief about a particular action is mentioned. None the less it is a perfectly good explanation; as with the incomplete explanations of the first type, its strength relies on the fact that the range of possible primary reasons is drastically reduced. We still do not know what exactly Davidson saw in his flipping the switch — it is possible that he wanted to turn on the heater and thought that this would be accomplished by flipping the switch —, but we know that a lot of possible reasons were not effective — he did not do it because he wanted to hear the news and regarded the switch to be connected with his radio-set or things like that. The same goes for explanations of the third kind, e.g.: "He did it because it made the apartment nice and cosy." This explanation, like the one before, gains its value not from actually mentioning primary reasons, but from the restriction of the range of primary reasons some of which were effective for the action in question. (It says a little bit less than the other one since it does not specify the kind of pro-attitude Davidson has with regard to a nice and cosy apartment, and it says a little bit more because it implies that the action was successful.) The importance of this Davidsonian account for the explanatory value of certain types of rationalizations lies in the fact that it requires no assumption about any things the explanatory sentences talk about. There is a standard objection to Davidson's action theory and philosophy of mind, namely that mentioning causes is not sufficient for a sentence to provide a causal explanation. In the light of his account of rationalizations of the second and third kind it becomes obvious that neither is mentioning a cause a necessary condition for causal explanations. Here, to my mind, is the key to Davidson's answer to Objection One. If reason explanations can be causal without reasons being causes (like in the example of Davidson's wanting a nice and cosy apartment), explanations by primary reasons (i. e. explanations of the first kind), too, could be causal without primary reasons being causes. What those explanations do is, again, narrow the class of events that encloses the cause. Knowing the primary reason gives one a much better idea of what a cause for the explanandum action looked like, but the cause itself does not belong to the primary reason. I take this to be what Davidson had in mind when in response to Objection One he said: Mention of a causal condition for an event gives a cause only on the assumption that there was also a preceding event. [...] In many cases it is not very difficult at all to find events very closely associated with the primary reason. States and dispositions are not events but the onslaught of a state or disposition is. [...] During any continuing activity, like driving, or elaborate performance, like swimming the Hellespont, there

276

Ralf Stoecker are more or less fixed purposes, standards, desires, and habits that give direction and form to the entire enterprise, and there is the continuing input of information about what we are doing, about changes in the environment, in terms of which we regulate and adjust our actions. {Actions, Reasons, and Causes, pp. 12 — 13).

These inputs of information — perceptions, feelings, and last but not least conscious deliberations — are events that in Davidson's words are "very closely associated with the primary reason" and therefore underlie the explanatory force of reason explanations. When we learn that Davidson flipped the switch because he wanted to turn on the light and believed that flipping the switch would be a turning on of the light, we get a good idea of the processes that could have been involved in the causal history of the switch flipping and those that could not. Davidson might have seen the switch at a certain distance, and this caused him to move his hand in that direction until he reached the switch, or he might have groped about in the darkness until suddenly feeling that he touched the device, and this caused him to move his hand in a certain way, or the like; but he did not stumble over a chair and fall against the wall and thereby flip the switch, neither did he feel an uncomfortable warmth causing him to take off his jacket which in turn caused his hand to flip the switch. So, to my mind, this is Davidson's answer to the question of what reasons are and how they are related to the actions they rationalize: Reasons are not causes, but giving a reason explanation, a rationalization, gives an idea of what the causes look like; therefore, rationalizations are causal explanations. An additional argument for the claim that rationalizations explain without reasons being causes can eventually be found in what I take to be the most attractive answer to the third and yet unresolved question of this paper, the question of what actions are. IV. It may be a matter of debate, or perhaps even better, a question to be addressed to Professor Davidson personally, whether he thinks that reasons are mental events or not, but it is beyond any doubt that he regards actions to be events. According to Davidson actions are (or at least belong to) those events that can be causally explained by primary reasons. He has defended this view in a number of articles, one of which, Adverbs of Action, even started with the words "Actions are events ..." (p. 230). Now, are actions really events? — In what follows I want to argue that there are two classes of actions that cannot be convincingly understood as events: First, actions that consist in the bringing about of something (which I shall call "productive actions")·, and secondly actions that consist in

Reasons, Actions, and their Relationship

277

the letting happen of something (which I shall call "admissive actions"'). And since there are so many actions that belong to one of those classes, one should not regard actions as events at all. But what then are actions? — Finally I shall try to sketch an answer which will have the advantage of still being in accord with almost all of Davidson's philosophical theory. The standard examples of productive actions are killings, although there are other more or less violent representatives to be found in literature: destroyings, thankings, divorcings, insultings, impregnatings etc. I shall concentrate on one of Davidson's own examples: 10 Suppose Arthur places a time bomb in a suitcase, and puts the suitcase on an aeroplane. The aeroplane is subsequently destroyed by the explosion, (p. 236)

What is obvious is that Arthur blew up the plane. And it may be taken for granted that he did it intentionally, so it was his action. Now, if actions are events, Arthur's destroying the plane had to be an event. And since events occupy a particular spatiotemporal location, this event has to have boundaries in space and time. The question is: When did Arthur's action take place? I want to show that there is no satisfactory answer to this question. (It is presumably equally difficult to tell where the action took place.) There is no special problem conjoined with the beginning of the action, for the sake of the argument it may be chosen arbitrarily. Suppose the action began when Arthur entered the airport. But when was the action finished? — Davidson's own answer is, in a sense, minimalistic: "Since Arthur did nothing more than place the bomb, it is natural to say [...]: his placing the bomb was his destroying the aeroplane" (ibid.). And because the placing of the bomb was over when Arthur left the airline counter his destroying the plane was over, too. This is an instance of Davidson's thesis that all actions are what Arthur Danto called basic actions. 11 But as Davidson is well aware, with regard to examples like the one under consideration this claim is hard to swallow. It entails that right after Arthur left the airport again, and even before the plane took off and long before it broke into pieces, Arthur had already destroyed it. Or to put it more figuratively: According to Davidson, there was an airplane moving toward the runway, which was destroyed an hour ago. — It is obvious that Davidson has to give very good arguments in support of this position. To my knowledge he proposed two lines of argument, one of which is discussed more frequently in literature — perhaps because it is easier to

10

11

Davidson discusses what I call productive actions in Agency, pp. 57 — 59, The Individuation of Events, pp. 177 — 178, Problems in the Explanation of Action, p. 38, and Adverbs of Action, pp. 236 — 239, from which the example is taken. Cf. Davidson's discussion in Agency.

278

Ralf Stoecker

counter —; and a second one that, I think, is much more interesting and revealing. The first argument traces back our uneasiness about Davidson's claim to an epistemic deficit. In the time span from the check of Arthur's suitcase to the explosion of the plane, nobody could know, whether Arthur would be successful or not. Hence, nobody could know whether what certainly was a placing of the bomb would also turn out to be a destroying of the plane. Therefore nobody would have been expected to say or believe, during this interval, that Arthur had destroyed the plane (or: that the plane had already been destroyed by Arthur). Since one could not know the truth of the claim, one is rather reluctant to accept it. Davidson supports this argument with an analogy: 12 Quite a long time ago Davidson's great-great-grandfather, Clarence Herbert Davidson of Inverness, had been born. Nobody in those days had the slightest idea that it was true that Donald Davidson's great-great-grandfather was born, they only knew that he was a child named Clarence Herbert. So, they would have been sceptical if confronted with such a claim. Still, it was true that Donald Davidson's forefather had been born. And, according to Davidson, the same goes for Arthur's destroying the plane. Although nobody, and in particular none of the poor victims, could know it, their plane had already been destroyed and they were killed. I do not know how much weight Davidson would want to attach to the grandfather analogy. To me it seems to indicate almost the opposite of what he intended it for. Though more than a century ago nobody knew that Donald Davidson's forefather had been born and consequently nobody claimed this to be true, now that we know Donald Davidson and his family tree, we can assert various truths about his ancestor; e.g. that once upon a time in the nineteenth century (I presume) Donald Davidson's great-great grandfather met a woman he liked. We easily accept this sentence as true, despite the fact that nobody could know this until 1917. In contrast we do not, even in the light of our retrospective information about the tragic development, accept it as true that the plane took off a few hours after it was destroyed by Arthur. In this crucial respect we make a difference between grandfathers and productive actions. As I said before, Davidson has another, stronger defense for his minimalistic construal of the temporal boundaries of productive actions. But let me postpone the discussion of this for a while and first ask, what alternative answers are there for the temporal localization of Arthur's action? Davidson's answer is so implausible, because it implies that the airplane is destroyed in advance of its destruction. Therefore, an alternative highly recommends itself, according to which Arthur did not destroy the 12

Cf. Problems

in the Explanation

of Action, p. 38.

Reasons, Actions, and their Relationship

279

plane until it exploded — the action ended with the destruction of the plane. As far as I can see, this second, more baroque answer has been the most popular one in literature. It has the advantage of circumventing the temporal problems of Davidson's position. But in return it has other perplexing consequences: Imagine Arthur leaving the airport, driving to a pub and taking a drink. An adherent of the second answer has to concede that Arthur, while sitting on the barstool and sipping his drink, is destroying an airplane. This in itself is a strange description, but it becomes even more peculiar in a case where Arthur does not know whether the bomb has already exploded or not (because it is connected to an altimeter not to a timer), and therefore has to be regarded as someone who does not know if he is still destroying a plane or has already finished his action. And it would be totally unexceptional in a case where Arthur sleeps while the plane is still in the air, or in a case where he is killed by his distrusting friends. Dead men don't destroy aeroplanes. 13 A second difficulty for this baroque account was brought to my attention by Wolfgang Künne. Obviously, when Arthur placed the bomb he tried to destroy the airplane; placing the bomb was his attempt to destroy the plane. And the attempt was successful, the plane burst into pieces. Yet, successfully attempting to destroy something presumably is identical with destroying it, and since Arthur's placing the bomb was his attempting to destroy the plane and was successful it has to be identified with his destroying the plane — which, in turn, supports Davidson's view and contradicts its baroque alternative. Obviously we are faced with a serious dilemma: Whatever time one chooses as the end point of Arthur's action, it forces us into saying strange things: Neither would one like to say that the action was finished before the plane exploded, nor would one like to say that the action continued after Arthur made his contribution to the intended outcome. Hence, for every action with a time interval between the agent's contribution and the occurrence of the intended result there is a period in which we are reluctant to say that the action was over as well as saying that it was still going on. But if actions were events we would have to make a choice and take one of these alternatives. This is the dilemma of the temporal localisation of productive actions. It has to be solved if we are to maintain the claim that actions are events. As mentioned above there is a second group of actions that resist a subsumption under the class of events, actions that are frequently grouped 13

O r take another example Davidson gives in The Individuation of Events (p. 177), where a murderer poisons the water tank of a spaceship, so that the victim dies after reaching his destination on Mars. If the second answer were correct, this killing would have continued for years.

