Colin McGinn-Philosophy of Language_ The Classics Explained-The MIT Press (2015).pdf

8/20/2019 Colin McGinn-Philosophy of Language_ The Classics Explained-The MIT Press (2015).pdf

http://slidepdf.com/reader/full/colin-mcginn-philosophy-of-language-the-classics-explained-the-mit-press-2015pdf 1/236



Philosophy of Language





The MIT Press

Cambridge, Massachusetts

London, England

Philosophy of Language

The Classics Explained

Colin McGinn



© 2015 Massachusetts Institute of Technology

All rights reserved. No part of this book may be reproduced in any form by any

electronic or mechanical means (including photocopying, recording, or information

storage and retrieval) without permission in writing from the publisher.

MIT Press books may be purchased at special quantity discounts for business or sales

promotional use. For information, please email [email protected].

This book was set in Stone by the MIT Press. Printed and bound in the United States

of America.

Library of Congress Cataloging-in-Publication Data

McGinn, Colin, 1950–.

Philosophy of language : the classics explained / Colin McGinn.

pages cm

Includes bibliographical references and index.

ISBN 978-0-262-02845-5 (hardcover : alk. paper)

1. Language and languages—Philosophy. 2. Language and languages—Philosophy—

Textbooks. I. Title.

P107.M38 2015

401—dc23

2014021824

10 9 8 7 6 5 4 3 2 1



Contents

Preface ix

1 Frege on Sense and Reference 1

1.1 Background 1

1.2 Identity 3

1.3 Additional Machinery 10

1.4 The Conception of Sense 12

1.5 Reference 16

1.6 Ordinary and Extraordinary Use 18

1.7 Further Points on Sense and Reference 20

1.8 Problems with Frege’s Theory 23

1.9 Extension of Frege’s Theory beyond Singular Terms 25

1.10 Further Aspects of Frege’s Theory 31

2 Kripke on Names 35

2.1 Background 35

2.2 Kripke’s Critique 39

2.3 Rigid Designation 42

2.4 Kripke’s Epistemic Objections 45

2.5 The Causal Chain Theory 48

2.6 Objections to Kripke’s Critique 49

2.7 The Social Character of Names 51

2.8 Essential Descriptions 52

2.9 Impure Descriptions 53

3 Russell on Definite Descriptions 55

3.1 Indefinite and Definite Descriptions 55

3.2 Three Theories of Definite Descriptions 60

3.3 Indefinite Descriptions and Identity 63

3.4 Russell’s Rejection of Meinong’s Ontology 65

3.5 The Details of Russell’s Theory of Descriptions 67



vi Contents

3.6 Problems with Russell 72

3.7 Primary and Secondary Occurrences 74

4 Donnellan’s Distinction 774.1 Introduction 77

4.2 Referential and Attributive Uses 78

4.3 Denoting and Referring 84

4.4 Truth-Value Gaps 85

4.5 Evaluating Donnellan’s Distinction 87

4.6 Implication and Implicature 90

4.7 Further Objections to Russell’s Theory 94

5 Kaplan on Demonstratives 975.1 Intension and Extension 97

5.2 Kaplan on Indexicals 100

5.3 The Two Principles of Indexicals 102

5.4 Context of Use and Conditions of Evaluation 105

5.5 Possible Worlds, Meaning, and Indexicals 109

5.6 Kaplan on “Today” and “Yesterday” 113

6 Evans on Understanding Demonstratives 115

6.1 The Fregean Theory of Indexicals 1156.2 The Point of Indexicality 118

6.3 Evans’s Theory of Sense and Reference for Indexicals 119

6.4 Saying versus Showing 122

6.5 Mock Sense 124

6.6 Empty Names 125

6.7 Evans’s View of Names 126

6.8 Evans on “Today” and “Yesterday” 128

6.9 Character, Content, and Information 130

7 Putnam on Semantic Externalism 133

7.1 Background 133

7.2 Twin Earth and “Water” 134

7.3 Meanings Are Not in the Head 135

7.4 Criticisms of Putnam 143

8 Tarski’s Theory of Truth 147

8.1 Background 147

8.2 Tarski’s Criteria of Acceptability 1498.3 Aristotle and the Redundancy Theory 151

8.4 Object Language and Metalanguage 155

8.5 How to Derive the T-Sentences 157

8.6 Satisfaction 159



Contents vii

9 Davidson’s Semantics for Natural Language 165

9.1 Background 165

9.2 The Merits of Tarski’s Theory as Applied to Meaning 168

9.3 Applying Tarski’s Theory to Natural Languages 175

9.4 Empirical Truth Theory 181

9.5 Criticisms of Davidson’s Theory 185

10 Grice’s Theory of Speaker Meaning 191

10.1 Background: Speakers and Sentences 191

10.2 Two Types of Meaning 193

10.3 What Is Speaker Meaning? 195

10.4 Consequences and Criticisms 199

Appendix: Kripke’s Puzzle about Belief 203

Notes 211

Index 215





Preface

This book is intended as a student text suitable for undergraduates taking

a typical philosophy of language course. But it takes an unusual form: it

undertakes to explain ten classic works in the field as clearly as I know

how. So it is not the typical general survey of issues, but instead focuses on

individual authors. It could also be used as an introductory text for gradu-

ate students with no background in philosophy of language. The book is

not geared specifically to students with a strong interest and background

in analytic philosophy; it aims to include students who may not even be

specializing in philosophy. The aim is to make difficult primary material

accessible to people who might otherwise struggle with it.

The book consists of ten chapters (plus an appendix), each of which

discusses in detail a single classic article. It is intended to be used in con-

junction with an anthology of classic texts. The anthology I have used is

Philosophy of Language: The Central Topics, edited by Susana Nuccetelli and

Gary Seahy (Roman & Littlefield, 2008). It could also be used in conjunc-

tion with A. P. Martinich’s The Philosophy of Language (Oxford UniversityPress, 2006), though the selection of articles is somewhat different in the

two books. I have found in teaching the subject that students need a thor-

ough, clear explanation of the classic texts, which by themselves they find

too difficult. Accordingly, the chapters in this book go through the classic

texts carefully and systematically. There is no attempt to give a general

survey of the literature, achieving complete coverage, and the book does

not deal with some of the more recent literature. The instructor would use

this book as a supplement to the original articles, sparing him or her a lot

of arduous exegesis.

I have generally included some evaluation and criticism of the views

and theories being expounded, but this is more to stimulate the student’s



x Preface

thought (and class discussion) than to contribute to the subject to the sat-

isfaction of my professional colleagues. I have always aimed to make the

material as simple as possible, without sacrificing accuracy. Everything isexplained from the ground up.

The book had an unusual gestation. It began when a student in my class

at the University of Miami, Colin Mayer, suggested that it would be useful

if there were a book offering the kinds of explanations I provided orally. I

agreed but was reluctant to write such a book myself, not wanting to give

up the time. He then suggested that he could transcribe the lectures from

recordings he had been making of my classes. We decided to give it a try. He

diligently set about the work. My task was to go over what he had writtenand revise it. I did that, finding it necessary to make many revisions (almost

every sentence). I did, however, try my best to preserve the original spoken-

word form of the lectures, thinking that this might make the material more

accessible. In pure writing there is always a tendency to value succinctness

and precision (not to mention elegance) over sheer comprehensibility. The

end result is a mix of informality and careful formulation. I am grateful to

Colin Mayer for undertaking this work, which could not have been easy,

and for his original suggestion.

I also had the assistance of Monica Morrison, who went over the raw

material of the transcriptions, cleaning up and formatting the text. All the

final text, however, is due to me. It was a much tougher job than I bar-

gained for, but I think the resulting book should be a boon for students

and teachers alike. I first taught philosophy of language some thirty-eight

years ago, and this is my distillation of many years of experience teaching

the subject. I hope it achieves its aim of conveying a rich body of thought

in an accessible form.

Colin McGinn

Miami, July 2012



1 Frege on Sense and Reference

1.1 Background

Before we begin to expound Frege’s views on sense and reference, a few

words about the general aims of the philosophy of language might be use-

ful. The most general thing we can say is that philosophy of language is

concerned with the general nature of meaning. But this is not very helpful

to the novice, so let us be more specific. Language is about the world—we

use it to communicate about things. So we must ask what this “aboutness”

is: what is it and how does it work? That is, how does language manage to

hook up with reality? How do we refer to things, and is referring to things

all that language does? Further, is referring determined by what is in the

mind of the referrer? If not, what else might determine reference? Some

parts of language we call “names,” but is everything in language a name?

How is a word referring to something connected to a person referring to

something? Do expressions like “Tom Jones,” “the father of Shakespeare,”

and “that dog” all refer in the same way? In what way do these types ofexpressions differ in meaning? How is a sentence related to its meaning?

Is the meaning the same as the sentence or is it something more abstract?

Can’t different sentences express the same meaning? What is a meaning?

Are meanings things at all? How is meaning related to truth? Whether what

we say is true depends on what we mean, so is meaning deeply connected to

truth? How are we to understand the concept of truth? What is the relation-

ship between what a sentence means and what a person means in uttering

a sentence? These are the questions typical of the philosophy of language.

In this book we will consider these questions by reviewing what the great-

est philosophers of language have said about them, beginning with perhaps

the greatest of them all—Gottlob Frege.



2 Chapter 1

Frege’s article “On Sense and Reference,” published in 1892, is the begin-

ning of modern philosophy of language, shaping the field ever since. We

shall therefore pay particularly close attention to its content, returning to itin later chapters. But before entering into detailed discussion of the article

it is important we gain some familiarity with two concepts: sentences and

propositions. A proposition is what is expressed by a sentence: the proposi-

tion expressed by a sentence constitutes the meaning of the sentence. Thus

it is possible for two different sentences to express the same proposition.

Two sentences that are synonymous with one another will express the same

proposition. Sentences can differ in their constituent words and be synony-

mous, having the same meaning, and thus express the same proposition.The following two sentences illustrate this point:

(1) John is a bachelor.

(2) John is an unmarried male.

The terms “bachelor” and “unmarried male” are synonymous, that is, they

have the same meaning; therefore, these two sentences express the same

proposition. Hence two different, nonidentical sentences of English can

express the same proposition. Two sentences from two different languages

can also express the same proposition. Here we have two synonymous sen-

tences of different languages, French and English:

(3) La neige est blanche.

(4) Snow is white.

Despite the fact that these two sentences are made up of different words in

two distinct languages, they still have the same meaning, and thus express

the same proposition.

With this understanding of how sentences relate to propositions, we can

now ask what a sentence is. A sentence is a collection of shapes, signs,

or acoustic signals. Different shapes of letters on paper or acoustic signals

in the air can correspond to the same proposition. Propositions, then, are

very different from sentences—more abstract than physical. A sentence is

the perceptible vehicle that expresses a proposition, and in addition can

be uttered by a person. When you utter a sentence like “Snow is white,”you thereby make a statement. A statement is a relationship between three

things: the speaker, the sentence, and the proposition. When a person

speaks he utters a particular sentence, and in so doing he makes a certain



Frege on Sense and Reference 3

statement. If a Frenchman utters the sentence “La neige est blanche,” he is

stating that snow is white, even though he did not utter that sentence of

English. However, since the sentence “La neige est blanche” is synonymouswith the English sentence “Snow is white,” the two different sentences

express the same proposition. A sentence in one language can be used to

report the same proposition expressed by a speaker who made a statement

using a different language. Sentences, statements, and propositions are sys-

tematically correlated, but they are not the same thing. A sentence is a

physical sequence, a statement is a human action, and a proposition is an

abstract meaning.

1.2 Identity

In “On Sense and Reference,” Frege is concerned with the relationship

between a sentence and the proposition it expresses. He is concerned with

discovering answers to the following questions: What exactly is the rela-

tionship between a sentence and the proposition that it expresses? When

is one proposition the same as another proposition expressed by a differ-

ent sentence? What constitutes a proposition? What is the meaning of a

word? The questions that concern Frege lead one to wonder how a sen-

tence, considered as an arrangement of shapes or a sequence of sounds, can

be meaningful. That is, we are concerned with sentences and their mean-

ings—how they are able to tell us things about the world. What kind of

thing is meaning?

Frege’s article discussing these questions is not straightforward—it con-

tains certain obscurities that are seldom if ever brought up by commenta-

tors, because they are difficult to interpret. In what follows, however, we

will bring out and clarify the obscurities in Frege’s article. First, let us exam-

ine the opening of “On Sense and Reference”:

Equality gives rise to challenging questions, which are not altogether easy to answer.

Is it a relation? A relation between objects, or between names or signs of objects? In

my Begriffsschrift , I assumed the latter.1

Though Frege is not explicit about what he means by “equality,” he is

using the term in a mathematical sense (not a social one!). The notion

of equality can be illustrated with a mathematical statement: “4 × 5 = 20.”

Contemporary philosophers use “identity” instead of “equality.” The exam-

ple “4 × 5 = 20” would be called an identity statement, asserting that the



4 Chapter 1

number 4 × 5 is identical to the number 20. It is these types of statements

that Frege intends when he uses “equality.”

Identity can also apply to other nonmathematical cases. There are afew things about identity that Frege does not mention. Philosophers often

distinguish between numerical identity and qualitative identity. Qualitative

identity occurs when two things are exactly alike. For example, two cars

that come from the same assembly line and have the same color and so

on would be said to be qualitatively identical. Frege, however, is primarily

interested in numerical identity. Numerical identity is the relationship a

thing has to itself. The relation is a very primitive and trivial one: every-

thing has the relation of identity to itself. Furthermore, numerical identitydoes not obtain between one object and another object, even if the two

objects are qualitatively identical. For example, two twins do not have the

relationship of numerical identity to one another—that relationship exists

only between one of the twins and himself.