280

Ralf Stoecker

under the heading of "letting happen" and that I have called admissive actions. Examples are a man who intentionally grows a beard, a farmer who lays fallow her land, etc. I want to represent them by a distressing, less fictitious example taken from applied ethics: Bertram is a severely handicapped child, a few days old. The pediatrician is absolutely certain that Bertram will never lead a human life, whatever that may mean. Then, Bertram gets an infection that by itself would be easy to cure. But the physician, in accordance with the parents and the medical staff, refrains from giving antibiotics to the child; a few days later Bertram dies. The doctor let Bertram die. She did it intentionally, it was her action. If this action were an event, it should have a place in space and time. Where and when did the action occur? — One may say that the action was finished when the doctor decided not to give antibiotics to her patient, then one would face problems similar to the ones in the airplane example. However, I think the new example is different from Arthur's actively destroying the plane and killing the passengers. In the case of Bertram's death it is reasonable to claim that for the whole period during which Bertram was dying the doctor let Bertram die. Bertram's doctor may have sat at the same time in the same pub as Arthur did and half an hour later Bertram died just the same moment the airplane exploded — still, Arthur, while sitting in the bar, did not destroy the plane, whereas the doctor did let Bertram die. So, in this temporal respect there is no dilemma, the action was not finished before the child died. But where did the doctor do it? — There she is sitting on a bar stool, raising her arm to take a puff at her cigarette. Is this moving her arm part of her letting Bertram die? What about her nervous drumming with the fingertips, or her thinking about her stolen car, do these actions belong to her letting Bertram die? — These are the sort of things that actually happened with the doctor while Bertram died. But I doubt that one should count those events as a part of her act of letting him die. And if they are not, I don't know any other events that could constitute this action. The problems raised by questions of the spatiotemporal localization of productive actions like destroying a plane or admissive actions like letting someone die, are obviously not due to a simple epistemic deficit. In some respect the questions must have been ill-advised from the beginning. In what respect? The radical answer that I eventually want to defend is that the spatiotemporal localization fails because actions are not events. But there is a more moderate answer that is compatible with the view that actions are events. Davidson gave it as his second defense of the assumption that Arthur's destroying the plane is identical with his placing the bomb and therefore ended long before the plane exploded. According to Davidson the difficulties with the question, of when Arthur destroyed the airplane, are due to a wrong construal of the logical form

Reasons, Actions, and their Relationship

281

o f the involved action sentences. Take a paradigm action sentence like "Sebastian ran h o m e " . As any other sentence it has the logical form o f an existential quantification: I f it is true, then there was at least one person named Sebastian and at least one action that was a running home by him. So, when did Sebastian run home? It is easy t o see that there may be more than one correct answer; he mave have run h o m e on Monday and have done it again on Tuesday. Turning a true action sentence into a " w h e n " question does not guarantee that there is exactly one correct answer. B u t there may also be n o correct answer at all. " J a c k and J i l l went up the hill." W h e n did they do that? T h e question has n o answer, i f J a c k went on Monday, Jill on Tuesday, because then there is n o time when they did it, there is only a time J a c k did it and another time Jill did it. 1 4 T h e sentence has t w o quantifiers over actions, one asserting the existence o f an action by J a c k , the other one asserting the existence o f an action by Jill. 1 5 Simply putting " w h e n " in front o f the sentence leaves open, which temporal location is asked for. N o w , the same is true, according to Davidson, for action sentences about productive actions. [...] 'Arthur destroyed the aeroplane' must have a form something like 'Arthur caused a destruction of the aeroplane.' But agents cause things to happen by doing something — perhaps by placing a bomb. [...] A fuller analysis of 'Arthur destroyed the aeroplane' then takes us to: There exist two events such that Arthur is the agent of the first, the second is the destruction of the aeroplane, and the first caused the second." {Adverbs of Action, p. 237).

I f Davidson is right, and the last sentence really gives the form o f "Arthur destroyed the aeroplane", it becomes clear why it is impossible to tell when A r t h u r destroyed the airplane. As in the J a c k and Jill example the " w h e n " - s e n t e n c e is not explicit as to the event in question. I f the event o f the plane's destruction is meant, the answer is that it happened a few hours after A r t h u r left the airport, but i f the event o f Arthur's doing something that caused the destruction is meant, this was already finished when he left.

14

It is not quite correct that there is no answer to the question about when Jack and Jill went up the hill. They did it in a particular week, month, year, century etc. And the same is true for Arthur's destroying the plane. But the requirement for events to have a spatiotemporal localization is stronger. For it to be satisfied, the "when"-question has to be answerable up to a certain degree of accuracy, a degree that is only determined by the familiar fuzziness of the localization of events. One can ask for the particular day and even hour when a revolution gets started, but not for the minute or second; still, revolutions are doubtlessly events. But the difficulties in localizing the time of Arthur's deed are certainly not due to its having extremely hazy boundaries.

15

As reference for the correctness of this reading of the sentence, cf. The Method of Truth in Metaphysics, p. 209.

282

Ralf Stoecker

Davidson explains why we sometimes go astray, when we simply put a "when" in front of an action sentence. But still Arthur did something that was his action, and if actions are events, Arthur's action must have a temporal location. Hence, in the paper just cited Davidson writes: No ambiguity attaches to the question 'When did Arthur's action of destroying the aeroplane occur?' for the description of this action is 'the action of which Arthur was agent which caused a destruction of the aeroplane', and this has a definite date, (ibid.)

That is how Davidson finally arrives at his original thesis that Arthur's action of destroying the plane was identical with his placing the bomb. But the argument has a weak point. It takes for granted that Arthur's action of destroying the airplane caused the destruction of the plane. And that is not the way we normally use verbs like "destroy". If Arthur's action of destroying the plane would have caused its destruction, then presumably along a causal chain of other causes, such as the loading of the suitcase into the plane's cargo compartment. This entails that we are willing to concede that Arthur's destroying the plane caused the transport of the suitcase. And this is quite implausible. But without this assumption the observation about the logical form of such action sentences does not support Davidson's minimalistic localization of Arthur's deed. A much more plausible interpretation of the logical form with regard to the question of when the action took place, is to be found in The Individuation of Events (p. 177): "To describe a pouring [of some poison] as a killing is to describe it as a causing of a death" — An action of killing someone or destroying something is not, according to the text passage cited, a cause, it is a causing. What I want to suggest now is, that this is exactly how we understand productive actions. Insulting is the causing of an injury, impregnating is the causing of a pregnancy and destroying is the causing of a destruction — a causing, not a cause. And only if causings are like causes events then insultings and destroyings are events, too. However, according to Davidson, they are not: It seems clear that it must be a general mistake to suppose that whenever an event is caused there must be something called a causing. [...] 'caused' relates events, as do the words 'before' and 'after'; it does not introduce an event itself. (.Problems in the Explanation of Action, pp. 36 — 37)

Since productive actions are causings and causings are not events, productive actions are not events; and, moreover, since causings are not entities at all, neither are productive actions — there are no killings, destroyings, insultings etc. It is obviously futile to look for the boundaries of things that do not exist, hence the problems bound up with the spatiotemporal localization of productive actions vanish immediately. But we trade them in for new problems. First of all it sounds strange, to say the least, that there should be no killings, destroyings and insultings anymore. The second problem

Reasons, Actions, and their Relationship

283

lies in the appraisal of those action sentences that seem to ground their truth on the existence of productive actions. There is certainly a difference between intentionally and inadvertently destroying something. But intentionally doing something entails, according to Davidson, that the action can be rationalized, i. e. causally explained. This in turn presupposes that actions are causal relata, that they are events, and thereby bluntly contradicts the thesis stated above that productive actions are not things at all, leave alone events. 16 A third problem for the eliminative attitude toward productive actions is the implausibility of its selectiveness. It seems arbitrary to exclude killings or destroyings from ontology while embracing other actions like arm movements, travelings, and the giving of philosophical talks. And, in particular, it would be odd to deny the existence of destroyings while accepting the act of letting someone die. An ontology comprising admissive but not productive actions is overtly queer. The only consolation I can offer for the first objection is a comparison with other presumed entities that are just as familiar as productive actions and still do not really exist, like properties, voices, moods, and of course reasons. If one accepts the ontological question of what there is as a reasonable question from the beginning, one has to reckon with the possibility that part of our familiar range of entities turn out to be homemade and therefore fictional. What about the other two objections to the view that productive actions do not really exist and that the belief in their existence is grounded on a misunderstanding of the logical form of certain action sentences? — One could try to meet at least part of the third objection by approaching admissive actions the way Davidson suggests for productive ones, i. e., as mistaken construals of the logical form of the particular action sentences. But the attempt to do it exactly the way Davidson proposed fails, and its failure is highly revealing. What is the form of sentences like "The doctor let Bertram die"? The analogy to Davidson's analysis of the Arthur example suggests the following: "There is something the doctor does, and there is Bertram's death, and the former causes the latter." Taking the causing of the death as an entity would then account for the failure to find a localization for the doctor's letting Bertram die. Yet, the proposed form 16

In the discussion of the question of whether reasons are causes I emphasized that causal explananda need not refer to causes. But this offers no escape from the objection now under consideration. Causal explanations explain the occurrence of an event by indicating part of its causal history. And although this is compatible with explanantia that do not refer to causes, it still presupposes that the explanandum sentence asserts that an event of a certain kind took place. The problem now is to bring this in agreement with first the claim, that one can intentionally destroy a plane, secondly that intentionality entails the possibility of a causal explanation, and thirdly that there are actually no destroyings.