We can now ponder the following: Is identity a relation? There are all

sorts of relations: left of, older than, belonging to a political party, or living

in a certain place. Each one of these examples illustrates a nontrivial rela-

tion, and therefore tells us something substantial about reality. However, in

the case of identity, it has been argued that the relation something has with

itself is trivial, and therefore gives no substantial information but provides

only a tautology. Frege continues his explanation of identity in the follow-

ing passage:

The reasons which seem to favor this are the following: a = a and a = b are obviously

statements of differing cognitive value; a = a holds a priori and, according to Kant, is

to be labeled analytic, while statements of the form a = b often contain very valuable

extensions of our knowledge and cannot always be established a priori. The discovery

that the rising sun is not new every morning, but always the same, was one of the

most fertile astronomical discoveries. Even today the identification of a small planet

or a comet is not always a matter of course.2

In the above Frege is concerned with statements that identify objects. An

identity statement using different names will have this form: “a = b” (“a

is identical to b”). There is one object that we have referred to with two

names, “a” and “b.” For illustration, let “a” be “4 × 5” and “b” be “20.”

We have referred to the object, a number, with the numeral “20,” as well

as with the expression “4 × 5,” and now we form a corresponding identity

statement. Two names that refer to the same thing create a true identity




statement when they are written down and have the symbol “=” between

them. On the other hand, if “a” does not denote something identical to

what “b” denotes, then we will produce a false identity statement.The essence of Frege’s point here is that when he wrote the Begriffss-

chrift , he thought that when we make a statement like “a = b” the relation

expressed by “=” is a relation between the names themselves. In this case,

the statement really is about the names “a” and “b” and not the objects to

which the names “a” and “b” refer. The names of the objects are separate

from the objects that they designate. During his Begriffsschrift days, Frege

thought that when he made an identity statement he was concerned with

the names that were in that statement. This is because the alternative viewseems to lead to absurdity:

Now if we were to regard equality as a relation between that which the names ‘a’ and

‘b’ designate, it would seem that a = b could not differ from a = a (i.e. provided a = b

is true). A relation would thereby be expressed of a thing to itself, and indeed one in

which each thing stands to itself but to no other thing.3

It would seem that taking “=” to relate objects, not names, would make “a

= b” express the same proposition as “a = a.” To illustrate this point more

clearly, we can use the example of the two names, “Hesperus” and “Phos-

phorous.” Venus is the planet that first comes up in the evening, and used

to be called “Hesperus.” “Hesperus” is a proper name denoting Venus; it

corresponds to the definite description “the evening star” (we discuss defi-

nite descriptions in more detail in chapter 3). In using the name “Hespe-

rus” we thus refer to Venus. Understanding advances in modern astronomy

that the ancients did not, we know that “Hesperus” refers to Venus. The

ancients, however, knew neither the name “Venus,” nor that Venus is aplanet and not a star. The same heavenly body is also seen in the morn-

ing—when viewed in the morning the ancients called it “Phosphorous, the

bringer of light.” Frege points out that the two different acts of naming

in fact correspond to the same object. In the example, the two different

names “Hesperus” and “Phosphorous” in fact correspond to the same heav-

enly body—Venus. It appears at one point in the sky in the evening and at

another point in the sky in the morning. The ancients did not know that

they were applying two names to the same planet. We can then say thatHesperus is identical to Phosphorous, stating a substantial astronomical

discovery. The ancient Babylonians were not able to assert that Hesperus is



6 Chapter 1

identical to Phosphorous, nor did they have any reason to think that. They

were ignorant of the identity.

The example of Hesperus and Phosphorous is a general illustration of thefollowing point: there are many cases where a single object has been given a

name, and then in another time and another context, given another name,

without anyone realizing that the object has been named twice. When the

identity is discovered, what the observer has learned, intuitively, is that one

thing has two appearances, and therefore that a = b. Therefore, the two differ-

ent appearances correspond to the same object, thus producing substantial

identity knowledge. In such a case “a = b” forms an informative identity

statement. We have expressed a proposition that is not trivial and gives usgenuine knowledge about reality. By contrast, an identity statement of the

form “a = a” (“Hesperus is Hesperus”) is not an informative proposition—it

is merely a tautology. The numerical identity—any numerical identity—

can be seen to hold without any empirical observations about the world

at all. In the case of Hesperus, if someone merely heard the name “Hespe-

rus,” she could know without observation that the statement “Hesperus

is Hesperus” is true. It is not possible to do the same with the statement

“Hesperus is Phosphorus.” This statement is informative, whereas the pre-

vious statement is not. Thus “Hesperus is Phosphorus” has empirical con-

tent and is synthetic (from Kant); but “Hesperus is Hesperus” is analytic, or

tautological, and is true simply in virtue of its meaning. To sum up, “a = a”

expresses an analytic, a priori proposition; “a = b” expresses a synthetic, a

posteriori proposition.

In the passages above from “On Sense and Reference,” Frege explains

how these two propositions (expressed by “a = a” and “a = b”) are com-

pletely different. For example, there could have been a time in the past

when people thought that every morning, a different fiery heavenly body

appears in the sky. Understanding that that heavenly body—the sun—is

the same one that appears in the sky every morning is a substantial empir-

ical discovery. We know that it has the same appearance, but sameness

in appearance does not entail that it is the very same object. But Frege

raises the following question: if equality is a relationship between an object

and itself, how could there be any difference between the propositionsexpressed by “a = b” and “a = a”? Would they not both be saying the same

thing, namely, that an object is identical to itself? In other words, wouldn’t

“a = b” express the same thing as “a = a”? So isn’t it better to suppose that




identity is really a relation between the names themselves, because they are

clearly different?

The sentence “a = a” expresses the proposition that a is identical to itself,and the statement “a is identical to itself” is analytic and a priori. However,

there is no way to argue that the statement “a = b” gives us the same propo-

sition as “a = a.” As we said, merely knowing a name allows one to say that

the object named is identical to itself. Even the ancients knew that Hespe-

rus is identical to Hesperus and that Phosphorous is identical to Phospho-

rous. What they did not know is that Hesperus is identical to Phosphorous.

Making the assumption that identity is a relation between an object and

itself appears to lead to paradox when thinking about identity propositions.Frege therefore thought when writing the Begriffsschrift that identity could

not be a relation between an object and itself. To avoid the paradox, the

two different sentences must state different propositions—but how?

If identity is a relation between names and not objects, then something

different is being stated in the two cases. Thus “a = a” informs us that the

name “a” denotes the same thing as the name “a.” On the other hand, “a

= b” informs us that the name “a” denotes the same thing as the name “b.”

Here we are no longer concerned with the objects themselves but with the

names of them. If we are really talking about the names, then we can see

how the two sentences produce different propositions. Why? Because “a =

a” contains the name “a” and only the name “a,” whereas “a = b” contains

the name “a” and the name “b.” The second sentence accordingly refers to

something the first one does not refer to, namely the name “b.” It contains

the name “b” and in this analysis the sentence refers to that name. This

explanation shows how these two sentences can express different proposi-

tions: they are about different things, because they are really about names,

not objects. The latter proposition is about the names “a” and “b” whereas

the former is only about the name “a.” This way of thinking is a natural

way to think about identity statements: an identity statement says that

one name denotes the same thing as another name—not that one object is

identical to itself.

It is not generally the case that sentences containing names are about

those names. In fact, sometimes statements have nothing to do with namesat all. Consider a statement where someone says, “Hesperus is bright”—

here he does not appear to be talking about the name “Hesperus.” Rather,

he is talking about a planet, which is Venus, and stating that it is bright.



8 Chapter 1

He is not saying the name “Hesperus” is bright. It is, of course, still possible

that the name “Hesperus” is bright (e.g., the name “Hesperus” is written as

a neon sign). However, in general, if someone says, “Hesperus is bright,” heis not saying that the name “Hesperus” is bright. We are not generally talk-

ing about our words, but using them to talk about something else.

Notice that there is a huge difference between a name occurring in

an ordinary statement where it refers to its bearer, and a name occurring

in quotation marks in a statement when it refers to the name. Generally

speaking, statements that use a name do not refer to that name. Therefore,

making the claim that an identity statement like “Hesperus is identical to

Phosphorous” refers to the names is to say something quite revisionaryabout that sentence. In actuality, the speaker intends for that statement to

refer to the planet Venus, and does not intend for that sentence to refer to

names of the object at all. This is sometimes called the use–mention distinc-

tion: we use the name to mention an object; we don’t use the name to men-

tion itself—except when we expressly want to talk about words, not things.

Looking back on his view in the Begriffsschrift , Frege now thinks he was

wrong to take the view that identity is a relation between names. He illus-

trates this point in the following passage:

What is intended to be said by a = b seems to be that the signs or names ‘a’ and ‘b’

designate the same thing, so that those signs themselves would be under discussion;

a relation between them would be asserted. But this relation would hold between the

names or signs only in so far as they named or designated something. It would be

mediated by the connection of each of the two signs with the same designated thing.

But this is arbitrary. Nobody can be forbidden to use any arbitrarily producible event

or object as a sign for something. In that case the sentence a = b would no longer

refer to the subject matter, but only to its mode of designation; we would express noproper knowledge by its means. But in many cases this is just what we want to do.

4

Frege had tried to avoid the problem in supposing that identity is a relation

between an object and itself because that would make identity proposi-

tions trivial. Bringing in the names themselves was intended to solve this

problem. The phrase “mode of designation” in the above passage is meant

to include the names themselves. But then the statement would refer to

a mode of designation, not to a state of affairs in the world. The mode of

designation then becomes what he here calls the “subject matter” of the

statement. Frege now finds this objectionable, because we would not be

expressing what he calls “proper knowledge.” The reader will wonder what




Frege means by the phrase “proper knowledge.” To learn that Hesperus is

Phosphorous is to learn something substantial, empirical, and a posteriori.

But what proposition have we learned? It is clearly not the proposition thata is identical to a. The proposition instead states that the name “a” denotes

the same thing as the name “b,” according to the earlier theory. However,

Frege raises the objection that knowing that one name co-denotes with

another is not enough to acquire “proper knowledge.” If we suppose that

proper knowledge is knowledge that goes beyond a tautology, does the

knowledge that “a” co-denotes with “b” go beyond tautology? Contrary to

what Frege implies, it can be informative to learn that one name refers to

the same thing as another name—very informative. It would be impossibleto possess this knowledge ahead of time by just knowing the names inde-

pendently. Through knowing the name “Hesperus,” one also knows that

Hesperus is identical to Hesperus. However, to discover that in addition

to this the name “Hesperus” denotes the same thing as the name “Phos-

phorous” is to learn something previously unknown. Effectively, we have

learned that two different symbols denote the same thing. Isn’t this “proper

knowledge”? It certainly isn’t a tautology.

But Frege is suggesting that learning that Hesperus is Phosphorous is not

only learning a linguistic fact but also understanding something significant

about reality and the objects in the world. This statement has revealed a

genuine empirical fact about two heavenly bodies. Frege’s earlier theory

does not capture the fact that the person who comes to know the statement

has learned something about the world. It reduces the fact learned to a

merely linguistic fact, but the fact learned is not merely linguistic in nature.

What one learns is not merely that the names have the same reference, but

that two appearances correspond to the same object. The object of one’s

knowledge, then, is not the same as that of someone who learns that one

name refers to the same thing as another name. That would be learning

something about two names, not two appearances. The real knowledge in

the sentence “Hesperus is Phosphorus” comes from understanding some-

thing empirical about reality, not just something about language. Frege’s

idea of “proper knowledge” is knowledge of the world, and not merely

linguistic knowledge. Thus he rejects the linguistic theory of the contentof identity statements, as well as the simple object theory—the theory that

identity statements are only about objects, not linguistic items.



10 Chapter 1

1.3 Additional Machinery

To capture what is grasped when someone learns that “a = b” is true, weneed another analysis of the proposition expressed by that statement. So

far, we have seen two propositions that “a = b” might express:

(5) a = a (the object is identical to itself).

(6) “a” denotes the same thing as “b.”

Of course, these two things are both things one can know, but they are not

what one learns from the proposition expressed by the sentence “a = b.” It

may seem that we have exhausted all the possibilities on this matter. If so,

this leads to a huge logical problem, because it means that we cannot even

explain such simple identity statements as “2 +2 = 4.” This logical problem

is why Frege is faced with the task of trying to account for something that

seemingly cannot be accounted for.

The purpose of “On Sense and Reference” is to bring in extra machinery

to account for the meaning of “a = b” beyond what we have talked about

so far:

If the sign ‘a’ is distinguished from the sign ‘b’ only as object (here, by means of its

shape), not as sign (i.e. not by the manner in which it designates something), the

cognitive value of a = a becomes essentially equal to that of a = b, provided a = b is

true. A difference can arise only if the difference between the signs corresponds to a

difference in the mode of presentation of that which is designated.5

Frege here introduces the notion of a “mode of presentation” without

much fanfare or explanation, and he contrasts it with a “mode of designa-

tion.” For Frege, the mode of presentation is what is essential to the mean-ings of the names “a” and “b,” the modes of designation—where the mode

of designation is simply the name considered as a sign. What is needed in

this account is a mode of presentation associated with the objects where

that mode is not to be identified with the objects themselves or with their

names. Frege states:

Let a, b, c be the lines connecting the vertices of a triangle with the midpoints of the

opposite sides. The point of intersection of a and b is then the same as the point of

intersection of b and c. So we have different designations for the same point, and

these names (‘point of intersection of a and b,’ ‘point of intersection of b and c ’)

likewise indicate the mode of presentation; and hence the statement contains actual

knowledge.6




This is a mathematical example, but we can think of other examples, by

returning to the evening star and the morning star, that illustrate the point

more clearly. The description “the evening star” refers to the same thing as“the morning star,” because those things are just Hesperus and Phospho-

rous, respectively. There are many instances of this same possibility, where

two descriptions pick out the same object. It need not be obvious to people

that these descriptions do refer to the same thing. All Frege wants his read-

ers to understand through the example is that two descriptions can refer to

the same thing—the intersection of these two lines and the intersection of

these other two lines are the same point.

The reader would naturally infer at this point that the mode of presenta-tion is connected to perception—it is the mode in which something percep-

tually appears, such that two different modes of presentation of something

correlate with different perceptual appearances. It is natural to assume

that two different ways an object is presented to somebody could produce

two entirely different appearances of that object to that person. A famous

example is of a mountain where someone approaches it from the east and

upon seeing it calls it Atlan. The same explorer approaches the very same

mountain from the west and calls it Athla. Of course, our explorer eventu-

ally discovers that he approached the same mountain twice, but from dif-

ferent perspectives. All of these examples illustrate the same point as Frege’s

triangle intersection example.

In addition to a name and its bearer, then, Frege has added the mode

of presentation of the bearer to somebody who uses the name. This brings

in additional machinery—some mode of presentation of both a and b. Let

“a” be associated with the mode of presentation MP1 and let “b” be associ-

ated with the mode of presentation MP2. Frege is arguing, in effect, that

if “a = b” is true the statement tells us truly that MP1 presents the same

object as MP2. Here the modes of presentation have replaced the names.

So understood, names are words with associated modes of presentation.

Now we see the difference between “a = a” and “a = b.” In “a = a” there is

only one mode of presentation, MP1, making that statement trivial. In “a

= b” there are two modes of presentation, MP1 and MP2, thereby creating

a nontrivial statement. It is nontrivial to find out that a single object hasthese two different modes of presentation. Hence, Frege’s solution to the

problem of identity statements is to bring in modes of presentation as the

missing ingredient.



12 Chapter 1

1.4 The Conception of Sense

The last sentence from the passage quoted above illustrates Frege’s view ofwhat he calls “actual knowledge.” We have already discussed how actual

knowledge is knowledge of the nonlinguistic world. It is not the names as

such that are important in this case, but the references of the names and

how they can appear or be “presented.” He continues:

It is natural, now, to think of there being connected with a sign (name, combina-

tion of words, letter), besides that to which the sign refers, which may be called the

reference of the sign, also what I should like to call the sense of the sign, wherein the

mode of presentation is contained. In our example, accordingly, the reference of theexpressions ‘the point of intersection of a and b’ and ‘the point of intersection of b

and c ’ would be the same, but not their senses. The reference of ‘evening star’ would

be the same as that of ‘morning star,’ but not the sense.7

In addition to the term “mode of presentation,” Frege has now introduced

another piece of theoretical machinery, the “sense.” He has so far explained

the sense as connected to the mode of presentation of the reference. Thus

in “a = b” the names “a” and “b” have the same reference but not the

same sense. To account for the proposition expressed by a sentence it is notenough to look at the sentence itself or at the reference of the words in the

sentence. To account for the proposition expressed by a sentence, another

level of semantic reality must be recognized—that of the sense. So, in addi-

tion to the reference of an expression in a language, the expression also has

a sense.

At this point Frege has established to his satisfaction that the meaning

of a name cannot be explained purely by its reference. Instead, the name

must be assigned a particular mode of presentation of its reference, and

the mode of presentation of the reference shows the true definition of the

name. Although the name refers to an object in the world, the real meaning

of the name comes not from what it refers to but from the mode of pre-

sentation. Therefore, Frege has shown us that a theory of language cannot

have only reference—it must have sense over and above reference.

So far, the word “sense” is merely a label. Frege has introduced this ter-

minology so that there is a mechanism to differentiate the various names,

since we have shown that it can be neither the reference nor the names

themselves that play this role. The sense, then, accounts for the cognitive

differences in names. But what is a sense? Frege uses the phrase “mode of




presentation,” and given his example of the triangle, it is natural to suppose

that the mode of presentation is a perceptual or psychological notion. Of

course, it is possible to see an object from different angles and perspectivesand not realize it is the same object you are seeing. The idea of sense can

be generalized beyond what we have talked about with the examples of

Hesperus and Phosphorous, or Frege’s own example of the triangle. But in

our examples and his it seems as though sense has something to do with

perceptual perspective—way of seeing. Notice in the previous passage that

Frege is not saying that the sense is identical to the mode of presentation;

rather, he says that the sense contains the mode of presentation. Strictly

speaking, then, Frege has introduced two extra levels of meaning: sense and mode of presentation, where the former “contains” the latter.

Not every expression in language that designates an object would natu-

rally be considered a proper name. A proper name is normally considered

an ordinary name, such as “Charles Dickens.” However, Frege also includes

other expressions under the heading of proper name that are generally not

called proper names. For instance, “the president of the United States in

2012” is said by Frege to be a proper name because it designates a particular

person, Barack Obama. Usually, such expressions are called definite descrip-

tions; however, Frege considers definite descriptions to be proper names.

Consequently, he thinks that both proper names and definite descriptions

have a sense and reference. In chapter 3, we will see that Bertrand Russell

argues that definite descriptions are not proper names at all, and that logi-

cally proper names are completely different from definite descriptions. In

his essay, however, Frege assumes that proper names and definite descrip-

tions are logically the same.

Frege’s main point is that every expression in either of these two catego-

ries—ordinary proper names and definite descriptions—has both a sense

and a reference. Further, it is the sense that contains informative value for

identity statements containing those proper names. Frege outlines this idea

in the following passage:

It is clear from the context that by ‘sign’ and ‘name’ I have here understood any des-

ignation representing a proper name, which thus has as its reference a definite object

(this word taken in the widest range), but not a concept or a relation, which shall bediscussed further in another article. The designation of a single object can also con-

sist of several words or other signs. For brevity, let every such designation be called a

proper name. The sense of a proper name is grasped by everybody who is sufficiently



14 Chapter 1

familiar with the language or totality of designations to which it belongs; but this

serves to illuminate only a single aspect of the reference, supposing it to have one.

Comprehensive knowledge of the reference is not to be obtained.8

Here Frege attends to the fact that people who understand a language will

grasp the senses of the names in that language. Hence the connection

between sense and understanding—one who grasps the sense will under-

stand the meaning of the names in the language.

A close scrutiny of the paragraph just cited will help us to figure out the

exact meaning of the term “sense.” There is a vital clue to the meaning of

“sense” when Frege states that the sense is something that “illuminates

only a single aspect of the reference.” From this, we can deduce that a senseis akin to a single aspect of an object. For it is natural up until this point

for the reader to assume that senses are something like concepts or ideas

in people’s minds. However, the above passage illustrates Frege’s rejection

of the idea that senses are anything mental. If the sense is an aspect of an

object, then it cannot be something in the person’s mind who understands

the expression—it is a part of the object, not the individual cognizing it.

Another way of interpreting this “aspect of an object” is viewing the

sense as a certain property an object has. For example, one of the proper-

ties of the moon is that it is arid. Obviously, objects have many differ-

ent properties, and different expressions can latch on to each one of those

properties as distinct from others. The sense, then, consists in latching on

to a particular property of the given object. As stated in the above pas-

sage, the mode of presentation is an aspect of an object. Those aspects will

exist regardless of whether anyone is there to know them, perceive them,

or apprehend them; objects have these properties, these aspects, indepen-

dently of human minds.

It is important at this point to note a natural interpretation of sense that

is flawed. Take the example of the definite description “the president of the

United States.” The reference of this definite description is a certain object

with various properties. Each of those properties that the object has is (or

corresponds to) a potential sense. In the case of this definite description,

one of these properties is an actual sense, because we have an expression

in our language that expresses that property—“the president of the UnitedStates.” That would seem to be the notion of sense that Frege has expressed

so far. However, there is a hole in this seemingly natural interpretation.

Since we know that the sense serves to illuminate this single aspect of the




reference, is it correct to suppose that the sense is an aspect of the refer-

ence? No, because a thing that illuminates an aspect is not identical to that

aspect. There is a distinction between the sense, the illuminator, and thething illuminated, the aspect. The thing illuminated is an aspect of the

object, a property. The sense is not identical to the aspect, though it is

closely related to it. The purpose of the sense is to illuminate the aspect; it

expresses it or contains it. To say that they are identical would be to ignore

a vital point in the above passage.

This distinction is a significant one for our purposes—if the sense were

identical to the aspect and the aspect is not itself representational, then it

follows that the sense is not representational. On the other hand, if thesense illuminates the aspect without being identical to it, then it can be a

representational entity. With this interpretation, sense becomes something

that represents an aspect of something. It is highly likely that this inter-

pretation of sense is the one Frege was going for—sense is something that

represents an aspect of an object. If we are trying to analyze an expression

like “the president of the United States,” we thus have four levels to exam-

ine: (i) the linguistic expression, (ii) the sense that illuminates the aspect,

(iii) the aspect illuminated by the sense, and (iv) the reference, an object.

In fact, very strictly, we might identify five levels in Frege’s theory, because

there is also the notion of a mode of presentation, which is contained in

a sense without being identical to a sense, and which serves to present an

aspect of the reference. The name expresses the sense, which contains the

mode of presentation, which illuminates the aspect, which is possessed by

the object of reference.

Several questions arise concerning the possibility of a regress of expla-

nation in trying to understand how reference works. If we think of sense

as referring to an aspect, then the idea of referring is presupposed by the

theory rather than explained. It matters whether or not we think that the

sense represents something because representation is a form of reference.

We must give a theory of reference to aspects before we can understand ref-

erence to objects. If the relationship between the sense and the aspect is one

of representation, we may question whether the relationship of reference

here is mediated by a further sense that presents the aspect. If the sense andthe aspect were related in representation, it would appear that this relation

would cause a regress. There is now something that lies between the sense

and the aspect—the mode of presentation of the aspect, that is, an aspect



16 Chapter 1

of an aspect. The possibility of regress raises an uncomfortable question for

Frege: is the sense to be taken as an aspect or something that represents an

aspect? Neither possibility appears satisfactory. If it is neither, then what isit exactly?

We saw in the previous passage that the expression illuminates a single

aspect of the reference but it does not illuminate every aspect of the refer-

ence. This is crucial to the whole picture Frege is painting, because a given

object can have several aspects and two proper names can latch on to these

different aspects. Therefore, when they are put together in an identity state-

ment, the statement becomes informative. If we knew every aspect of every

object, we would not gain information with identity statements, becausewe would already know everything. For example, we would know that the

evening star is the morning star. But because we do not know a given object

in all of its aspects we are in a position to be informed of something when

we are told that a = b. I can know one thing about an object without know-

ing everything about it.

1.5 Reference

We should examine the following passage to aid in the discussion of the

relationship between signs, senses, and references:

The regular connection between a sign, its sense, and its reference is of such a kind

that to the sign there corresponds a definite sense and to that in turn a definite

reference, while to a given reference (an object) there does not belong only a single

sign. The same sense has different expressions in different languages or even in the

same language. To be sure, exceptions to this regular behavior occur. To every ex-

pression belonging to a complete totality of signs, there should certainly correspond

a definite sense; but natural languages often do not satisfy this condition, and one

must be content if the same word has the same sense in the same context. It may

perhaps be granted that every grammatically well-formed expression representing

a proper name always has a sense. But this is not to say that to the sense there also

corresponds a reference.9

The relationship as explained above is rather fluid—the very same sense can

be expressed by two different signs, as in the case of synonymy. Synonymy

can exist within a language or across different languages. For example, Eng-lish speakers would say “snow” and French speakers would say “neige.” Fur-

ther, because of ambiguity it is possible to have one sign that corresponds

to two different senses—“bank” could mean a bank of a river or a bank for




money. Ordinary proper names, such as “Bob,” in our language have a simi-

lar problem of ambiguity since many people have the same name. The same

name has many different senses depending on whom or what it names.Concerning reference, Frege believes that a single reference can have

many senses corresponding to it and can have many signs corresponding to

it. However, there cannot be one sense that corresponds to several different

things, since a sense uniquely determines its reference. In Frege’s system,

the reference does not determine the sense, because there can be many dif-

ferent senses for the same reference. In contrast, the sense does determine

the reference, because the same sense cannot fix two different references.

A sense must always have one specific reference to which it corresponds.Therefore, the determination goes from sense to reference but not con-

versely. Furthermore, there is no determination from the sign to the sense.

Although every expression should have a definite sense, it is possible

for expressions not to have senses. For example, someone could make up

words like “fedneep” that are nonsense—they are signs that lack a sense.

However, to make a statement with meaning Frege states that the sign

should have a sense:

The words ‘the celestial body most distant from the Earth’ have a sense, but it is very

doubtful if they also have a reference. The expression ‘the least rapidly convergent

series’ has a sense; but it is known to have no reference, since for every given conver-

gent series, another convergent, but less rapidly convergent, series can be found. In

grasping a sense, one is not certainly assured of a reference.10

The general point may be lost to the reader since Frege’s examples are rather

technical. Only astronomers would understand the former example, and

mathematicians the latter. The general idea underlying the examples is that

you can form definite descriptions that do not refer to anything. Take the

following example of a definite description: “the polka dotted president

of the United States.” There has never been a polka dotted President of

the United States, so descriptions like those do not refer to anything at all.

There is a reason why descriptions such as “the polka dotted president of

the United States” must have a sense even though they do not have refer-

ence. For us to be able to construct meaningful and true statements such

as “the polka dotted president of the United States does not exist,” thedefinite description itself must be meaningful. This is just one example,

but there are infinitely many definite descriptions that have sense and are

therefore meaningful, but lack reference. Therefore, it is possible to have



18 Chapter 1

sense without reference, and to form proper names that have sense but no

reference.