284

Ralf Stoecker

is incorrect; there is nothing the doctor does while she lets Bertram die that causes his death. What she does is sitting on the bar stool, drinking, smoking, thinking etc. None of these events is a cause of the child's death. Consequently, admissive actions cannot be approached in the way Davidson proposed for productive ones. But they can be approached almost in the way of Davidson's proposal. The only modification needed is, to my mind, still very much in Davidson's sense.17 What is merely required is moving back one step in his argumentation for the logical form of productive action sentences quoted earlier (the passage from Adverbs of Action, p. 237). At first Davidson assumes that they have the form: "Arthur caused a destruction of the aeroplane"; and then he continues: "But agents cause things to happen by doing something", leading to his final proposal. What I want to suggest is another understanding of Arthur's causing the destruction of the plane, an understanding that does not presuppose that one can only cause something by doing something: In a more general sense people as well as material objects can be said to cause events whenever reference to them provides a causal explanation of the occurrence of the particular event. At this point the results from the first part of my paper fit in very well. Causal explanations do not always say what the causes of the explanandum event are, they sometimes merely indicate the range of possible causes. Hence, although a person can cause something by doing something, it is not true that she has to. And according to my reading of Actions, Reasons, and Causes this is the way reason explanations usually work, they explain by merely indicating causes, without specifying them. Now, the range of causes indicated by a reason explanation sometimes (and perhaps often) comprises events inside the particular person (e.g. Davidson's perception of the switch), but it can also be a range of external causes — even causes very far away from the person that has the reasons. This, I take it, is the case in admissive actions. Bertram's doctor believes that it is her professional duty not to keep a patient alive who has no hope of leading a human life; she is certain that Bertram is such a patient and she wants to do her duty. These reasons rationalize her letting Bertram die — not by causally explaining her smoking or thinking about her stolen car, 17

In his response at the ZiF conference Professor Davidson preferred another way to handle admissive actions, the way he approached omissions in his Reply to Bruce Vermaßen. I cannot do justice to this line of argument here, but it has to be emphasized that admissive actions are, at least, a very special kind of omissions and that, therefore, any treatment of omissions or negative acts (as they are sometimes called) has to take into account the fundamental difference between e.g. intentionally not giving antibiotics and intentionally letting someone die. In the latter case and not in the former, there is obviously something happening (a person is dying) and the agent admits that it is happening. In this important respect admissive acts are different from other negative acts.

Reasons, Actions, and their Relationship

285

but by causally explaining Bertram's death. Knowing the doctor's attitudes we know better what kinds of causes led to his death. (For example, that he did not have an allergic reaction to antibiotics, nor was he killed by an evil nurse.) I hasten to admit that this is a very sketchy and speculative application of my considerations about Actions, Reasons, and Causes. None the less it may provide a way out of the problems discussed. As with productive actions, admissive actions do not really exist, like the former they are the product of a reification of the person's causing something. And as in their case this does not force us to regard the particular action sentences as false. It is still true that people intentionally let other people die; because this only entails that their reasons causally explain the other people's deaths. A second advantage of the proposed approach is its universality. As stated in the third objection to it, Davidson's view of productive actions is handicapped by being highly selective and, in particular, by being not applicable to admissive actions. The suggested approach to admissive actions, in turn, may be easily put to use also for productive actions and even for actions in general. Arthur longed for some acknowledgment from his political group, and he was sure that blowing up the airplane would get him the desired appreciation. This causally explains the destruction of the plane, since it dramatically narrows the range of its possible causes. As I said earlier: Doing something that causes something else is one kind of causing the latter. Now it would still be odd to claim that there are no acts of killing or letting someone die, while there are movements, travelings and talkings. If "The doctor let Bertram die" takes the form of "The doctor caused Bertram's death", one may wonder if "Carla moved her hand" should not be read as "Carla caused her hand to move". There is no ontological pressure for this claim, since it is usually easy to identify the moving of the hand with the hand's moving. Yet, as Davidson pointed out in Problems in the Explanation of Action, there are ways to move a hand that forbid such an identification (e.g. when a partially paralyzed person moves her hand with the other one). 18 In those cases moving a hand is a productive action and therefore ceases to exist. In view of this strange consequence, it is not too absurd to regard simple actions to be causings as well, with the final consequence of denying their existence. There certainly is an event of a hand moving through the air, and it is usually easy and convenient to call this event the person's action. But as long as one wants to maintain that there is one characteristic feature shared by all actions, I suggest that it is to be found in the fact that actions are causings — 18

Ibid. p. 37.

286

Ralf Stoecker

causings o f e v e n t s like the m o v e m e n t o f the hand, the destruction o f the airplane, o r the death o f a child, and causings that o n l y o b t a i n w h e n the causing p e r s o n has reasons that rationalize the caused e v e n t . Reasons are n o t causes, t h e y can at best be said to explain causally, and w h a t they explain is not the o c c u r r e n c e o f an action b u t o n l y o f an e v e n t that may in p r i m i t i v e cases and f o r reasons o f c o n v e n i e n c e be identified w i t h the action. Strictly speaking, the action is n o t the e v e n t but the causing o f the e v e n t . A n d f r o m an o n t o l o g i c a l p o i n t o f v i e w this is n o t e n o u g h to s u r v i v e . 1 9

REFERENCES

Davidson, Donald: Essays on Actions and Events [EAE], Oxford 1980 — Inquiries into Truth and Interpretation [ITI], Oxford 1984 — Actions, Reasons and Causes, in: [EAE] — Adverbs of Action, in: Vermazen, Hintikka (eds.), "Essays on Davidson: Actions and Events", Oxford 1985 — Agency, in: [EAE] — Freedom to Act, in: [EAE] — How Is Weakness of the Will Possible?, in [EAE] — The Individuation of Events, in: [EAE] — Intending, in: [EAE] — The Logical Form of Action Sentences, in: [EAE] — Mental Events, in: [EAE] — The Method of Truth in Metaphysics, in: [ITI] — Problems in the Explanation of Action, in: Ph. Pettit et al. (eds.) "Metaphysics and Morality — Essays in Honour of J.J.C. Smart", Oxford 1987 — Psychology as Philosophy, in: [EAE] — Reply to Bruce Verma\en, in: Vermazen, Hintikka (eds.), "Essays on Davidson: Actions and Events", Oxford 1985 — Reply to Quine on Events, in: LePore/McLaughlin (eds.), "Actions and Events", Oxford 1985 For further bibliographical information see the selected bibliography at the end of this volume.

19

I want to thank Peter Bieri, Jack Cahoon, Toni Koch, Wolfgang Künne and Peter Lanz for their helpful criticisms of an earlier version of this paper.

Donald Davidson: Reply to Ralf Stoecker This is a paper rich in ideas. I will be able to comment on just a few of the interesting issues Ralf Stoecker raises. The first one has to do with the question how we want to use the word "cause". Everybody allows that most talk about causality is interest-relative; what we call "the" cause of some event is some feature chosen from the totality of causal factors which particularly interests us, something we find surprising or out of the ordinary. In giving a causal explanation of an event we normally take for granted a great deal of background; what we typically want to know is what to add to that background to make the occurrence of the effect intelligible. If we take this generous attitude towards the nature of a cause, all kinds of things can be causes. People can be causes, stones can be causes, states can be causes, events can be causes. In everyday discussion we may call any of these things causes. To call reasons causes doesn't go against this common way of speaking or thinking about what's involved when we ask about causes, even though "primary reasons", as I have used the phrase, are certainly not events. At the same time I would give a certain priority to events. Suppose that what is to be explained is the fact that an aeroplane blew up, or that a hand moved, or that a child died. The "total" cause of such an event must always include an event or change, whatever we happen to mention as "the" cause. This claim doesn't imply that we necessarily know what the relevant change was. In many situations where the effect is an event, such as the explosion of a boiler, we're sure that there was some prior change that constituted part of the causal nexus although we are not in a position to say exactly what that change was. All we know is that the pressure in the boiler was very high, that the boiler had been previously weakened by rust, and so we say, that's why the boiler exploded. No event has been mentioned in the description of "the" cause. Or we may say that the pressure became too great; but this is also a way of expressing our ignorance. In the same way, we often know that it was a condition for somebody acting intentionally that he had certain beliefs and desires, and we call these "a" or "the" cause of the action. But beliefs and desires are not changes. They are states, and since I don't think that states are entities of any sort, and so are not events, I do not think beliefs and desires are events. So when we mention beliefs and desires to explain an action, we

288

Donald Davidson

are describing key aspects of the circumstances under which the agent acted. Typically, in fact, the connection between an agent's having certain attitudes and his acting is closer than these remarks suggest, for it is changes in the attitudes, which are events, and which are the often unmentioned causes. We actually mention an event if we give as a reason someone acted a perceiving, a noticing, or a surge of passion. And we can often turn a causal explanation which mentions beliefs or desires into an explanation which refers to an event or events by saying the cause of the action was the advent of one or both of the belief-desire pair. Up to this point I don't think I have disagreed with anything Ralf Stoecker has said, except that I'm distinguishing two uses, a broad popular use, and a rather more limited use, of the notion of cause. The more limited use allows only events to be causes; the broader use accepts objects, conditions, and, of course, events, as causes. So I quite concur that using the word "cause" as Stoecker does, and "reasons" as I did in "Actions, Reasons, and Causes", reasons are not causes. I do not believe I have ever made the mistake of supposing that beliefs and desires are events. Stoecker asks why I suppose there are events, and he speculates about the ontological status of states. It seems to me he does not quite appreciate my main reasons for postulating events as particulars. My main reasons are semantical: I accept an ontology of events because that ontology provides the only account I find persuasive of the semantics of a large category of sentences and the entailment relations of those sentences. I do not think the lack of a perfectly general and useful criterion of event identity is any more serious for events than for objects; one only gets fairly solid criteria when one considers sorts: sorts of objects or sorts of events. States are another matter. Not only do we have no good idea how to individuate them, but, more important, there seems no clear semantic need to treat them as entities, just as there are no compelling reasons to introduce entities corresponding to predicates in giving the semantics of predicates. We can for many purposes, at least, think of something being in a given state simply as a matter of some predicate being true of that thing. Thus someone believes that coal is black if and only if the predicate "believes that coal is black" is true of that person. (This does not give the logical form of belief sentences on my view, but the general point about the predication of states to objects or locations remains.) This brings me to the second part of Stoecker's paper, which proposes that there are no such things as actions. This thesis seems to me very hard to defend; defending it would require, among other things, rethinking the semantics of a large number of sentences. Since Stoecker makes no suggestions about what the semantics of action sentences should be to accommodate his thesis, I will not speculate on the outcome of an attempt to provide such a semantics. But it is worth discussing the difficulties in