1.6 Ordinary and Extraordinary Use

Frege applies his discussion of sense, signs, and reference to the ordinary

use of words in our language, but not only to this:

If words are used in the ordinary way, what one intends to speak of is their reference.

It can also happen, however, that one wishes to talk about the words themselves or

their sense. This happens, for instance, when the words of another are quoted. One’s

own words then first designate words of the other speaker, and only the latter havetheir usual reference. We then have signs of signs. In writing, the words are in this

case enclosed in quotation marks. Accordingly, a word standing between quotation

marks must not be taken as having its ordinary reference.11

If words are used in an ordinary way, then in using a word one is intend-

ing to speak of the object the word refers to. For example, when someone

uses the words “Barack Obama,” he will usually intend to speak of Barack

Obama, and therefore Barack Obama is his reference. However, words are

not always used in an ordinary way. Therefore, it is not in every case that weare speaking of the reference of a word. It is also possible that one can talk

about only the words themselves. Likewise, one can talk about the sense of

a word. For example, “the sense of ‘Barack Obama’” refers to the sense of

that name, not to its reference. Take caution when parsing these sorts of

sentences. For example, if one writes “the sense of Barack Obama” instead

of “the sense of ‘Barack Obama’” one has confused the sense of a human

being (whatever that may be) in the first case with the sense of a name in

the second case. Barack Obama does not have a sense, because he is a per-

son, not a piece of language. Quotations give us a device to prevent us from

falling into such a logical error. When writing about the sense of an expres-

sion as opposed to the reference of an expression, quotation marks can be

used to form the appropriate expression. Therefore, when talking about

signs and the sense of signs we must be careful about our use of quotation

marks so that what we say makes sense.

Further, when reporting what someone else has said, words do not have

their usual reference. In this case, the quoted words are signs of signs. Most

of the time words are signs of objects, but in the case of quoting the words

of another person the quoted words become signs within signs. Therefore,




“‘Barack Obama’” is a sign of a sign. Let us look at two examples to further

illustrate these points:

(7) The word man

(8) The word “man”

The second example is correctly expressed because the quotation marks

show that it is a word that is referred to. In the first example without quota-

tion, “man” refers to a certain species or gender, not to the word itself. In

spoken language, we can use such techniques as intonation of voice, body

language, or say “quote” and “unquote.” Frege thought that ordinary natu-

ral language was quite defective in this way and that it should be clearerwhen one attempts to talk about words themselves and not what they are

about.

There are many other places in “On Sense and Reference” where Frege

attempts to deal with how words function in normal and abnormal speech.

He writes:

In order to speak of the sense of an expression ‘A’ one may simply use the phrase ‘the

sense of the expression “A”’. In reported speech one talks about the sense, e.g., of

another person’s remarks. It is quite clear that in this way of speaking words do not

have their customary reference but designate what is usually their sense. In order to

have a short expression, we will say: In reported speech, words are used indirectly or

have their indirect reference. We distinguish accordingly the customary from the in-

direct reference of a word; and its customary sense from its indirect sense. The indirect

reference of a word is accordingly its customary sense. Such exceptions must always

be borne in mind if the mode of connection between sign, sense, and reference in

particular cases is to be correctly understood.12

Consider someone who says, “John said that Barack Obama is great.” Noticehere that “that” has been inserted in the sentence with no quotation marks

at all. This example illustrates indirect speech. Someone also could have

said, “John said, ‘Barack Obama is great,’” and it would have served much

the same purpose. But it may be that John, contrary to the latter statement,

is not an English speaker. For example, John could have uttered an Italian

sentence, “Barack Obama e meraviglioso” (translation: “Barack Obama is

wonderful”). An English speaker would take the Italian words and trans-

late them into an English sentence, thus forming a statement of indirect

speech. Frege thinks that in indirect speech the expressions that follow a

word like “that” do not have their ordinary reference. Instead, these words

refer in that context to their ordinary sense not their ordinary reference.



20 Chapter 1

To give you a better sense of what Frege has in mind, let us take an

example of someone who utters a sentence containing an expression that

has no reference. Suppose John says, “The polka dotted president of theUnited States is great.” In this case, that statement has no reference, and we

have reported the sentence in the direct speech form. However, if we put

it into the indirect speech form, then we might be taken to suppose that

there is such a thing as a polka dotted president, contrary to our intentions.

If the definite description were taken to refer to its normal reference, then

that part of the sentence would have no reference at all. Furthermore, if

the part of the sentence had no reference, something true could not have

been said. To avoid these consequences, Frege thinks that we refer insteadto the customary sense of the expression and use it abnormally in that

particular context. Since the customary sense exists, there is no part of that

sentence that lacks a reference. Paraphrasing the idea into explicit form,

what is really being said when someone says “John said that Barack Obama

is great” is “John said something expressing the proposition that Barack

Obama is great.” It is almost as though the individual who utters these

words is talking directly about the sense that someone’s words have and not

the reference of what he says. When we are reporting what someone said,

the interest does not lie in whether or not what the person said was true or

really achieved objective reference. Rather, the interest lies in the content

of what the person said, and therefore in the sense of the words he used.

In this complex sentence, there is no reference to Barack Obama at all. The

only thing that is referred to is the sense of the name “Barack Obama.” This

solves the potential puzzle of reporting a thing that a speaker says that may

not refer to any real object. So there may not be a reference for “the polka

dotted president,” but there is a sense of that expression, and that is what

matters in reporting the content of what someone said.

1.7 Further Points on Sense and Reference

It is wrong to suppose that words can be used only to talk about their cus-

tomary references. We have seen how it is possible to talk about words, and

the sense of words, without talking about the reference of those words.Concerning this point, Frege states the following:

The reference and sense of a sign are to be distinguished from the associated idea. If

the reference of a sign is an object perceivable by the senses, my idea of it is an in-




ternal image, arising from memories of sense impressions which I have had and acts,

both internal and external, which I have performed. Such an idea is often saturated

with feeling; the clarity of its separate parts varies and oscillates. The same sense is

not always connected, even in the same man, with the same idea. The idea is subjec-

tive: one man’s idea is not that of another. There result, as a matter of course, a vari-

ety of differences in the ideas associated with the same sense. A painter, a horseman,

and a zoologist will probably connect different ideas with the name ‘Bucephalus’.

This constitutes an essential distinction between the idea and the sign’s sense, which

may be the common property of many and therefore is not a part of a mode of the

individual mind. For one can hardly deny that mankind has a common store of

thoughts which is transmitted from one generation to another.13

In this passage, Frege sharply distinguishes between ideas present in peo-ple’s minds and the sense and reference of words. To reiterate a point made

above, Frege does not think that the ideas present in people’s minds have

anything essentially to do with sense and reference at all. A psychological

idea may be necessary for a human being to grasp a sense, but that does not

mean that the sense is the same thing as the idea.

First, depending on who you are, a certain word will bring different ideas

to mind. For example, an equestrian will have a different idea come to

mind when he hears the word “horse” uttered than when a zoologist hears

the same word. Frege thinks that the sense of the word “horse” is the same

for both of those individuals—the only difference lies in the different men-

tal associations each person has for that word. Furthermore, over time an

individual can come to have different emotional associations with the same

word. In that case, Frege does not think that the sense changes; rather, the

mental associations do. Mental associations can change, but the sense will

stay the same.

The second reason he gives for making this distinction is that mankind

acquires a stock of knowledge, a series of propositions we believe, and we

pass those propositions on from generation to generation. Therefore, in

a nonpsychological sense, the same thought (or proposition) is transmit-

ted from one generation to another. This process concerns something that

transcends the individual persons and the minds that are responsible for

the transmitting. For example, consider Isaac Newton in the eighteenth

century with various thoughts going through his mind. Suddenly, he statesthat gravity obeys the inverse square law and writes it in his Principia. After

this event, everyone who reads Principia acquires that thought, down the

ages, until the present day. Knowing such a thing is different from knowing



22 Chapter 1

Newton’s subjective, psychological ideas. Hence, when Frege speaks of

thoughts he refers to something that is objective and transcends time—

a thought is the objective unchanging sense of a sentence. Thoughts, inFrege’s use, are abstract entities.

Ideas are not the same as senses; rather, they are things that perish when

the mind that has them perishes. Ideas are not really shared by people.

Senses, however, are shared by people and do not perish with an individual

mind. For Frege, senses have the same objectivity and mental independence

as references. The sense of the word “gravity” existed back in Newton’s time

and we grasp that same sense now. Therefore, many subjective ideas can

correspond to the same objective sense. Frege’s general purpose in arguingfor senses to be objective is to show the objective basis for mathematics and

science in general.

It is important to note that ideas can also be objects of reference. In

normal speech, people do not typically talk about ideas. People have ideas

all the time, but they do not usually refer to them. For example, if someone

says, “It’s raining outside,” she is not saying anything about ideas at all. If

she were talking about ideas, she would say something like, “My idea that

it is raining outside is well founded.” Just as senses and words can be the

objects of reference, so too can ideas be the object of reference.

Frege constructs a complete picture for organizing all of these aspects

of language by forming a system of levels—words, ideas, senses, and refer-

ences. He illustrates his leveled system with an analogy:

The reference of a proper name is the object itself which we designate by its means;

the idea, which we have in that case, is wholly subjective; in between lies the sense,

which is indeed no longer subjective like the idea, but is yet not the object itself. The

following analogy will perhaps clarify these relationships. Somebody observes the

Moon through a telescope. I compare the Moon itself to the reference; it is the object

of the observation, mediated by the real image projected by the object on the glass

in the interior of the telescope, and by the retinal image of the observer. The former

I compare to the sense, the latter is like the idea or experience. The optical image in

the telescope is indeed one-sided and dependent upon the standpoint of observa-

tion; but it is still objective, inasmuch as it can be used by several observers. At any

rate it could be arranged for several to use it simultaneously. But each one would

have his own retinal image. On account of the diverse shapes of the observer’s eyes,

even a geometrical congruence could hardly be achieved, and an actual coincidence

would be out of the question. This analogy might be developed still further, by as-

suming A’s retinal image made visible to B; or A might also see his own retinal image

in a mirror. In this way we might perhaps show how an idea can itself be taken as an




object, but as such is not for the observer what it directly is for the person having the

idea. But to pursue this would take us too far afield.14

There is the telescope, the object observed through the telescope, the opti-cal image on the lens of the telescope, and the retinal image on the eye of

the observer. The retinal image is also an optical pattern that is projected

through the lens of the eye and passes on to the retina. There appear to be

three levels: the object out there, the optical image on the lens, and the

retinal image. Frege compares the optical image to the sense, and the idea

to the retinal image. The retinal image is different for each individual per-

son who looks through the telescope because we all have different retinal

structures. However, he thinks that the optical image is the same, even

though people observe it with different retinas. Therefore, the sense is an

objective thing in the same way that the optical image is an objective thing,

and different from the retinal image, which is subjective and depends on an

individual’s physiological makeup.

1.8 Problems with Frege’s Theory

In an earlier section, we discussed how Frege explains that “a = b” could not

state what he had previously held, namely that the name “a” denotes what

the name “b” denotes. He argued that his earlier thoughts on this were

incorrect because if the sentence says that “a” denotes what “b” denotes

then it is about not the objects those names designate but the names them-

selves. His solution to this problem is to bring in the notion of sense, which

contains the mode of presentation of the object. Associated with the name

“a” and the name “

b” there are particular modes of presentation, and this

fact accounts for the informative value of “a = b.”

To analyze “a = b” with Frege’s notions of sense and mode of presenta-

tion, we can consider a situation where MP1 is associated with the name

“a” and MP1 presents what MP2, associated with the name “b,” presents.

According to his theory, what makes a sentence such as “a = b” informative

is that one mode of presentation presents the same thing as another mode

of presentation.

Some readers may wonder why the same objection Frege makes againstthe name theory could not be raised against his own theory. On the surface,

the statement “a = b” appears to be about the objects a and b. However,

Frege’s theory is focused not on the objects themselves but on the mode



24 Chapter 1

of presentation of those objects. Common sense would tell us that “a = b”

does not seem to be about modes of presentation at all but about objects.

For example, few people would think that a statement involving the name“a” (e.g., “a is a planet”) is about a mode of presentation, unless the mode

of presentation itself is explicitly under discussion. It is natural to assume

that the statement is about an object and that the object is a planet. If

names are generally not about modes of presentation, we may wonder how

identity statements could be about modes of presentation. The problem is

that the subject matter of “a = b” is not the name “a” or the name “b,” nor

the mode of presentation of a and the mode of presentation of b, but the

objects a and b. At no point are we talking about words or the modes ofpresentation they allegedly express.

Frege raises no objections to himself in regard to this matter. However,

the question is a rather uncomfortable one because it exposes a gaping hole

in the theory he proposes in “On Sense and Reference.” If “a = b” is only

about objects, then he has regressed to his original problem: “a = b” states

that an object is identical to itself. Frege solves the problem of informative

value, but the way he solves it seems to raise the same kind of objection

he has against the names theory, which we discussed at the beginning of

this chapter. The only difference between these two is that one theory deals

with purely linguistic knowledge and the other deals with knowledge of

modes of presentation. Through the latter theory, Frege has shown us that

one mode of presentation can correspond to the same object as another,

but that does not allow the identity statement “a = b” to be about the actual

objects themselves. There is definitely a challenge here that Frege fails to

address, considering that his own theory commits him to something objec-

tionable by his own standards.