Reply to Ralf Stoecker

289

my theory of action that drive him to his revolutionary conclusion. The difficulties he thinks he sees are two. One concerns what he calls "admissive actions", cases where an agent allows something to happen that would not otherwise have occurred. Stoecker distinguishes such cases from what have been called negative actions where the agent simply refrains from doing something, but I do not see what the interesting difference is. Whether it is a matter of not saving another person's life when one has the means, or of not smoking, an agent intentionally alters the course of events. The point in both cases is that what the agent "does", his "action", is not an event. Suppose we decide to say that allowing someone to die, like quitting smoking, is an action. Then some actions are not events. I see no great harm in talking this way. It is perhaps a question how apt we are to call such omissions and allowings actions, since it is natural to say in such cases that the agent intentionally refrained from acting: no event, and no action. The decision here is, as far as I can see, purely verbal. Stoecker's second point concerns a puzzle I think I was the first to state clearly. It's the puzzle that's now often called "the time of the killing", which raises the question when an agent killed his victim when there is a delay between the originating act and the death of the victim. Much of Stoecker's discussion is directed to what I wrote when I first addressed the problem and not to the modifications I have since introduced in the attempt to make my solution more acceptable. Let me spell out one of these modifications. For reasons that are not unique to actions we must, I think, analyze sentences that contain causal verbs like "destroyed" or "broke" (in their transitive uses) as involving two events. Thus "The ball broke the window" has to be analyzed to at least this degree of complication: "There were two events, a motion of the ball, and a breaking of the window, and the first event caused the second". The complication is required in order to explain the relation between the transitive and the intransitive uses of verbs like "break" and "destroy", why it is that "The ball broke the window" logically entails "The window broke". Similarly, "Arthur destroyed the plane" becomes (roughly) "There was an action of which Arthur was the agent, there was some event that was a destruction of the plane, and the first of these events caused the second." This means, as Stoecker explains, that if we ask, "When did Arthur destroy the plane?", we're asking about two different events, and they may have different dates. The right response to the question is therefore: Which event are you asking about? But if we must pick out one of those events to call the action, then it's the thing that Arthur did that resulted in the destruction of the plane. The further twist, which Stoecker misses, is that on this analysis there is no need to infer, from the fact that Arthur's placing of the bomb was his destroying of the plane, that he destroyed the plane

290

Donald Davidson

before the plane was destroyed. Arthur did not do something that had the consequence that the plane was destroyed until the plane was destroyed. The point has to do with tense, not epistemology as I earlier thought. Stoecker's only further difficulty with my analysis of actions characterized in terms of their consequences is that we would not normally say that Arthur's action of destroying the plane caused the destruction of the plane. Of course we wouldn't say it, since "Arthur destroyed the plane" logically entails that the plane was destroyed. My analysis explains why we wouldn't say it; but does Stoecker have an alternative semantic analysis of the sentences that explains the entailment? [In response to a question regarding the relationship between logical form and conceptual analysis:] The question where the line comes between conceptual analysis and logical form is not always a clear one. There are examples where no more is involved than questions of efficiency, simplicity and degree of fidelity to surface syntax. But once you have set your standards for a semantical theory there are many cases where the decision is straightforward. Take, for example, the relationship between "John broke the vase" and "The vase broke", that is, between a transitive and an intransitive use of the same verb. On my proposal about the logical form of the first sentence, it entails the second sentence by the laws of first-order quantification theory. An alternative would be to treat the transitive and intransitive forms of "break" as separate predicates, and to treat the relation between them as purely contingent, or as established by a new rule of inference, that is, by appeal to the logical analysis of the two predicates. This would be possible, but intuitively unsatisfactory, and semantically inefficient. Intuitively unsatisfactory, the theory would require that the same verb have semantically unrelated meanings. The two meanings would be related only by an ad hoc piece of analysis. Inefficient, since there would be many other verbs each of which would demand its own separate rule of inference. In this case the decision is easy, though not forced. It is forced in many cases, however: for example, if we demand that the primitive vocabulary of the language be treated semantically as finite, we cannot construe adverbs simply as part of the spelling of the verbs they appear to modify.

The Explanatory Force of Action Explanations 1 PETER L A N Z

An undisputed datum for all theories of action is that mentioning pairs of beliefs and desires often explains why an agent acted as he did. She bought a copy of Joyce's Ulysses, because she desired to have her own copy of that work and believed that buying a copy of Ulysses would satisfy that desire. This is as familiar and pervasive as it is central to our understanding of what an action is. An action is something that fits this sort of explanatory schema, epitomized long ago by Aristotle in the practical syllogism·. The contents of a belief and a desire provide the premisses of the argument, and performing the action is the drawing of the conclusion. This is not to be taken as a piece of introspective psychology but as a philosophical analysis of what it is to give an explanation of an action: It is to give the reason, why someone acted as he did. Now, if mentioning χ explains y, then one can ask: In virtue of what does mentioning χ explain y? One would expect that there is some sort of systematic connection between χ and y, if mentioning χ explains y. A correct explanation is somehow anchored in such a non-accidental relation between χ and y. The idea here is that the concept of an explanation is not a purely psychological or pragmatic one. Satisfying a cognitive need of some cognitive agent by mentioning χ is not all there is to explaining y by mentioning x. A correct explanation points to a real relationship between χ and y which is not only in the eye of the explainer nor only in the eye of the consumer of the explanation. To return to reason-explanations the question is: Is there a pattern, i.e. a non-accidental connection between reasons and actions, in virtue of which mentioning a reason explains a given action? Related to this question are others: What does such a pattern look like? Is it significantly different from patterns which underlie explanations in the physical sciences? Is there discontinuity rather than continuity between these possibly different forms of explanation?

1

I am very grateful to Louise Röska-Hardy who corrected my English. — I benefited from the many discussions I had with Ralf Stoecker on Donald Davidson's philosophy during the preparation of this conference.

292

Peter Lanz

Philosophers have produced three different accounts of the nature of the pattern underlying action-explanations and consequently of the explanatory force of such explanations. There is the causal-nomological account, adopted among others by Ayer, Paul Churchland, Fodor, Follesdal, Hempel and Alvin Goldman. There is the purely normativeintentionalist account, adopted among others by Anscombe, Dennett, Kim, Melden, Stoutland and v. Wright. And there is the distinctively Davidsonian account which tries to integrate the causal and the normativeintentional accounts by rejecting those elements in the first two accounts that seem to preclude such an integration. [I doubt that the simulationapproach developed by Stich and recently refined by Gordon and Goldman is really an independent alternative to Professor Davidson's proposal. These philosophers are placing more stress on certain psychological pecularities of a fact which Professor Davidson has always taken to be very important to the task of understanding others, namely, "the need to view others, nearly enough, as like ourselves" (Davidson 1980: 239).] The first attraction of Professor Davidson's account is that it seems to be descriptively or phenomenologically correct. The unprejudiced observer discovers that there are two relations holding between the explanans, i.e. the belief-desire pair, and the explanandum, i.e. the action. One is a logical relation: the contents of the belief and desire imply that there is something valuable or desirable about the action. The contents of the belief-desire pair, which constitute the reason for which one acted, recommend the action and, therefore, they make intelligible why one performed the action. Acting from a recommendation he judges to be the comparatively best is what we expect from an agent who has his wits about him. The other relation is a causal relation: The reason must have been efficacious in bringing about the occurrence of the action, otherwise the reason could not explain the action. In virtue of the first relation reasons are rationalisers·, knowledge of the reason gives us an understanding of what the agent sees in his own action; in virtue of the second relation reasons are movers. To say that reasons are causes then means that persons are moved by what recommends and thereby rationalizes their actions. Reason-explanations give us the sort of causal information we need to understand others as persons, i.e. as beings that lead their lifes guided by their own desires and beliefs, pursuing their own projects, recognizing opportunities, obstacles and objections and reacting to them. It follows that neither relation exhausts the explanatory force of reason-explanations. Both relations have to be present if a reason-explanation is to do its proper job. In the light of a primary reason, an action is revealed as coherent with certain traits, long- or short-termed, characteristic or not, of the agent, and the agent is shown in his role of Rational Animal (Davidson 1980: 8).

The Explanatory Force of Action Explanations

293

Without rationalizing the action a cause cannot help to explain the action as an action, i.e. as something the agent untertook, because he saw something valuable or desirable in it. Without causing the action the reason cannot help to explain why the action occurred. Another attraction of this account is that it fits well with a relatively clear account of the sort of autonomy involved in human agency. This autonomy results if the causal force of what determines the course of action taken by the agent reflects the rational force of the reasons considered by the agent. One distinctive element in Davidson's picture is that neither empirical laws nor principles of rationality or principles of practical reasoning play any important role in accounting for the explanatory force of actionexplanation. Here, both the defenders of the causal-nomological doctrine as well as the defenders of an intentionalist account disagree. Proponents of the causal-nomological account argue that mentioning a reason is explanatory, because there is a law subsuming the belief-desire pair and the action. According to Hempel, it is the law which "confers explanatory force upon the explanans" (Hempel 1966: 99f.). The argument is that only knowledge of some law or law-sketch rationally grounds the conviction that mentioning a belief-desire pair amounts to picking out a causally and explanatorily relevant factor. It is assumed that our concepts of belief and desire, taken by themselves, are explanatorily neutral. They only become explanatorily useful as elements of lawful statements. If you accept this, you will sympathize with the following question, which Ayer raised in a critical discussion of Professor Davidson's views: "... if mental events cannot be sufficiently pin-pointed to be candidates for subsumption under strict laws, why should it be thought that they can be sufficiently pinpointed to be identified as causes and effects?" (Ayer 1982: 188). — Moreover, that there are laws covering reason-explanations is the central argument for those who think that psychology is continuous with the rest of science. Seen from this perspective abandoning the causal-nomological account of reason-explanation seems to many philosophers to be too costly, because it sets intentional psychology apart from the rest of science. The proponents of an intentionalist account, however, hold that the principles which underwrite reason-explanations are not concerned with the causal order but with the rational order "which [in Anscombe's words] is there whenever actions are done with intentions" (Anscombe 1957: 80). Reason-explanations explain because they are backed by principles which help us identify the reasons that make a particular action justified, intelligible, rational, meaningful, or somehow significant to us. The relation is a logical or rational or normative relation. In other words, the relation is not a relation of making the case but of making intelligible, in

294

Peter Lanz

short, a kind of rational relation. Mentioning the attitudes explains the action, because it makes the action intelligible. The sort of law mentioned by the proponents of the first proposal is not an empirical law but a principle of practical reasoning, a norm to which rational agents are responsive. Who is right? — It is interesting to note that when it comes to the task of producing an example of a corresponding law or a corresponding principle there is much more agreement. All parties in the dispute take their cue from the practical syllogism. Proponents of both camps tend to produce something like the following statement as a first attempt at a law or at a principle of rationality: (S) Given any agent x, if χ desires that p and believes that doing a will bring about p, then χ does a. Considering (S) to be an empirical law, however, creates difficulties. First of all, as (S) stands, it is false. It is all too easy to give counterexamples. χ may desire something else, say q, more strongly than he desires p. χ may believe that Λ is a means of attaining p, but not the most efficient, or cheapest, or elegant, or enjoyable means of attaining p. Even if χ believes that a is the best means of attaining p and has no overriding incompatible desire q, χ may not know how to execute a, or knowing how, may be unable to do it. Of course, the proponent of (S) can improve (S) by adding to the antecedent a further clause each time the critic comes up with a type of counterexample. The result will be a much more complex statement like this: (L) For any agent x, if 1. χ desires that p 2. χ believes that performing a is a means to bring about p (under circumstances C) 3. there is no alternative option believed by χ to be a way of bringing about p that under the circumstances C is preferred by χ 4. χ has no desires q or s or t that override p 5. χ· knows how to do a 6. χ is able to do a then 7. χ does a (or ceteris paribus χ does a). Now, does (L) look like an empirical law? Consider first the method by which we arrived at (L); this method has nothing to do with observation, data collection, experimentation and induction, but rather with probing our own understanding of what an action is, of what we would count as a well-founded objection to a proposed reason-explanation and of what

The Explanatory Force of Action Explanations

295

we assume to be fulfilled when a reason explains an action. The antecedent of (L) does not give us necessary and sufficient conditions for χ does a: deviant causal chains, weakness of will, a strong liability to hesitation, forgetfulness, absentmindedness are cases where either antecedent and consequent are not connected in the right way or the conditions mentioned in the antecedent are not sufficient to set off the action. Nor can we simply assume that we can test whether the conditions hold independently of the action. What, for example, is evidence that there is no overriding desire before χ acted? It seems, therefore, that (L)'s antecedent rather gives us a list of conditions we believe to be fulfilled, when someone did a. In other words, it would be difficult to understand what it could mean that someone actually performed the action a, but one or more of the antecedents of (L) were not fulfilled. (L) articulates something of our reflective understanding of our concepts of belief, desire, and action. This becomes still more evident when one observes the following: You can exchange any of the the six antecedents of (L) for the consequent and thereby articulate part of your understanding of the concept which now, after the exchange has been made, shows up in the consequent. Exchange for example 1. χ desires that p for 7. χ does a\ the result will be: (L)dcsire

1. χ does a 2. χ believes that performing a is a means to bring about p (under circumstances C) 3. there is no alternative option believed by χ to be a way of bringing about p that under the circumstances is preferred by χ 4. χ has no desires q or s or t that override p 5. χ knows how to do a 6. χ is able to do a then 7. χ desires that p. Now you have formulated some of the things involved in believing χ desires that p. Here one might object that the relation between (L) and (L) desire exemplifies the same sort of transformation we are familiar with from the case of equations in physics. 2 Consider for example Boyle's law: PV 2

=rT

Professor Eike von Savigny put forward this objection in the discussion of this paper.

296

Peter Lanz

where Ρ is pressure, V is volume, Τ is temperature, and r is a constant that varies with the gas under consideration. From this law we infer that there is an inverse ratio between the pressure and the volume of an ideal gas, given the temperature of the gas remains constant. But is this similarity, pointed out by the objection, not only superficial? The reasons, in virtue of which the transformations in the two examples exist, are different: The relation between (L) and (L) dcsire states a normative constraint we expect rational agents do respect in their attitudes and in their behaviour. Boyle's law and its corollaries restrict the range of physically possible states of ideal gases. Answering some further pertinent questions will confirm the impression that (L) is not an empirical law. By pertinent questions I mean questions we are used to asking about genuine empirical laws. How could (L) be tested or even be refuted? (L) does not seem to be used as something which should be tested, but rather as something which our findings must conform to, if we are to be able to make sense of our findings in the realm of intentional human behaviour. If an agent does not act in the way we have reason to expect, given the beliefs and desires we have attributed to him, we do not take this as evidence of the need to revise or to improve or even to replace (L). Rather we naturally conclude that some of our attributions were mistaken, that the evidence we relied on was misleading or that the agent changed his mind between our reading the signs and his doing the action. On no account do we tolerate the idea that there are non-rational principles connecting belief-desire pairs with actions. The reason for this is that (L) is already involved in the inference back from behaviour to belief and desire. How do we tell, for example, that our neighbours believe a drawing of Mont Ventoux by Cezanne is worth more than a photo of that mountain, or prefer the former to the latter? Ask them or offer them a choice of one or the other. They all answer in favour of the Cezanne or they all take the drawing by Cezanne. But these behaviours are marks of their belief that the drawing by Cezanne is worth more than the photo only if their behaviour reflects their beliefs and desires in accordance with (L). If you find someone among your neighbours who has the tick of always, almost automatically answering in favour of a drawing or taking a drawing in choices of that type, even if the drawings are of no worth whatsoever, you may hesitate to infer from his behaviour to a belief in the afore-mentioned way. In order to use behaviour as a guide to belief, we have to hold the agent's desires constant. And in order to employ behaviour as a guide to his desires, we have to hold beliefs constant. Any action can be the result of almost any belief, provided the agent has the appropriate desire, and vice versa. If we know what someone's beliefs and desires are, then (L)

The Explanatory Force of Action Explanations

297

will tell us what actions he will undertake. If we know what actions a person performs and we know his beliefs, then (L) will tell us what he wants. And if we know his wants and what actions he performs, then (L) will tell us what he believes. But without at least two of the three, belief, desire and action, the third is not determinable. The sort of coherence among beliefs, desires, and actions, required by (L) and its variants is as much a necessary condition for the determination of these items as it is a precondition for the explanatory force of reason-explanation. This fact has important consequences: • It is not possible to give a psychological characterization of an agent in intentional psychological terms using his behaviour as a source of evidence for this characterization and at the same to hold that it is an open and purely empirical question whether that agent is rational. Contrary to what Hempel maintained, being rational is not an empirical psychological predicate on a par with remembering the president's speech or being tired. Rationality does not enter the picture as an extra premise in action-explanations, because it is a precondition of an agent's having the desires and beliefs these explanations ascribe. • (L) cannot be refined in the ways empirical laws are typically refined, namely by finding more and more ways to measure the initial conditions (the conditions mentioned in the antecedent of the law), ways which are independent of the law in question. • Contrary to the more optimistic outlook of the proponents of the causalnomological account, intentional psychology is somehow insulated from the rest of science. • Abandoning (L) amounts to abandoning the subject matter of intentional psychology. I take it that these observations and arguments bolster a thesis naturally associated with the so-called logical connection argument. The thesis is that we never can find descriptions of beliefs, desires and actions fulfilling two conditions: (a) they can be used in action-explanations; (b) they are independent enough from one another to enable us to frame laws about them that have much informative content or improvable predictive power. Professor Davidson has demonstrated that this thesis does not sustain the conclusion of the logical connection argument, namely, that reasons cannot be causes. Distinguishing between events and their descriptions and between relations holding between events and relations holding between descriptions of events is sufficient to block the fallacious reasoning. Other reasons tell against the purely intentionalist account: • The terms ordinarily used to connect beliefs and desires with actions are mostly cognates for causation: Beliefs and desires determine action, produce it, result in it, bring about action.