Philosophers have approached this problem differently. In Tractatus

Logico-Philosophicus, Ludwig Wittgenstein claims that these sorts of iden-

tity statements are ill formed. In natural language, Wittgenstein argues, we

can make such statements, but they express trivial propositions and not

substantial propositions. Wittgenstein thought that statements like this

must be eliminated from an ideal language because they do not make any

sense. However, Frege does not make any objection of that sort—instead heattempts to make a seeming triviality into something substantial. Though

Wittgenstein’s solution to this problem is to eliminate that sort of sentence

from an ideal language altogether, Frege tries to give a theory of it. He never

considers Wittgenstein’s more radical eliminative suggestion.




1.9 Extension of Frege’s Theory beyond Singular Terms

With an understanding of how sense and reference apply to singular terms,we will now examine how Frege extends this theory of expressions beyond

proper names and definite descriptions. In the text, Frege introduces his

theory by giving some arguments for his principal tenets. An explanation

of his overall theory is helpful before examining the text closely.

Singular terms, as we have seen, are subsentential expressions. It is

understandable to suppose that if Frege’s theory is applicable to singular

terms, the parts of sentences, then it should also be applicable to full sen-

tences. For example, consider the sentence “Hesperus is a planet.” Fregeargues that his theory can be extended so that the entire sentence has a

sense and reference. One of the odd things about Frege’s theory is that it

is clear that singular terms have references, but he must persuade us that

they have a sense in addition to reference; but the opposite problem holds

for full sentences—we can all agree they have sense, but we have to be per-

suaded that they have reference. In the case of our example, the sense of the

sentence is the nonpsychological thought expressed by it—the proposition

that Hesperus is a planet. The claim of reference seems much more difficult

for Frege to justify, and he produces a few different arguments for why the

whole sentence has a reference.

It is clear to the reader at this point what Frege means by the sense of

a sentence, but what about the reference of a sentence? First of all, Frege

thinks the reference of a sentence is its truth-value. The truth-value, for

Frege, is an object. There are only two truth-values, true and false; Frege

refers to them with the names “the True” and “the False.” If someone utters

a true sentence such as “Hesperus is a planet,” then its truth-value is the

True—an object—because it is true. If instead the speaker had said, “Hespe-

rus is a man,” that statement would be false, so the truth-value would be

the False.

To reiterate, for Frege, all true sentences refer to the truth-value the True

and all false sentences refer to the truth-value the False. Here the concept of

truth-value has nothing to do with value or ethics. Sometimes, particularly

in journalistic writing, “truth-values” have a completely different meaningpertaining to ethics. When Frege refers to truth-values, though, he is not

speaking of values in ethics. Frege makes two stipulations concerning the

truth-value of a sentence. The first is that a truth-value is the reference of a



26 Chapter 1

sentence; the second is that the reference of a sentence is an object. Right

away, we can see how peculiar both of these claims are. To say that a sen-

tence refers to its truth-value seems to misuse the word “refers.” After all,this word “refers” is the same word he uses for a singular term referring to

the object it designates (e.g., “Hesperus” refers to Venus). This type of rela-

tion of reference holds between names and objects, but to suppose that sen-

tences refer to anything in the same way names do is to part from what we

accept in ordinary language. People normally think that parts of sentences,

the singular terms, refer to things, but not whole sentences. For instance,

what is the reference of the sentence “Hesperus is a planet”? Naturally, it

would seem that the reference of this sentence would have something to dowith Venus, since it contains the name “Hesperus.” However, Frege thinks

the reference of that sentence is the truth-value the True—an object—since

the statement is true. To say that a true sentence refers to the truth-value

the True is certainly not the ordinary use of the word “true.” It is logical

to assume that a sentence has a truth-value—it is either true or false—but

it is still unclear why Frege claims that a sentence has as its reference a

truth-value.

His second claim, that the truth-value is an object, is just as counterin-

tuitive. In ordinary language, we would not assume the predicate “is true”

refers to an object. Frege does not specify a special meaning for the word

“object.” He seems to be using the word “object” in an ordinary way, as

when it refers to an external thing in the world (e.g., a person, a planet, or

a house). His claim that the True is an object too is quite strange. It means

that in Frege’s complete list of all the objects in the world, in addition to

ordinary objects—every person, planet, elementary particle, and so on—we

would also include the True and the False. Hence, Frege considers the True

and the False as entities to which one can intelligibly refer. Although these

two doctrines seem strange, their purpose from the theoretical point of view

is not as puzzling—by introducing these notions Frege can extend the theory

of sense and reference to whole sentences. Then, not only will the singular

terms have a sense and a reference, but the sentences the terms are a part of

will also have a sense and a reference. The sense is the thought the sentence

expresses, the reference is the truth-value, and the truth-value is an object.This is nice and neat, to be sure, but it sounds extremely far-fetched.

Theoretically, in extending the whole apparatus to sentences, another

possibility arises—the extension of sense and reference to complex




sentences. Consider an example where someone says, “Hesperus is a planet

and Mars is a planet.” In this case, the truth-value of the sentence depends

on the truth-values of each of the component sentences. The applicationof Frege’s theory to this example would show that the sentence before the

conjunction refers to the object the True and the sentence after the con-

junction also refers to the object True. Therefore, the truth-value of a con-

junction of two sentences both referring to the True would be the True.

These examples illustrate how Frege attempted to extend the theory of

sense and reference beyond the simplest case, where it seems very plausible,

to more complex cases, where it seems less plausible. Now that we have

examined more generally the two basic doctrines of Frege’s extension ofsense and reference to full sentences, we can begin to look in detail at his

arguments in the essay itself. Frege begins his discussion in the following

passage:

So far, we have considered the sense and reference only of such expressions, words,

or signs as proper names. We now inquire concerning the sense and reference for

an entire declarative sentence. Such a sentence contains a thought. Is this thought,

now, to be regarded as its sense or its reference? Let us assume for the time being that

the sentence has reference. If we now replace one word of the sentence by anotherhaving the same reference but a different sense, it can have no bearing upon the ref-

erence of the sentence. Yet we can see that in such a case the thought changes. For

example, the thought in the sentence ‘the morning star is a body illuminated by the

Sun’ differs from the thought in the sentence ‘the evening star is a body illuminated

by the Sun.’ Anybody who did not know that the evening star is the morning star

might hold the one thought to be true and the other false. The thought, accordingly,

cannot be the reference of the sentence but must rather be considered as the sense.15

Here Frege assumes that the reader will question why sentences should

have a reference. If we assume that the sentence has a reference, then it

is possible for the sentence to refer to its expressed thought. Whatever the

reference of the sentence is must be invariant under substitution of terms

in the sentence that have the same reference. Therefore, whatever the refer-

ence is must be something that is uniquely determined by the references of

each of the terms in the sentence. Take the following example:

(9) Hesperus is F and Phosphorus is F. (“F” stands for any property.)

These conjuncts, according to Frege, express two different thoughts, where

“Hesperus is F” expresses thought T1 and “Phosphorus is F” expresses

thought T2. The question is whether the reference of “Hesperus is F” is T1.



28 Chapter 1

For Frege, whatever the reference is, it will be preserved when anything is

substituted with the same reference for any term in the original sentence—

because the reference of the whole is a function of the reference of its parts.Suppose that in the statement above we now substitute the name “Hes-

perus” for the name “Phosphorus.” Since they have the same reference,

swapping the names should be possible without affecting the statement’s

truth-value. Of course, the resulting statement is still true, because Hes-

perus is F and Phosphorus is F. However, the sentence “Phosphorus is F”

does not have the same sense as the sentence “Hesperus is F.” Since they

do not express the same sense, it follows that they do not express the same

thought. If they express two different thoughts, those thoughts cannot bethe reference of the sentence. In other words, if the thought were the ref-

erence of the sentence, it could not be true that the reference of the sen-

tence depends on the reference of each part of the sentence. Therefore, the

thought is not the reference of the sentence.

Despite all of our discussion so far, the question remains: why does Frege

think that the sentence refers to anything? Why does he think that it refers

to a truth-value, and why does he think that the truth-value is an object?

The premise of concern in Frege’s argument is based on the example of

the sentence “Odysseus is a brave man,” which contains the empty name

“Odysseus,” a name with a sense but no reference. Such cases are common

for scholars of epic poetry and mythology. In such cases, the thought itself

is what is important and not the truth-value. However, if our interest lies in

what is true in reality, then we must look at the reference of the sentence

“Odysseus is a brave man.” Only by determining the reference is it possible

to determine if the object referred to in the sentence, Odysseus, has the par-

ticular property attributed to it. Therefore, the truth-value of the thought

lies not only in the thought itself but also in what the thought refers to,

since the reference determines the truth-value.

This premise of Frege’s—that the truth-value of a thought is determined

by the references of the parts of the sentence—seems logically sound. He

continues in the following passage with an explanation of how to extend

this premise to the sentences themselves having references:

The thought remains the same whether ‘Odysseus’ has reference or not. The fact that

we concern ourselves at all about the reference of a part of the sentence indicates

that we generally recognize and expect a reference for the sentence itself.16




Frege does not qualify his statement here, and he makes a huge logical leap.

Unless he is able to give a more thorough defense of his reasoning, there

is absolutely no reason why sentences should have reference just becausetheir parts do. Furthermore, if our interest is in the truth-value of the sen-

tence, and the truth-value can be known through the parts of a sentence,

there is no reason to concern ourselves additionally with the reference of

the sentence. If the term in the sentence (e.g., “Odysseus”) refers to some-

thing real, that makes the truth-value of the sentence the True, assuming

the denoted object has the attribute predicated. Frege gives no explanation

for why we should also acknowledge that the sentence itself has a reference,

and the passage above is the only place where he tries to defend this posi-tion. The sentence may indeed have the property of being true, but it is a

further question whether sentences refer to the True.

Although this part of Frege’s argument is flawed, he makes two further

claims that should be investigated. First, he claims that sentences have

truth-value, and then he claims that the reference of a sentence is its truth-

value. He concludes that the reference of a sentence is its truth-value in this

passage:

We have seen that the reference of a sentence may always be sought, whenever the

reference of its components is involved; and that this is the case when and only

when we are inquiring after the truth-value. We are therefore driven into accepting

the truth value of a sentence as constituting its reference. By the truth value of a sen-

tence I understand the circumstance that it is true or false.17

Frege concludes here that the reference of a sentence must be its truth-

value. The only reasoning behind his conclusion is that the truth-value

of a sentence is something that is determined by the reference of its parts.

This point can also be made through our earlier examples of substitution

arguments. When substituting coreferential singular terms, the truth-value

is preserved. The truth-value of “Hesperus is F” remains the True when we

substitute “Phosphorus” for “Hesperus.” Therefore, if the reference of the

sentence is preserved by substituting coreferential singular terms, then it

can be argued that the truth-value is the reference. However, some prob-

lems arise from this conclusion.

Although something may in fact be preserved under substitution ofcoreferential terms, there is no reason to call what is preserved the reference

of the sentence. Furthermore, in addition to the truth-value, there is some-

thing else such a substitution preserves that Frege never considers—what



30 Chapter 1

we call a fact, a state of affairs that makes a statement true. In this respect,

the fact stated by “Hesperus is a planet” would be the same fact as stated

by “Phosphorus is a planet,” because facts concern objects and properties,not the words used to describe them. The fact that makes the first state-

ment true is the fact that makes the second statement true—that a certain

object has a certain property. When we substitute one coreferential name

for another, truth-value is preserved, but so is the fact that makes those

statements true. In other words, the state of affairs that corresponds to the

sentence is preserved. So why not say this is the reference?

In addition to the truth-value, then, the fact is also invariant under sub-

stitution of coreferential terms. This proposal is less counterintuitive thanFrege’s: in Frege’s view, every true sentence has the same reference and every

false sentence has the same reference. However, it is not true that every true

sentence corresponds to the same state of affairs. Therefore, the state of

affairs is a much more useful concept than truth-values in this case—if sen-

tences have references at all, states of affairs would seem a better choice. If

we only suppose that the reference of a sentence is its state of affairs, then

our only components are the sense and the state of affairs. There is no need

to speak of truth-values as objects of reference. This proposal is much more

logically sound than making the odd claim that the sentence refers to its

truth-value, and that all true sentences have the same reference. Another

way to challenge his argument is to propose that the sentence has no ref-

erence at all and only expresses a thought. It is clear why singular terms

should have reference, but his reasoning for why thoughts should have

reference is absent of any intuitive or argumentative justification.

A problem also emerges if we take a closer look at Frege’s proposal that

the truth-value of a sentence is an object. Contrary to Frege’s proposal, the

truth-value appears to be a property of something that the predicate “is

true” ascribes. Why does he think that “is true” is a singular term for an

object, the True? In fact, he has to completely deny the way languages are

structured when using this concept of truth. Instead of a sentence standing

in relation to an object called “the True,” why not just say that being true

is a matter of a sentence having the property of being true? Transforming

the truth-value from a property to an object is an unnecessary step thatFrege takes in attempting to extend his theory of sense and reference to

sentences. Sentences are just not like singular terms.




There is still, possibly, one explanation that Frege could produce by

drawing on an earlier theory he has about complete expressions, incom-

plete expressions, and objects. His view is that a complete expression alwaysdesignates an object whereas an incomplete expression always designates a

concept. His notion of object is extremely broad and is whatever is referred

to by a complete expression. Singular terms are complete expressions and

sentences are complete expressions. It is clear why sentences are complete

expressions—they are used to make statements. Why he thinks a singular

term is a complete expression is more obscure, because a singular term can-

not be used to make a statement. But, since Frege thinks that proper names

are complete expressions and complete expressions designate objects, andsince he thinks that sentences are complete expressions, he concludes that

both must designate objects. He argues they must, because that is what he

means by an object—something that is designated by a complete expres-

sion. Therefore, the object that a sentence must designate is a truth-value

(even though it could have been a state of affairs).

The natural objection to this idea is that he uses a completely technical

sense of the word “object,” since he claims that an object is to be defined as

whatever a complete expression denotes. Of course, it is possible to define

“object” that way, but he has shifted the sense of the word “object” from its

ordinary sense to his own technical sense. In the same way he stipulated a

new meaning for the word “object,” he could have stipulated that whatever

is denoted by a complete expression is a dog. Frege could then argue that

he has a technical interpretation of the word “dog” such that “dog” means

whatever is designated by a complete expression. In doing so, Frege would

have completely changed the meaning of the word “dog” and used it to

refer to the truth-value, in the same way he used the word “object.” The

suspicion is that he has taken over the meaning of the word “object,” which

has a well-established meaning and use. Someone can stipulate whatever he

likes, but this does not mean he has discovered anything significant, such

as that truth-values are objects (or dogs).

1.10 Further Aspects of Frege’s Theory

A sentence does not, for Frege, always refer to a truth-value, any more than

a singular term always refers to its customary reference, because sometimes

it shifts its reference. Remember that if a name is quoted in a sentence, it



32 Chapter 1

does not refer to its customary reference but to the name itself. In the same

way, quoting a sentence would result in a reference to the sentence itself,

and not its truth-value. According to Frege, that is not the only case of refer-ence shift, or the most interesting case. Sentences refer to something other

than their truth-value when they occur in what are called opaque contexts.

Consider this example: “John said that Hesperus is a planet.” Now in this

example there is a subsentence: “Hesperus is a planet.” Here Frege thinks

we are referring neither to the truth-value of that subsentence nor to Hes-

perus. In such opaque contexts, “Hesperus is a planet” now refers to the

thought John expresses when the sentence occurs outside that context. On

the other hand, if it occurs on its own, it expresses its customary sense andrefers to a truth-value. But when it occurs in the opaque context the refer-

ence shifts. The name “Hesperus” refers now to the sense it normally has,

the customary sense, and the whole sentence no longer refers to its truth-

value but to its customary sense, which is a thought. So it is not true that for

Frege the sentence always refers to its truth-value (this may make us wonder

why he is so convinced that it ever refers to a truth-value). The basis for this

shift of reference lies in the fact that when a sentence occurs in this kind

of context, its truth or falsity does not matter to the truth or falsity of the

whole sentence. For example, if Jane says, “John said that Hesperus is cream

cheese,” Jane said something true even though John said something false.

Whether what John said is true or not does not matter so far as Jane’s report

is concerned, as long as Jane quotes him properly. Since the truth-value of

her statement depends only on the accuracy of the quotation, Frege thinks

that the truth-value of this opaque context sentence depends solely on the

sense of those words. All words then refer to at least two things, according

to Frege: ordinary uses of words refer to their ordinary reference, but in

opaque contexts they refer to their ordinary sense.

Although words in opaque context all have references, we may wonder

whether or not they all have distinct senses. The sense of the name “Hes-

perus” in an ordinary context cannot be the sense of the name “Hesperus”

in an opaque context. Otherwise, the sense would be identical to the refer-

ence, since the reference now is its ordinary sense. To solve this problem,

Frege proposes that there must also be an indirect sense. Now, in additionto every name having two references depending on the context, it also has

two senses. The name has its ordinary sense and also the sense it has when

it occurs in an opaque context. We can understand why the indirect sense




has to exist, given Frege’s assumptions, but we do not know what it is. Since

it is referred to, there must be a sense that refers to it. The sense is a mode

of presentation, so the indirect sense is a mode of presentation of a mode ofpresentation. What kind of creature is that?

Another way to explain Frege’s proposal is to consider someone looking

at an object from a particular perspective. Frege would introduce the notion

of an indirect perspective, a perspective on a perspective. But what is that

exactly? It is not possible to have two perspectives on a perspective, because

movement (a different position in front of the object) would cause a new

perspective. Further, Frege does not tell us what this new perspective-on-a-

perspective might be. Is it possible to perceive a perceptual perspective froma specific perspective? He explains the ordinary sense well enough with the

examples of the triangle and planets, but he never gives an example of the

senses that correspond to these words when they occur in opaque contexts.

We are left wondering how there can be a mode of presentation of a mode

of presentation. At this point the theory is generating consequences that

are completely detached from anything that has a clear articulation. If we

give Frege the benefit of the doubt, then there must be cases where there is

a mode of presentation of a mode of presentation of a mode of presentation

(e.g., Jane says, “John said that I said that Hesperus is cream cheese”). There

is no explanation as to what that third-order mode of presentation might

be. The multiple modes of presentation are all meant to be distinct from

each other, but we do not know what they are.

Even with these difficulties in his theory, we must not overlook how

attractive Frege’s theory is from a theoretical point of view. It has a simple

structure, with only a few components. Further, it is a unique semantic

theory that had not existed prior to its introduction in his essay. Frege has

attempted to build a kind of mathematical theory of meaning, elegant and

economical. However, he runs into trouble when he tries to apply the the-

ory to natural language, which is not so streamlined. He tries to squeeze

too many disparate things into his mathematically inspired model. But

Frege’s contribution to the philosophical understanding of the semantics

of language is tremendous. In many ways “On Sense and Reference” was

the essay that began the discussion on how to develop a rigorous theory oflanguage. Though some of Frege’s doctrines in this essay are highly ques-

tionable, his idea of sense and reference for singular terms influenced phi-

losophers far into the future, and we will often return to it.





2 Kripke on Names

2.1 Background

We will now jump forward eight decades. The reason for this is that Frege’s

theory of sense for names received its most sustained criticism in 1972,

though criticism had been brewing for a while. Thematic continuity thus

trumps chronological continuity. In this chapter we will discuss the descrip-

tion theory of names and Saul Kripke’s critique of it in Naming and Neces-

sity .1 Since Frege is widely credited with inventing the description theory of

names, Kripke’s critique is directed largely at Frege and those who followed

his lead. Frege’s essay “On Sense and Reference” contains a footnote that

states the theory that Kripke is criticizing. So here is footnote 4 of that essay:

In the case of an actual proper name such as ‘Aristotle’ opinions as to the sense may

differ. It might, for instance, be taken to be the following: the pupil of Plato and

teacher of Alexander the Great. Anybody who does this will attach another sense to

the sentence ‘Aristotle was born in Stagira’ than will a man who takes as the sense of

the name: the teacher of Alexander the Great who was born in Stagira. So long as the

reference remains the same, such variations of sense may be tolerated, although they

are to be avoided in the theoretical structure of a demonstrative science and ought

not to occur in a perfect language.2

The point Frege makes in this footnote is that when different people speak

a language containing a single proper name they can associate different

descriptions with that name. Since that is possible, the proper name to

which speakers assign those two or more different descriptions is ambig-

uous. Such ambiguity is a defect of natural language. In a properly con-structed scientific language, the same proper name would not be allowed

to have two or more different senses by being associated with two or more

different descriptions. Still, in ordinary language, people may well assign



36 Chapter 2

different descriptions to the same name. Frege is here assuming that what

people mean by a name is expressed in a definite description, and he is con-

cerned that the descriptions can vary, thus producing unwanted ambiguity.In Naming and Necessity Kripke is interested not in the ambiguity issue

but rather in the underlying theory of the meaning of names. He is inter-

ested in the theory of names that supposes that the meaning of a name—

its sense—is given by a definite description. Frege writes the footnote as

though this theory does not warrant a discussion and merely raises the

specter of ambiguity in natural languages. Perhaps he regards the descrip-

tion theory as self-evidently true, so in need of no defense.

Before we discuss any of Kripke’s critical points, it is important to havea basic understanding of the description theory of names. Take an example

of a proper name like “Aristotle.” The name “Aristotle” refers to a long-dead

individual. In the present day, someone can say, “Aristotle was a great phi-

losopher,” and refer to that long-dead individual, and there is no ambiguity

as to whom he means. There was a certain individual back in Ancient Greece

and that very man is the man we refer to when we say “Aristotle” today. Of

all the billions of people who have lived we manage to pick out just one of

them with the name “Aristotle.” Remarkable! How do we do that? Certainly

not just in virtue of the sound the name makes when you utter it. Also, we

can make true statements about this man like “Aristotle wrote The Metaphys-

ics.” We refer to a unique individual and then we say something true about

him. Thus names permit a remarkable feat of linguistic time travel, homing

in on a man who existed over two thousand years ago.

The question arises: how can we refer to such a long-dead individual by

using a name? We do not see evidence of how in the name itself. The name

is just a piece of language—a shape or a sound. It would be impossible to

scrutinize the name as it is written or pronounced and somehow deduce

the identity of the man to whom it refers. To answer this question, philoso-

phers following Frege have wheeled in the description theory.

The description theory uses definite descriptions that can be applied to a

certain individual and no one else to enable a speaker to refer to that indi-

vidual. Aristotle can be referred to with the definite description “the best

pupil of Plato.” Definite descriptions enable the speaker or writer to refer toa certain individual by combining a number of different words, such that

that combination of words refers only to that particular individual. In addi-

tion to “the best pupil of Plato,” other examples of definite descriptions



Kripke on Names 37

would be “the tallest man in Australia” or “the president of the United

States.” The key point here is that the description must refer to one indi-

vidual and one individual only. There is only one tallest man in Australia,just as there is only one president of the United States and one best pupil of

Plato. Such descriptions are uniquely identifying.

The definite description “the best pupil of Plato” refers uniquely to Aris-

totle in virtue of the fact that Aristotle alone fits that description. In other

words, he uniquely satisfies the terms in that description. He was a pupil

of Plato, and he was the best pupil of Plato, and this definite description

expresses those properties. Therefore, when that definite description is used

it does not refer to anybody else but Aristotle. Definite descriptions containa predicate (“is the best pupil of Plato”) and only one object (Aristotle) satis-

fies that predicate.

Initially, it may seem as though the name “Aristotle” is not made up of

the terms in the definite description and that the name does not express

any of Aristotle’s properties. After all, it does not on its face express any

properties that a certain individual had way back in ancient Greece. There-

fore it cannot refer in the way the definite description refers, because it does

not have the same semantic nature. But according to the description theory,

the name “Aristotle” does work in the same was as a definite description.

According to this theory, the name is in fact synonymous with the descrip-

tion. The name “Aristotle” is used as a short form of the definite description

“the best pupil of Plato” for purely practical reasons. It is inconvenient to

continually refer to someone with a lengthy definite description. Instead

of repeatedly saying “the best pupil of Plato,” we abbreviate this definite

description to a synonymous name, “Aristotle.” We could shorten it even

further if we liked (e.g., to the name “Ari”), but it all accomplishes the same

purpose—to make it easier to refer to that particular individual. Thus names

are just condensed definite descriptions, and their mode of reference is the

same as that of descriptions.

In other words, the definite description defines the name “Aristotle.”

The name “Aristotle” is therefore a disguised form of the definite descrip-

tion. Notice that this theory is a surprising theory because on the face of

it the name is not a definite description; this is why it is thought to bea disguised definite description. Now we know that the name “Aristotle”

refers to Aristotle because it is short for a definite description of Aristotle.

Since the definite description refers to him, the name “Aristotle” also refers



38 Chapter 2

to him. So, if John says to Jane, “Whom do you mean by ‘Aristotle’?” she

could reply, “I mean the best pupil of Plato,” and her statement would be

an example of the description theory of names.To understand the description theory, it is important to see how it works

and what its commitments are. The first thing to consider is that according

to this theory the sense of the name “Aristotle” is expressed by the definite

description “the best pupil of Plato,” so that when names differ in their

sense, they are short for different definite descriptions. Since the sense of

the definite description constitutes the sense of the name, we can apply

Frege’s account of the sense of definite descriptions in terms of modes of

presentation, as discussed in chapter 1. Therefore, a definite descriptiongives a mode of presentation containing a specific aspect of the refer-

ence. Two names with the same reference can express different definite

descriptions.

The sense is what is understood when a name is spoken or written. In

understanding the name “Aristotle” one grasps the sense of the name and

therefore the sense of the associated definite description. The theory of

descriptions, then, is a theory of what understanding the name consists in,

and what one’s grasping the meaning of a name is a grasping of.

The theory also tells us what constitutes the information value of the

name. Informative identities can be formulated with names, and the asso-

ciated defining definite descriptions give their information value. In the

case of the names “Hesperus” and “Phosphorus,” the descriptions are “the

evening star” and “the morning star,” respectively. In our discussion about

identity statements using names in chapter 1, we saw that the information

value of these two names differs, since the two definite descriptions are

not synonymous with one another—one says “evening star” and one says

“morning star.” To determine what proposition is expressed by “Hesperus is

Phosphorus” we must substitute the descriptions for the names. Since the

two descriptions are not synonymous, those types of descriptions differ in

their information value; so the names that abbreviate them have a different

information value.

Further, the theory of descriptions explains what uniquely determines

the reference of a name. The definite description refers to only one par-ticular individual. For example, the definite description “the best pupil

of Plato” is a unique condition that only Aristotle satisfies. Therefore, the

definite description determines the reference of the name. This part of the



Kripke on Names 39

description theory is in accordance with Frege’s theory of sense and refer-

ence, as discussed in chapter 1, since sense is held to determine reference.

The sense incorporates the description, the description determines the refer-ence, and so sense determines the reference. For these reasons, when some-

one utters the name “Aristotle” she is referring to one individual only. The

description is what “targets” the name’s reference to a particular individual.

Finally, the theory explains how name reference comes to be introduced.