298

Peter Lanz

• Intentionalists cannot account for the difference between a reason for acting and a reason for which one acted. They cannot explain how it is possible that a proposed reason-explanation is wrong although the mentioned reason rationalizes the action. • We try to influence the behaviour of other persons by influencing their beliefs and desires. How is this possible, if beliefs and desires are not causal factors controlling behaviour? • Many emotions depend on beliefs and desires. One becomes sad when one hears that a friend took the airplane now in the hands of the hijackers; one is relieved when the friend calls to say that he missed the airplane. How is this possible if beliefs and desires are not causes? • To say that a behaviour χ is an element of a pattern with meaning Μ or a behaviour χ forms, together with other elements, a pattern with meaning Μ does not explain why χ occurs. But how can desire-belief pairs be explanatory when it is neither an empirical law nor a principle of reason which confers explanatory force on them? Professor Davidson's answer is: "... the propositional attitudes are by nature explanatory" (Davidson 1986: 204). This is the case because it is part of our concepts of a belief and a desire that they have rational effects one upon the other and on behaviour. Beliefs breed other beliefs and, together with desires, actions, on the whole, along rational lines. There is a conceptual connection between reasons and actions. This does not preclude causal explanation as the following example shows. We may explain the fact that some iron filings moved towards a piece of steel by pointing out that this piece of steel is a magnet. The magnet caused the iron filings to move. Nevertheless there is a conceptual connection between the descriptions used in the explanandum and those used in the explanans. To be a magnet is to be something which ceteris paribus attracts iron filings in its neighbourhood. The explanatory force of this explanation does not depend on a law. If there is a law at all, then it is the low-level statement of tendency you are familiar with if you know the meaning of the word magnet·. A magnet is something which tends to move iron filings. I take it that Professor Davidson's answer in the case of the explanatory force of beliefs and desires essentially parallels the case of magnets or more generally the case of causal, functional and dispositional concepts, i.e. concepts which partly are defined in terms of their typical causes or effects or in terms of both of them. What distinguishes the concepts of belief and desire from causal concepts like magnet, sunburn, poison, sleeping pill, migratory and sedentary birds etc. is this: they are "multitrack" dispositional concepts (Ryle). There is neither a short nor a long list of typical effects a belief must have to be, say, the belief that Paris is

The Explanatory Force of Action Explanations

299

the capital of France. This contrasts with a dispositional concept like migratorj bird. This concept is defined in terms of a characteristic seasonal behaviour. Correspondingly, one expects science to reveal one speciesspecific mechanism which underlies this behaviour. We therefore find a hierarchy of explanations beginning with the ecological one which gives something Dennett called a "free-floating reason". Swallows, taking them as our example of migratory birds, being insect eaters, must migrate, because they would starve to death if they tried to winter in central europe. This explanation leaves open how swallows are able to do what the given reason requires them to do. At the next level one introduces talk of a sensibility or liability, i.e. of dispositions and behavioral regularities: Swallows have acquired a genetic constitution in the course of the evolutionary history of their species which induces them to respond appropriately to the proper stimuli from the environment. At the next level biologists search for internal mechanisms and intrinsic physiological causes: Swallows flee south because migration is tied in with photoperiodicity: they respond to the decrease in day length and they are ready to migrate as soon as the number of hours of daylight have dropped below a certain level. One might then also detect an immediate or triggering cause: A cold air mass with a sudden drop in temperature affected the birds, already in a general physiological state of readiness for migration, so that they actually took of on that particular day. The case of action-explanation is different. The reasons are not freefloating but represented by the agent himself. And, more to the point: The beliefs and desires an agent has often do not require or determine one unique course of action. Agents have to make up their own minds and they can do this in different ways whithout thereby violating "norms of rationality". These norms are loose and therefore it is often rightly disputed what it is that reason requires (a fact, familiar to philosophers, I suppose). The empirical and therewith explanatory import comes in when one is presented with specific desires and beliefs. Then one can employ one's own ability to reason practically and utilize this ability vicariously: which course of action should one take given a specific set of attitudes? Of course, rationality is presupposed here too, but not as a very specific set of norms or an articulated theory specifying in detail what is required, but rather as the assumption that other people exemplify in their reasoning and in their actions patterns which we can recognize as thoughts and actions, i.e. as things, which, by and large, hang together in rational ways. In this sense reason-explanations are norm-based explanations because they make use of the idea that an agent does what is required, given his own beliefs and desires. If an agent fulfills this expectation then our conception of him as a rational agent is confirmed and we can make further explanatory use of what we think are normative implications of

300

Peter Lanz

his own specific beliefs and desires. As in the case of design-based or function-based explanations, we simply assume that the underlying mechanisms are in good working-order. To put it in other words: Causal concepts like the concepts of belief and desire are concepts of the sort of difference a state or a property is apt to make and not concepts of how this difference is brought about or produced in organisms. Knowledge of the rational abilities of an agent can sustain our confidence that a certain behavioural effect will show up, even if we do not have any knowledge of mechanisms, i.e. of the ways the organism produces this effect. To know that someone has certain beliefs and desires is to have a certain amount of causal information about him; but this sort of causal knowledge is not knowledge of mechanisms or of executing processes. It is causal knowledge about a rational agent. I suppose that this also provides a plausible way understanding Professor Davidson's claim that "the mental is not an ontological but a conceptual category" (Davidson 1987: 46). To see rational causes at work does not commit one to a non-physicalist ontology, but rather reflects the cognitive needs of rational agents in their interactions among one another. We therefore end with a sort of anthropocentfic causality, which concedes that in the case of the physical properties which science will reveal to be causally relevant to what goes on in the brain, there will be no kinds which strongly enough correlate with what we are used to calling beliefs and desires. Consequently, intentional psychological predicates seem to be on a par with colour predicates. On the side of the physical properties science reveals to be causally relevant for the colour impression which a certain stimulus induces in us, there are no properties which classify objects, surfaces and media into kinds which correlate well enough with those kinds of objects, surfaces and media which fall under our predicates red, green, yellow or blue. Intentional psychological predicates then seem to be as irreducibly anthropocentric as colour predicates are. In both cases the reason is somewhat the same: The facts cannot be stated without reference to the pecularities of our own human nature; in the first case the pecularities concern our reasoning powers; in the second case the pecularities concern our visual system.

REFERENCES

G. Ε. M. Anscombe 1957, Intention, Oxford: Blackwell. A. J. Ayer 1982, Philosophy in the Twentieth Century, New York: Random House. Donald Davidson 1980, Essays on Actions & Events, Oxford: Clarendon Press.

The Explanatory Force of Action Explanations

301

Donald Davidson 1986, Judging Interpersonal Interests, in: Jon Elster & Aanund Hylland (eds.) 1986, Foundations of Social Choice Theory, Cambridge: Cambridge UP: 195 — 211. Donald Davidson 1987, Problems in the Explanation of Action, in: Ph. Pettit et al. (eds.) 1987, Metaphysics and Morality. Essays in Honor of J. J. C. Smart, Oxford: Blackwell: 35-49. C. G. Hempel 1966, Explanation in Science and in History, in: William H. Dray (ed.) 1966, Philosophical Analysis and History, New York: Harper & Row: 95 — 126. Peter Lanz 1987, Menschliches Handeln zwischen Kausalität und Rationalität, Frankfurt/M.: athenäum.

Donald Davidson: Reply to Peter Lanz I can express nothing but admiration for and agreement with this paper. It seems to me right in every respect, which leaves little for me to add. I particularly admire the treatment of the so-called laws of action which have sometimes been proposed by Hempelians, Paul Churchland, and others. Lanz brings out clearly the fact that what looks like a law may in certain cases simply be (part of) an analysis of the concept of an action, or the concept of belief, or the concept of desire. All these concepts are revealed as the kind of concepts they are, in part, anyway, by these "laws" connecting them with one another. I like the emphasis on the fact that explanations of actions make heavy use of causal concepts, i. e. concepts that have the notion of cause irreducibly built into them. It seems right that the notion of a belief, for example, is in part the concept of a state which has certain causal properties just as an action is the concept of something that is caused in certain ways. And like Lanz, I see no way of getting rid of these features of psychological concepts without simply changing psychology into something different. There is nothing wrong, of course, with trying to be more "scientific" about human behavior: the result of such attempts just won't answer to the particular interests that we have in reasons and actions. It is obvious that we use causal concepts in our ordinary everyday understanding of all sorts of things that are not actions. Many of us can do no better in trying to explain why some iron-filings moved than to say there was a magnet in the neighborhood; we know, of course, that magnets tend to move iron filings. The concept of a magnet is a causal concept: something wouldn't be a magnet if it didn't cause iron filings to move under certain conditions. Employed as explanations, appeal to such causal tendencies and dispositions leaves something to be desired because it fails to tell us what it is about an object that makes it behave as it does. To attribute a causal power to something is simply to say there is something about it that will cause certain effects given the right conditions. This is essentially Quine's approach to the analysis of disposition concepts and it seems to me the right one. The difference between disposition concepts in the physical sciences and in psychology is that we can hope that they will be replaced without loss in the physical sciences (as the concept of a magnet is not needed now that we can say explicitly what the properties

Reply to Peter Lanz

303

are that create magnetic fields), while it is only by changing the subject that they could be eliminated from psychology. I have one small suggestion concerning the final sentence in Lanz's paper, in which he proposes that it is due to the "reasoning powers" of human beings that certain basic psychological concepts are not reducible to the concepts of the hard sciences. I agree that it is because these concepts operate in a conceptual domain infused with rationality that psychology is different, but I see this difference as due to our special interest in interpreting human agents as rational agents rather than to special powers in those agents. We can try applying psychological concepts to other entities, or to the non-rational behavior of human agents. This has often enough been done; the results just don't seem very interesting to some of us.

Mental Concepts: Causal because Anomalous PETER B I E R I

1. In his recent paper Representation and Interpretation1 Professor Davidson says something striking about mental concepts which may provide a better understanding of Anomalous Monism. I want to use this occasion to explore his remark and to ask whether I got it right. 2. In Freedom to Act2 we read: Unavoidable mention of causality is a cloak for ignorance; we must appeal to the notion of cause when we lack detailed and accurate laws. In the analysis of action, mention of causality takes up some of the slack between analysis and science, (p. 80)

In Representation

and Interpretation

the same point recurs:

It is often thought that scientific explanations are causal, while explanations of actions and mental affairs are not. I think almost exactly the reverse is the case: ordinary explanations of action, perception, memory, and reasoning, as well as the attribution of thoughts, intentions, and desires, is riddled with causal concepts; whereas it is a sign of progress in a science that it rids itself of causal concepts, (p. 22 f.)