When a particular name becomes introduced in a language, it can be intro-

duced through a definite description. By contrast, we can imagine a situa-

tion a couple of thousand years ago where a baby is about to be baptized,

and the priest asks, “What is the name of the baby I am about to baptize?”The mother says, “Aristotle,” and the priest continues, “Let the baby before

us now be called henceforth ‘Aristotle.’” Another example is a definite

description that denotes an individual who is not in close proximity to the

speaker. For instance, one could say, “I will call the tallest man in Australia

by the name ‘Herbert.’” The point is that descriptions can be used to intro-

duce names and bring them into the language.

2.2 Kripke’s Critique

The description theory was extremely popular among philosophers for a

long time, and the main tenets of the theory remained virtually unchal-

lenged from the time Frege introduced it until Kripke raised objections

in 1972. Naming and Necessity contains a series of lectures that generated

a considerable amount of controversy because Kripke claimed that the

description theory was completely wrong. Moreover, he seemed to prove

that it was completely wrong, which was shocking to philosophers, since

the theory had been well established for over seventy years. Kripke’s argu-

ments were received with a good deal of surprise by the philosophical com-

munity, because the description theory seems like such a natural theory.

It just has so much going for it. It is important to note that this theory

describes the psychological condition of a person who understands or uses

a name. The idea is that if the name is synonymous with a description, then

a description must be psychologically present in the mind of the personwho utters the name. The theory tells us what it is to know the meaning of

a name. We must now turn to Kripke’s critique of the theory, fully aware of

its merits and content.



40 Chapter 2

The description theory says that a name “ A” is synonymous with

a description “the F .” Now consider the sentence “ A is the F .” This sen-

tence will have several properties. First, it will be known to be true a priori.Without any empirical investigation, this sentence can be known to be

true simply by understanding the name “ A.” If “ A” is synonymous with

“the F ,” all one needs to know is the meaning of the name “ A” to know

that A is F . Compare “Bachelors are unmarried males”: there is no need to

know anything more than what “bachelor” means to know that bachelors

are unmarried males. However, if someone said, “Bachelors are unhappy,”

that illustrates an example of an a posteriori statement—one that requires

research into the empirical world to determine if it is true. The truth of thatstatement cannot be shown just by virtue of the definition of “bachelor.”

According to the description theory, “ A = the F ” is analytic—true by defini-

tion—and a priori, because the description gives the meaning of the name

and no more.

A further property of “ A = the F ” is that it must be a necessary truth. If a

truth is analytic, it is true in all possible worlds. Given that the two terms

are synonyms in that statement, the statement is a necessary truth, just as

“ A = A” is a necessary truth. It will thus follow that A is F in every possible

world, just because “ A” means “the F .” Thus, according to the description

theory, the proposition expressed by “ A is the F ” will be a priori, analytic,

and necessary. These are straightforward consequences of the description

theory. Notice that not every description you couple with the name will

have those consequences, because not every description is supposed to be

synonymous with the name. Only certain descriptions are synonymous

with the name. When someone says, “Aristotle,” he might mean the best

pupil of Plato, but he can go on to attribute other properties to Aristotle

that are not contained in the meaning of “Aristotle,” for example, having

a mole on his left elbow. Some definite descriptions, then, will give rise to

statements that are a posteriori, synthetic, and contingent. Clearly, some

of the things that are true about Aristotle are true about him only contin-

gently. The main point to understand, though, is that some of the descrip-

tions are true of him analytically and a priori, according to the description

theory.Given what the description theory entails, Kripke’s question is the fol-

lowing: is it true that there is a description “the F ” such that it generates a

proposition expressed by “ A is the F ” that has these three characteristics?



Kripke on Names 41

So, is it true that “Aristotle was the best pupil of Plato” is a priori, analytic,

and necessary? If this is true, then the description theory is correct; but if

not, it is incorrect. Kripke argues that there is no description, or cluster ofdescriptions, regularly associated with a name that generates these three

characteristics. Thus the description theory of names has to be false.

Kripke first argues against the necessity of the description. He uses the

same example as Frege (“Aristotle”), so we can use our definite description

of Aristotle here as well (“the best pupil of Plato”). He attempts to show that

the fact that Aristotle was the best pupil of Plato is a contingent truth and

not a necessary truth.

Of course, it is not being disputed that Aristotle was the best pupil ofPlato, because he wrote a number of the formative texts of Western phi-

losophy and is one of the most influential philosophers of all time. In the

real world, there is not much debate about Aristotle being the best pupil of

Plato. In our world, Aristotle was indeed the best pupil Plato ever had (he

always got an A+). However, Kripke asks us to consider alternative reali-

ties—possible worlds—where this may not have been the case. There is the

actual world, the world we reside in now, where things are a certain way. In

this world Aristotle was a philosopher, the sun rises in the east, and a man

walked on the moon. Then, there are possible worlds, which are alterna-

tives to the actual world, in which different things are the case.

Imagine Aristotle was born in the same year, had the same parents, and

lived in the same household. However, in this alternate reality, he has an

accident as a child where he banged his head on a Greek sculpture and suf-

fered enough brain damage to prevent him from any further academic pur-

suits. Although this did not (thankfully!) happen in our world, it could have

happened in another world. Such events could contingently happen. If that

had been so, Aristotle would not now be called the best pupil of Plato—he

wouldn’t have been a philosopher at all. There are less extreme examples

of possible worlds in which the Aristotle we know could have turned out

differently. If Aristotle had strong musical interests, he could have attended

another school that was not Plato’s academy to develop his musical talents.

Therefore, Kripke argues, it is quite contingent that he became a philoso-

pher and not something else, such as a harp player.The point is that there are many contingent facts about people that can

be expressed in definite descriptions. It is not necessary that we pursue one

particular path in life (e.g., a philosopher). We could have easily pursued



42 Chapter 2

other paths, just as Aristotle could have. These facts are contingent; they are

not necessary facts like 2 + 2 = 4 or all bachelors are unmarried males. They

could have been otherwise.Since it is only a contingent fact that Aristotle was Plato’s best pupil, the

statement “Aristotle was the best pupil of Plato” expresses only a contin-

gent fact and not a necessary fact. But if “ A = the F ” is not necessary, then

the name “ A” does not mean the same thing as the description “the F .”

Therefore, the description theory is false. This argument of Kripke’s can be

called the “modal argument” because it deals with questions of modality,

that is, what is necessary and contingent.

Frege (and later Russell) thought that when using a name like “Plato” or“Aristotle” we have in mind some famous deeds of the individual denoted.

Eventually, the description of those famous deeds becomes synonymous

with the name. Kripke’s objection to this proposal is that when a person

performs those famous deeds, he has not necessarily performed them. It is

conceivable that he might not have performed such deeds, and therefore it

is not a necessary truth that he performed those deeds.

2.3 Rigid Designation

At this point, Kripke explains his concept of rigid designators and non-

rigid designators. To begin, we can first discuss the non-rigid designator.

Again, Kripke brings up the idea of possible worlds. Let’s consider the defi-

nite description “the most famous pupil of Plato.” In the actual world, it

designates Aristotle, but it does not designate him in every possible world.

In some possible worlds, Aristotle does not even exist, since it is not true in

every possible world that Aristotle’s mother gave birth to him. Therefore,

the definite description “the most famous pupil of Plato” is a non-rigid des-

ignator, meaning it designates different objects in different possible worlds

from what it designates in the actual world. The non-rigid designator itself

stays the same when considering every world, but in different worlds it

designates different individuals or objects, depending on who does what

in that world.

A rigid designator, then, is one that designates the same object in everypossible world. Kripke argues that, for example, proper names are rigid

designators. Before we explain what that means, let us examine a conse-

quence of that for the description theory of names. If it is true that definite



Kripke on Names 43

descriptions are non-rigid designators, and if it is true that names are rigid

designators, then it cannot be true that names are synonymous with definite

descriptions, because they are semantically different. If Kripke shows thatnames are rigid designators and definite descriptions are non-rigid designa-

tors, he will have shown the description theory to be false. In other words,

he will show that names refer to the same thing in all possible worlds, but

definite descriptions refer to different things in different possible worlds.

The reason Kripke holds that a name is a rigid designator is that a name

refers to one specific individual and only to that individual from world to

world. He holds that the name “Aristotle” designates the same person in

all possible worlds. Suppose in the actual world the only person with thename “Aristotle” was that particular Greek philosopher. Now could “Aris-

totle” have denoted anyone other than the actual Aristotle we refer to with

that name? That is, could Aristotle be anyone other than Aristotle? Clearly

not. Given the meaning of “Aristotle” as it now exists, it cannot denote

anyone other than the person it actually denotes. Someone other than Aris-

totle could have been denoted by “the most famous pupil of Plato,” but no

one else could be Aristotle himself. We use the name to pick out a specific

individual and this reference stays constant from world to world. It is as if

the name gets hold of a certain individual and won’t let him go as we tra-

verse modal space, whereas descriptions allow us to vary our reference as we

travel from world to world.

Kripke makes this point using a number of different names (e.g.,

“Moses”), but the same point applies in every case. We can summarize his

argument in the following way: if the description that is held to be syn-

onymous with the name is a description that records famous deeds of the

bearer of the name, and those famous deeds are contingent properties of

the bearer, then they cannot hold with necessity of that individual. There-

fore, they cannot be synonymous with that name. To put it differently, the

descriptions of famous deeds give rise to non-rigid designators like “the

most famous pupil of Plato,” but names are rigid designators—so the latter

cannot mean the same as the former.

It is important to note a couple of things about the force of this argu-

ment so far. The first point is that the argument works only if the descrip-tion expresses a contingent property of the object denoted. However,

questions may arise as to whether or not every description in a language

gives only a contingent property of the object. Kripke himself acknowledges



44 Chapter 2

that descriptions are not always non-rigid designators and that there are

instances where they are rigid designators. To illustrate this point, consider

“3 is the successor of 2.” This sentence has the same logical form as “ A

=the F .” The numeral “3” is a name of the number 3, and that number must

be identical to the successor of 2—no other number than 3 could be the

successor of 2. This statement is a necessarily true one, not a statement of

contingent fact. It could not have been the case that in other worlds 3 is

the successor of the number 82. Since the successor of 82 is 83, 3 cannot be

83, because it is built into the nature of 3 that it is not 83. Therefore, the

definite description “the successor of 2” is a rigid designator for the number

3. There is no possible world in which that description can designate any-thing other than the number 3.

The modal point Kripke makes about the description theory is based on

descriptions that designate famous deeds that are rooted in contingency.

But what if the description described aspects of the reference that are not

merely contingent? In that case Kripke’s modal objection would not apply.

If there are properties of human beings that are necessary properties of

them in the same way that being the successor of 2 is a necessary property

of 3, that would show the theory of descriptions to be less vulnerable than

Kripke claims.

In some of Kripke’s other work, he discusses something called the neces-

sity of origin. This idea stipulates that the essence of a person comes from

the origin he actually has. In other words, there is no possible world where

Aristotle existed and came from different parents than the ones from which

he actually came. In different possible worlds, even if there were an indi-

vidual who resembles Aristotle down to the last detail, he would not qualify

as being Aristotle unless he had the actual Aristotle’s origins. We can express

this essentialist claim in a definite description: “the person with origin O.”

Now we can say, “ A is (necessarily) the person with origin O,” or “Aristotle is

(necessarily) the person who came from parents A and B.” We can agree with

Kripke that this statement expresses a necessary truth. In that case, there is

no refutation of that version of the description theory on the basis of non-

rigidity and contingent properties, because now in every possible world Aris-

totle satisfies that description: he is necessarily the person with origin O. Themodal argument works only if the description is contingent, but not all are.

In addition to the necessity of origin, there are different theories about

personal identity. One theory is that a person is identical to his brain.



Kripke on Names 45

Under this theory, if the brain of Aristotle were transplanted into the body

of Einstein, the resulting person would be Aristotle. Since Aristotle’s brain

carries his identity, it does not matter what body his brain has been trans-planted into. Take a person with brain B. If Aristotle is the person with

brain B, nobody could be Aristotle without having brain B, and anybody

would be Aristotle who has brain B. Therefore, the description “the person

with brain B” always designates Aristotle in every possible world, so that

that description is necessary or rigid. That description will not result in

these modal objections, objections having to do with the contingency of

the property expressed.

In Naming and Necessity , Kripke never considers these types of rigiddescriptions. He does build a convincing argument against the famous

deeds version of the description theory, but we have no reason to take the

famous deeds theory to constitute the entire scope of the description the-

ory. Even if Frege and Russell were fixated on famous deeds, many other

examples of descriptions do report something non-contingent about an

individual. We must next consider Kripke’s other objections to see if they

overcome this limitation.

2.4 Kripke’s Epistemic Objections

One of Kripke’s nonmodal objections has to do with whether or not some-

thing is a priori. If a statement is analytic—true by definition—then it must

be a priori— knowable without examining the world. If it is not a priori, then

it is not analytic. If it is not analytic, then the terms are not synonymous;

and if they are not synonymous, then the description theory is false. Kripke

gives the example of the physicist Richard Feynman. He supposes that

someone knows that Feynman is a physicist but does not understand his

specific contribution to physics. Most people are not experts in physics and

will not be able to tell you what Feynman’s unique discoveries were but can

still say, “Feynman was a famous physicist.” If the same person was asked

who Gellman was, he could say, “Gellman was a famous physicist too.” It

is evident that with these two descriptions, nothing distinguishes the two

physicists from one another—both are simply “a famous physicist.” Theperson who made those statements does not have sufficient knowledge in

his mind to descriptively define Feynman or Gellman. Kripke’s point is that

the same information will be associated with the names by our nonexpert



46 Chapter 2

speaker, but this information is not sufficient to pick out one physicist from

the other. Therefore, the descriptive information in the speaker’s mind does

not determine the reference of the names—yet the speaker does manage torefer to specific distinct individuals. He does not know any definite descrip-

tion true of his reference, so he certainly doesn’t know any such description

a priori. Even though the speaker cannot distinguish between Feynman and

Gellman, he is not referring to Gellman when he uses the name “Feyn-

man.” In this case, the speaker does not have the kind of knowledge that

the description theory says he should have in order to understand the

name. Therefore, he does not know a priori that Feynman is the F for some

F that uniquely identifies Feynman. The speaker does not know the descrip-tion a priori that Feynman is the F , because he does not know that Feynman

is the F at all. So it cannot be descriptions in his mind that fix the reference

of the name as he uses it.