3. So far the picture is this: The notion of cause plays the role of indicating an epistemic deficit: We know there are strict laws but we don't know them yet. Our concepts indicating causal dispositions and causal relations, as well as the concept 'cause' itself, are place-holders for the laws to be discovered. Once these laws are known, the causal concepts become dispensable. In a sense, the vocabulary of causation is no more than a piece of heuristics. 4. As far as I remember, the notion of a mechanism does not play a systematic role in Davidson. However, it will be useful for later purposes to rephrase the above point in terms of mechanisms: The use of causal concepts indicates that we lack knowledge of the detailed mechanisms which are governed by the yet unknown strict laws. 3 5. Now suppose that in the domain of mental concepts, too, the notion of cause had its preliminary, purely heuristic role, and that reference to ' In: K.A.MohyeldinSaid/W.H.Newton-Smith/R.Viale/K.V.Wilkes (eds.), Modelling the Mind, Oxford: Clarendon Press 1990: 1 3 - 2 6 2 In: Essays on Actions and Events, Oxford: Clarendon Press 1980: 63 — 81 3 "In general,..., appeal to causal powers and dispositions reveals ignorance of detailed explanatory mechanisms and structures": Representation and Interpretation: 25

306

Peter Bieri

causal dispositions were understood here exactly as in the domain of physical concepts like, e.g., 'solubility' (Davidson's example). This could mean two things: Either we consider replacing reference to ordinary mental dispositions by reference to yet-to-be-discovered strict psychophysical laws which will be couched in a revised or entirely new mental vocabulary. Or we consider moving on to detailed physiological mechanisms and the corresponding strict laws, leaving behind any reference to the mental. In either case the original mental concepts introduced in action explanations (the concepts specifying reasons) would become dispensable, and we would lose what is constitutive of the mind and of mental explanations: the normative, rational features. Conclusion: There must be a different, nonstandard interpretation of the causal character of mental concepts. 6. This reflection is supposed to provide some background and motivation for the following claim which strikes me as novel: In order to keep intact the normative features that help define beliefs and other thoughts, a degree of looseness in their connections with events as described in noncognitive terms is required. The 'unscientific' concept of cause takes up the slack. This slack is not the slack of ignorance: it is the slack that must exist between two schemes of description and explanation, one, the mental, being essentially normative, the other not. (Representation and Interpretation: p. 25f; italics mine).

And this point is characterized by Davidson as providing "the hint of an explanation of the irreducibly causal character of the concepts we apply to thinking and acting" (p. 25; italics mine). 7. I read this passage like this: The content and thus the identity of (intentional) mental events is defined by the normative features of rationality. If we thought of mental events as covered by strict (psychophysical) laws, thus construing their connection with the non-intentional as "tight", we would have to construe their identity as defined by their nomological role, their role in natural laws — a consequence which conflicts with the priority of normative considerations emerging from the theory of interpretation. 4 Therefore, the connection must be construed as irreducibly "loose". The notion of cause, when applied to the mental, is a conceptual instrument to capture or express this (non-accidental, essential) looseness. Whereas in the domain of the physical causal talk foreshadows the discovery of more tightness in terms of mechanisms and strict laws, the notion of cause has here the opposite function of indicating the impossibilitj of further coherent steps towards more tightness of connection. In this sense mental concepts have, as Davidson says, an "irreducibly causal character", whereas the concepts applied to the physical are, one might say, "reducibly causal" 4

Cf. J. Kim, Psychophysical Laws, in: E. LePore/B. McLaughlin (eds.), Actions and Events, Oxford: Blackwell 1985: 3 6 9 - 8 6

Mental Concepts: Causal because Anomalous

307

in the sense that concepts of causal dispositions can, in priciple, be replaced by the non-causal vocabulary of strict laws. Slightly exaggerating: The notion of cause as built into mental concepts has a job to do which is much more systematic and substantial than usual: the job of keeping the mental and the physical schemes apart, thereby doing justice to the autonomy and irreducibility of the mental domain. 8. This way of looking at the causal character of mental concepts is, I think, relevant to the understanding of Anomalous Monism. It has been said 5 that Anomalous Monism fails to account for the causal role of mental properties, i. e., of mental events as such, under their mental description. Roughly, the charge is based on the following reasoning: For every event, there is, in principle, an explanation in terms of detailed physical mechanisms and strict physical laws. This is true for all behavioral events including actions. The explanations in question will refer to mental events, but this reference will be confined to their physical properties. Their mental properties could become relevant only if there were strict laws tying them to the physical properties, either in terms of strong supervenience or reduction. But according to Anomalous Monism there cannot be such laws, due to the constitutive normative features of the mental domain. Therefore, mental properties are epiphenomenal. 9. Viewed in the light of Davidson's recent claim this argument looks wrong-headed. True, the mental properties of mental events (their contents) do not fit in the world of strict laws and mechanisms. Their normative feature prevents that. But it is a mistake to conclude from this fact that mental properties cannot be causal. Quite the opposite is true: They are causal, and irreducibly so, because they are anomalous. The objection misconstrues the notion of cause by assuming that for something to be causal is for it to be embedded in mechanisms governed by strict laws. On the contrary, the level of mechanisms and strict laws is the level where causal concepts (and the concept 'cause') are no longer needed, and if a certain range of concepts is essentially causal, this means that the topic of these concepts (mental content) is not part of the world of laws and mechanisms. So it is wrong to think: (A) The mental events' (or mental properties') being anomalous prevents them from being causal. What is true is, rather: (B) The mental events' (or mental properties') being anomalous makes them (irreducibly) causal. 10. Suppose we follow the reasoning I am trying out here. We could then say this: Anomalous Monism is not at all in danger of undermining the causal role of the mental and the causal character of mental concepts. Objections to this effect seriously misconstrue the role played by the 5

See, e.g., J. Kim, The Myth of Nonreductive Materialism, Proceedings and Addresses of the American Philosophical Association 63 (1989): 31—47

308

Peter Bieri

concept of cause in the mental domain. Something very different is true: Anomalous Monism explains the particular sense in which mental concepts and mental events (as falling under these concepts) are causal. As to the charge of epiphenomenalism, the reply will now be this: Mental content, being normatively identified, can, indeed, play no role in (the science of) the fine-grained mechanisms that lead to behavior. In this sense it may perhaps be called 'epiphenomenal'. But this is compatible with the causal character of mental concepts (and, therefore, with the causal character of the events decribed by these concepts), for their irreducibly causal character, correctly understood, tells us precisely this: The idea of mental content does not have the job of teaching us anything about mechanisms. That's not how this idea works, and it is precisely this insight which is captured by the causal aspect of mental concepts and which is articulated in Anomalous Monism. 11. I am by no means sure that this extrapolation from the passage quoted fits Professor Davidson's intentions. But it has one thing to recommend itself: It would explain his stoic reaction to all charges of epiphenomenalism, a reaction I could watch a year ago in this very room. I shall now raise a few questions which come to mind when you compare the above proposal with other things Davidson has been saying. 12. Question 1: How is the claim that the notion of cause is 'unscientific', i. e., non-nomological, compatible with the 'Principle of the Nomological Character of Causality': "Events related as cause and effect fall under strict deterministic laws"? (.Mental Events6: p. 208) In particular: How can this principle be reconciled with the idea that, in the case of mental concepts, the notion of cause serves to indicate that there cannot be strict laws ? It seems to me that there might, as a matter of fact, be two different notions of cause in play here, or perhaps better: two different perspectives on the concept. One of them is the perspective familiar from Causal Relations1·. ... a singular causal statement 'a caused V entails that there is a law to the effect that 'all the objects similar to a are followed by objects similar to b' and that we have reason to believe the singular statement only in so far as we have reason to believe there is such a law. (p. 160)

According to this perspective, the existence of strict laws is of the essence of something's being a cause: If we know in advance (a priori) that there are no mechanisms and laws to be discovered, the very idea of causation loses its hold. According to the new perspective (or, at any rate, different perspective, for it might have been present in Davidson's thinking 6 7

In: Essays on Actions In: Essays on Actions

and Events, op.cit.: 207 — 27 and Events, op.cit.: 149 — 62

Mental Concepts: Causal because Anomalous

309

all along), on the other hand, invoking causal dispositions and applying the concept of cause is one thing, envisaging mechanisms and strict laws is another thing. You can coherently have the first thing without having the second thing. And more: Whereas in the first perspective the transition from causal statements to laws is all of a piece, this transition involves, in the second perspective, a change of topic: You leave behind the entire notion of cause. It is this separation which is presupposed when the notion of cause is assigned the job of indicating the autonomy (through normativity) of the mental. Now suppose someone were to say this: "Okay, I understand this new account of the concept of cause, and I can see how it is supposed to work with mental concepts. But until I am told how the old and the new account can be made to harmonize, I don't feel I have actually learned anything new. For it seems to me that calling mental concepts 'causal' in this new sense is really no more than just repeating that they are anomalous. But the promise of Anomalous Monism was to explain how mental events could, despite their being anomalous, have a causal role in the old (nomological) sense." What would Professor Davidson reply to such an objector ? 13. Question 2: If mental concepts are "irreducibly causal" in the particular sense explained: What happens to the argument for Anomalous Monism, in particular to the argument for token identity ? That argument says that mental events must have a physical description because, being causes, they must instantiate strict laws, and the only strict laws there can be are physical laws. Obviously, this argument operates with the old, nomological notion of causation. If, however, we follow Davidson's new explanation of the causal character of mental concepts, it would seem that the fact that mental events are conceived of as causes does not force us to postulate that they instantiate laws; on the contrary. But then the "demonstration of identity" (.Mental Events·, p. 224) does not get off the ground. What am I missing ? 14. Question 3: According to Davidson's semantic externalism 8 , thoughts and beliefs are, at least in the basic cases, necessarily about the sorts of objects and events that cause them. So mental content results from mind's causal properties. These causal properties cannot be understood in the nomological sense, for that would require strict psychophysical 8

I have in mind A Coherence Theory of Truth and Knowledge (in: D. Henrich (ed.) Kant oder Hegel, Stuttgart: Klett-Cotta 1983: 4 2 3 - 4 3 8 ) , The Myth of the Subjective (in: Bendikt, Burger (eds.), Bewußtsein, Sprache und die Kunst, Wien: Verlag der Osterreichischen Staatsdruckerei 1988: 45 — 54) and Epistemology Externalised (Dialectica 45 (1991): 191-202).

310

Peter Bieri

laws. Once again, we have here (this time at the input end) a connection between events described in non-cognitive terms and events described in cognitive terms. And, once again, this connection must be kept "loose". Is it correct to say, then, that the causal relations which generate (or are used in determining) mental content are "irreducibly causal" in the sense discussed earlier?

Donald Davidson: Reply to Peter Bieri I'm gratified to have the passages quoted by Peter Bieri taken so seriously. Unfortunately, much that I write will not stand such careful scrutiny. But I do take the theme o f these passages seriously, though perhaps I don't think they explain quite as much as Bieri is suggesting. Let me try to clarify the point I was trying to make. "Unavoidable mention o f causality is a cloak for ignorance": almost everything here needs explaining. By mention o f causality, I mean the use o f concepts (expressions) which cannot be understood without appeal to the notion o f causality. Examples are the concepts o f acting with an intention, o f favism, o f an infectious disease, o f an igneous rock. An intentional action is one caused (in the right way) by beliefs and evaluative attitudes; favism is a severe allergic reaction caused by eating the broad bean or inhaling its pollen; an infectious disease is either "capable o f causing infection" or "caused by a microorganism" or both; an igneous rock is a rock caused by solidification from a molten state. Use o f these concepts in an explanation always involves explicit or implicit appeal to a rough generalization, a less than precise "law"; if we can do no better than use such concepts, we admit we do not know a strict or precise law governing the event or state to be explained; or we may know such a law but not know enough to apply it. Causal concepts short-circuit part o f the explanatory process — a part that a strict law would fill in. T h e short circuit occurs because the explanation depends on characterizing something as the sort o f thing that has a certain cause or a certain effect (under the "right" conditions, other things being equal, etc.), thus sparing us the need to say what it is about the object or substance or state that explains why it acts as it does. When is the use o f a causal concept "unavoidable"? As a practical matter, whenever we cannot get along without it. But there is a sense in which we sometimes believe that we could eliminate causal concepts if we just knew more. Thus I suppose it possible that there are no meteorological or geological events that cannot in theory be explained by the physics we know; in theory we could describe these events (given knowledge o f particular facts more detailed than we ever expect to have) in the language o f pure physics. We could then apply the laws of physics to give explanations and predictions o f the weather or plate tectonics as precise as the

312

Donald Davidson

fundamental laws of physics allow. In this attenuated sense, causal concepts may be avoidable in the natural sciences. This would mean that all the concepts we use in the natural sciences could be expressed in the vocabulary of physics. The basic concepts of psychology are, I have argued, not reducible to those of physics, and therefore do not lend themselves to inclusion in a systematic, precise and closed system of description and laws. This is why, particularly when we want to talk of the relations between events described in psychological terms and events described in physical terms, we make necessary use of causal concepts; these give us the sort of forgiving conceptual connections and explanations we want. What then, are we "ignorant" about when it comes to explaining psychological events? We don't know precise laws for explaining and predicting them; but unlike the situation in the natural sciences, this isn't because we haven't discovered them yet; it's because there are no such laws. So I think it was misleading of me to suggest that the use of causal concepts in psychology was a cloak for ignorance. It only makes sense to speak of ignorance when there is something there to know. Now let me try to answer Bieri's three questions. 1) How, Bieri asks, can I reconcile my view that if one event is the cause of another then those events fall under a strict law with the claim that because mental concepts are irreducibly causal they cannot fall under strict laws? The answer, which I gave in "Mental Events", is that events "fall under a strict law" if there exist descriptions of the events, that is, ways of classifying them, such that those descriptions instantiate a strict law. But this clearly does not imply that all descriptions we may give of those events will instantiate a strict law, or any law at all. Mental concepts provide ways of characterizing or describing events which don't fall under strict laws. Causality relates events however described; laws relate events under specific descriptions. Bieri wrongly interprets me when he says, "in the case of mental concepts, the notion of cause serves to indicate that there cannot be strict laws". The notion of cause does no such thing; it is the fact that the concepts are causal concepts that makes those concepts ineligible for inclusion in strict laws. I certainly do not think there are two concepts of cause. There are, however, two concepts that are easily confused: that of causality, which is a relation between events, and that of causal explanation, which is a relation between propositions or sentences. The former is independent of our ways of describing and classifying events; the latter is totally dependent on it. Question 2 was: If mental concepts are irreducibly causal, what happens to the argument for token identity? Bieri's trouble lies, it seems to me, in a certain ambiguity in his concept of causality; some of the time he thinks it implies laws, some of the time not. It does imply laws in this sense: if

Reply to Peter Bieri

313

c and e are related as cause and effect, there exist descriptions of c and e that instantiate a strict law; it does not imply laws in this sense: if c caused e, this is true no matter how c and e are described. There is not, so far as I can see. any conflict between these claims. Bieri says, "If we follow Davidson's new explanation of the causal character of mental concepts, it would seem that the fact that mental events are conceived of as causes does not force us to postulate that they instantiate laws; on the contrary." Up to the semicolon this is right: there is nothing about the argument for the irreducibly causal character of mental concepts that forces us to say anything about laws at all. Where I disagree is with the "on the contrary", for the irreducibly causal character of mental concepts tells us nothing about whether other concepts are geared to laws. It is causal relations, not concepts, that imply the existence of laws. Question 3: Is it true that "the causal relations which generate (or are used in determining) mental content are 'irreducibly causal'"? The answer is once again that it is concepts that are or are not 'irreducibly causal', not causal relations. The question therefore makes no sense. f In response to a question concerning the relationship between causation and causal laws:] The trouble is due to a failure to take seriously the difference between events and their (various) descriptions. It makes no sense to say causation is a relation which "almost by definition cannot be a nomological relation". Laws deal with types of events and therefore the statement of a law requires particular classifications of events. Laws deal with entities only as they satisfy specific concepts. Causal relations, on the other hand, have nothing to do with how we describe or classify events, or the concepts under which we conceive them, since causal relations hold between events however we happen to describe or classify them. So for me it is simply wrong to say the "function" or "essence" of the concept of cause is somehow to shield causal relations from laws. Causal concepts don't sit well with strict causal laws because they enable us to evade providing strict laws; causal relations are another matter.

Evaluative Judgements THOMAS

SPITZLEY

Although the title of my paper is fairly general, the actual topic on which I shall concentrate is considerably more modest. I shall restrict myself to making some comments on Professor Davidson's very influential and important theory of evaluative judgements. My paper consists of three parts. In section I) I illustrate the notion of a prima facie judgement and the relevance of such judgements. In part II) I investigate in more detail a component of prima facie judgements, viz. the evidence component. Thereby I want to answer two questions: a) what does the evidence component consist of, and b) is somebody who holds a prima facie judgement committed to the truth of what is expressed in its evidence component? Section III) is devoted to some problems concerning the so called all things considered judgements, especially to the question of what the logical form of all things considered judgements is like. I. The Relevance of Prima Facie Judgements If we want to explain why someone (intentionally) performs a certain action a, e.g. why he keeps a promise, we can do this by saying that he assents to the moral principle "It is right to keep a promise" and believes that his action a is a keeping of a promise. This explanation can be formulated as a practical syllogism: (1)

It is right to keep a promise,

(2)

a is a keeping of a promise,

therefore (3) a is right. Traditionally (1) is taken to have the logical form of a simple universal generalisation, so that in formalizing (1) we would get (Γ)

(x) (x is a keeping of a promise —» χ is right).

Let us imagine further that the agent assents not only to the moral principle (1) but also to the moral principle

316

Thomas Spitzley

(4)

Lying is wrong

and that (5)

Λ

is a lie.

Then we can derive the conclusion (6)

a is wrong

from the premises (4) and (5). That means, however, that one and the same action, viz. Λ, is right as well as wrong, and so we are confronted with an explicit contradiction. From this result Davidson rightly concludes that it is wrong to formalize sentences which express moral principles as simple universal generalisations. Instead he proposes to understand moral principles and other pro attitudes like tastes, values, inclinations, or preferences as prima facie judgements. 1 The moral principle "It is right to keep a promise" is therefore not to be formalized as in (Γ), but rather as (7)

pf (x is right; χ is a keeping of a promise). 2

In (7) appears a prima facie operator which ... relates propositions. In logical grammar, 'prima facie' is not an operator on single sentences, much less on predicates of actions, but on pairs of sentences related as (expressing) moral judgement and ground. 3

(7) must be understood in the following way: instances of the open sentence after the semicolon express thoughts which, if true, can serve as reasons for judgements which are expressed by the corresponding instances of the open sentence before the semicolon. 4 So prima facie judgements are conditional judgements as opposed to unconditional judgements like e.g. (6). The decisive point here is that (7) together with (2)

a is a keeping of a promise

do not justify the passage to (3) 1

2

3 4

a is right.

Cf. D. Davidson, "Replies to Essays I - I X " (Replies), p. 202, in: B. Vermazen & Μ. B. Hintikka (eds.), "Essays on Davidson·. Actions & Events", Clarendon Press, Oxford 1985, pp. 1 9 5 - 2 2 9 . Cf. D. Davidson, "How is Weakness of the Will Possible?" (Weakness), p. 38, in his "Essays on Actions and Events" ( E A E ) , Clarendon Press, Oxford 1980, reprinted with corrections 1982, pp. 2 1 - 4 2 . Davidson, Weakness, p. 38; cf. Replies, p. 202 and p. 207. Cf. Davidson, Replies, pp. 202f.

Evaluative Judgements

317

Analogously, the moral principle that lying is wrong: (8) pf (x is wrong; χ is a lie), and the belief (5) a is a lie do not allow me to conclude: (6) a is wrong. This feature, as well as others, prima facie judgements have in common with probability judgements in analogy to which Davidson modelled the prima facie judgements. What one can derive from the premises (8) and (5) is, according to Davidson, (9) pf 0 is wrong; (8) and (5)). 5 And correspondingly it follows from (7) and (2) (10) pf {a is right; (7) and (2)). In (9) and (10) it becomes obvious that in this form of practical reasoning the conclusion is a complex sentence which contains its premises. (6), the conclusion of a classical practical syllogism, corresponds only to a component of (9), viz. to the evaluative component, and the premises (8) and (5), from which (9) is derived, too, constitute a component of (9), viz. the evidence component. (6), "a is wrong", gives us a clear recommendation of what to do or, in this case, rather what to refrain from. (9), however, does not at all express a recommendation on the basis of which one really should act, since it is perfectly compatible with a further judgement the agent might hold, viz. that due to some other of its properties the action a is right. Such a judgement is e.g. (10). Whereas in (9) the evaluative component is "