Now consider a case where someone comes along and tells our naïve

speaker, “Feynman is the man who originated the parton model.” Our

speaker clearly learned something from his informant, contained in the defi-

nite description about Feynman. However, as Kripke points out, this knowl-

edge is not a priori. According to the description theory, if a description is

synonymous with a name the corresponding statement should be known a

priori. But the person who heard that Feynman is the man who originated

the parton model knows something empirically about Feynman, not a pri-

ori. Kripke’s point is that for any description that a person associates with a

name, the description is always known empirically, not analytically. State-

ments that report such famous deeds are always synthetic, never analytic.

The second point Kripke makes is based on the Gödel–Schmidt example.

Many people who have heard of Kurt Gödel will know him as the math-

ematician who proved the incompleteness of arithmetic. Therefore, we

can refer to Gödel with the definite description “the mathematician who

proved the incompleteness of arithmetic.” Kripke asks us to suppose that

Gödel had not proved that theorem at all, but that it was instead proved

by an obscure figure named Schmidt. He also asks us to suppose, hypo-

thetically speaking, that Gödel had plagiarized his incompleteness theorem

from Schmidt, and Gödel had unjustly received the accolades for devisingthe proof.

In Kripke’s thought experiment, the man referred to when someone

says “the mathematician who proved the incompleteness of arithmetic”



Kripke on Names 47

is Schmidt, not Gödel. In this case, the speaker has a false belief about

Gödel—he thinks that Gödel invented the proof, but he did not. His false

belief about Gödel cannot then constitute the description that determinesthe reference of the name “Gödel” whenever he uses it. He refers to Gödel

with “Gödel,” while the description refers to Schmidt.

Another example of a Gödel–Schmidt type situation that Kripke does

not use is the case of perceiving an object. A description theory of seeing

would maintain that a description in the mind of the perceiver is what

determines which object is seen. Imagine the description here is very closely

related to the appearance of what is being seen. The appearance is like the

description, and the object and the viewer’s relation to it can be likened tothe object being referred to with the name. This description theory tries to

analyze the relation of seeing an object. That is, the object being seen is

determined by an appearance that is in the viewer’s mind, which translates

into a description.

The first objection to this theory is that there could be another object in

the world that is exactly similar to the one the viewer originally saw. So the

perceptual experience of the viewer cannot be the only determinant of the

object being seen, since there could be many such objects. The seen object

cannot be uniquely fixed by the person’s qualitative experience.

Equally, we are familiar with perceptual illusions that mirror the Gödel–

Schmidt case. Suppose someone views an object and he experiences a per-

ceptual illusion with respect to that thing. Does that mean he is not really

seeing that thing? No; he sees it, but his experience misrepresents it. Nor is

it the case that he really sees some remote object that fits his experience bet-

ter. The lesson is that what determines the object of perception is certainly

not the internal nature of the viewer’s experience by itself—this can misrep-

resent the object. The internal nature of the viewer’s experience plays a role,

but it is not the only factor that fixes the perception relation. The object

you are seeing is rather the one that actually causes you to have the visual

experience. The causal theory of perception proposes that the object being

seen is the thing that causes the perceptual experience. The object that best

fits one’s experience need not be the cause of the experience.

Consider referring with proper names along the lines of our perceptualexample. What fixes the object of reference is not merely what is going

on in the speaker’s mind in terms of descriptions. Rather, it is an exter-

nal relationship between the speaker and an object of another kind. This



48 Chapter 2

relationship could be of the causal kind, as in the perception case. Kripke’s

own theory later defends the view that the object of reference is what

causes one to use a name, not what best fits the description in the speaker’smind. This analogy to perception helps articulate the intuitive faults in the

description theory raised by the Gödel–Schmidt case and others like it.

If the objections Kripke raises through the Feynman and Gödel–Schmidt

thought experiments are correct, then it follows that the classic description

theory is incorrect. Descriptions in one’s mind cannot determine reference

because one might not have any definite description in mind (as in the Feyn-

man case), or the description might not fit the actual reference (as in the

Gödel–Schmidt case). Therefore, there is no such description that determinesthe reference of the name. This concludes Kripke’s argument against the

description theory, which includes the modal part and the epistemic part.

Though we have already considered some possible counterarguments to

the modal part of Kripke’s argument, the epistemic part looks extremely

convincing. However, since the description theory solves so many seman-

tic conundrums concerning names, we must ask what alternative theory

might to put in its place.

2.5 The Causal Chain Theory

If the description theory is incorrect, then the first question we must

address is how to solve Frege’s problem of the informative value of identity

statements, discussed in chapter 1, which Kripke barely mentions. But he

does put forward the chain of communication theory of naming. He argues

that we do not refer to something with a name by having a description in

our minds that picks out that object. Rather, naming is a much more social,

interactive phenomenon than that picture would suggest. Kripke suggests

that we must consider these social realities when someone is given a name.

We can refer back to our example of Aristotle being baptized. The baby,

Aristotle, is given a name, and then people who were present at his baptism

begin using his name. Five years later, say, people who have never seen

Aristotle may refer to him by name. Then, after decades of interaction with

people, Aristotle one day dies, but people still refer to him. Kripke thinksthat the reason why people can still talk about Aristotle after his death is

that they spoke to people who knew Aristotle, and then picked up their

reference from those people.



Kripke on Names 49

Kripke describes a historical situation in which each speaker is a link in a

chain, each intending to refer to the same person with the name “Aristotle”

as the previous person in the chain did. Here, the reference is preserved bythe intention to refer to the same person as the speaker referred to from

whom we originally got the name. This chain continues on and on through

the centuries, down until the present time, where any of us can say, “Aristo-

tle is a great philosopher.” So, we can refer back to Aristotle because of this

long chain of linguistic connections stretching back to his baptism.

Notice that Kripke emphasizes that it is not that a speaker has a descrip-

tion of this chain in mind; rather, being a link in the causal chain makes

one refer to that original individual. In other words, when referring to Aris-totle one does not need to have a description of him in mind but just to be

a link in the right causal chain. This example is somewhat like our example

of the perception case, except that it is social. In the case of perception, the

objects out there are causing the experiences in the viewer. Similarly, in

Kripke’s view, an object out there is causing this long chain of communica-

tion that causes one to say the name “Aristotle.” Because of that long causal

chain, anyone suitably connected to it can now refer to that person. The

description an individual has in mind does not matter in this case; rather,

being embedded in this causal chain with other speakers is what matters.

These speakers form a long chain going back in time to the point where

Aristotle was first called by the name “Aristotle.” This is the alternative pic-

ture Kripke paints for us as to how reference works and what determines it.

2.6 Objections to Kripke’s Critique

Kripke knows he is not giving a theory of necessary and sufficient condi-

tions, because the causal chain theory faces some prime facie problems.

However, he still believes that it paints a better picture of reference than the

description theory. He acknowledges the fact that the causal chain could be

interrupted at various points. There are many examples of this. Someone

along the chain might not intend to refer to the same person, or she might

make a mistake with the name, or somehow shift the reference of the name.

But the really troublesome issues that arise if we accept Kripke’s theory arethe problems about the sense of names, raised by Frege. If Kripke rejects

the description theory, then he does not believe that the sense of a name is

equivalent to a description. How then does he account for the informative



50 Chapter 2

value of “Hesperus is Phosphorus”? As an alternative theory, Kripke men-

tions John Stuart Mill’s view, that the meaning of a name is simply its bearer.

However, as we saw in considering Frege’s work, this view cannot handle thecase of “a= b,” where “a” and “b” refer to the same object (e.g., “Hesperus”

and “Phosphorus”). If the Millian view is true, then “a = b” has the same

cognitive content as “a = a”. Frege’s description theory solves that problem;

but Kripke, in rejecting the description theory, appears to be left with only

the Millian view, which does not adequately explain the sense of a name. It is

not as if in rejecting the description theory we can embrace a nice alternative

theory, the Millian theory—that just leads straight to Frege’s problem. We are

thus left with a nasty dilemma on our hands.Because of these difficulties, a second look at the description theory

is warranted to determine if Kripke’s arguments really refute it. We have

already covered objections to aspects of Kripke’s modal argument that

could resuscitate the description theory. However, Kripke’s epistemic argu-

ments require a different set of considerations. First, we could decide that

the description theory is a theory of sense, but not reference. Kripke has

refuted the use of the description theory to determine reference with the

Gödel–Schmidt example, but we could still suppose that the description

constitutes the sense of a name so far as its cognitive content is concerned.

On this approach, two names can have two different cognitive values, con-

tained in descriptions, without supposing that the descriptions that consti-

tute the cognitive value also determine the reference of the name. We can

think about it just like the case of perception. When one sees an object,

there is a cognitive, psychological component of experience and an extrin-

sic component of an object causing the experience. In the same way, there

could be a two-factor structure with names. The descriptions could be con-

sidered the cognitive, psychological content of the name, and the causal

chain could be what determines reference. According to this solution, we

take a two-factor approach to the meaning of names: the reference-deter-

mining part, along the lines of Kripke’s theory, and a more psychological

part that characterizes what is in a person’s mind when he understands the

name. Thus the description constitutes the psychological side of meaning,

but the referential side is determined by a Kripkean causal chain. This two-factor approach solves the problems raised by Frege, while still accepting

Kripke’s counterexamples. However, we still face the problem of answering

Kripke’s epistemic arguments against the description theory.



Kripke on Names 51

If Kripke’s epistemic arguments refute the description theory in its classi-

cal form, it would only be possible to have a description theory that some-

how accommodates the force of those arguments. In the Gödel–Schmidtthought experiment, one individual in the linguistic community refers to

Gödel by using the name “Gödel,” despite having in mind an incorrect

description of the reference. However, Kripke does not mention the fact

that certain members of the community do have in mind a uniquely iden-

tifying, correct description of Gödel. If language is as social as Kripke takes

it to be, then an individual who believes the wrong description of Gödel

is connected to other individuals who know correct descriptions of him.

Therefore, an individual’s reference is fixed by being part of a linguisticcommunity in which some people associate correct descriptions with the

name—though not all do.

2.7 The Social Character of Names

Kripke’s epistemic objections deal primarily with descriptions on the level

of an individual. But if the description theory is focused on the level of the

community instead of the individual, then the objections that applied only

to an individual with an incorrect description fall apart. According to the

socialized description theory, the reference of a name is fixed by the people

who have in their minds the correct description. We thus come to the idea

of linguistic deference. The people who are least knowledgeable about the ref-

erence of a particular name will defer to those who are most knowledgeable.

To illustrate deference and the social description theory, let us consider a

historical case similar to the Gödel–Schmidt example, which Kripke also

mentions. Giuseppe Peano was an Italian mathematician who axiomatized

arithmetic, so there are various axioms that are called “Peano’s axioms.”

However, according to authorities, Peano was not in fact the man who

invented those axioms. Richard Dedekind, another nineteenth-century

mathematician, proposed this collection of axioms, and Peano published a

more precise version of them. Peano had cited Dedekind’s work, but some

people wrongly attributed the axioms to Peano, and so they became known

as Peano’s axioms. Many people in our linguistic community thus have afalse belief about Peano. If someone uses the name “Peano” thinking that

he satisfies the definite description “the man who axiomatized arithmetic,”

that does not mean he is referring to Dedekind with “Peano.” The reason



52 Chapter 2

is that there are people in the community who know other correct descrip-

tions that apply to Peano, such as “the man who cited Dedekind’s inven-

tion of the axioms.” In this way, the description theory can be true for theprimary users of the name and the mathematical authorities, the people to

whom others defer in using the name “Peano.” The descriptions used by

the authorities overrule those of the odd misinformed speaker. The descrip-

tive beliefs of the authorities fix the reference of the name, not those of the

uninformed.

Another example that illustrates this point involves the scientific terms

that are used by nonexperts. Certain terms like “DNA” find their way into

popular culture, even though most people have a poor understanding ofthe terms. Although people use the term “DNA” all the time, few people

can refer to DNA with a unique scientific description and truly understand

it. However, the people who do not understand “DNA” borrow their refer-

ence from people who do have in mind an adequate description. If no one

had in mind a correct description of DNA, nobody could refer to it. When

a name comes into a language, its reference is fixed by the description that

introduces it into that language. Kripke himself does not deny this possibil-

ity, because he accepts that names can be introduced by means of a descrip-

tion. The fact that some people do not really know what names mean does

not show that those names do not have meaning—as with “DNA.” Kripke’s

epistemic arguments do not refute the description theory when the descrip-

tion theory is proposed as a theory of the language of a community. Kripke’s

arguments do not refute the description theory as modified to include this

social element, though they do refute the individualistic form of the theory.

We can say that a definite description determines the reference of the name

in a community, because people can defer linguistically.

2.8 Essential Descriptions

Given the additions and modifications to the classic description theory,

you may be wondering how it is possible to formulate the right kind of

description. Consider a person with brain B, such that whoever has that

brain is that person. The description “the person with brain B” cannot failto apply to the person who has that brain. Someone could say, “Aristotle

might not have been a famous philosopher,” and that is a true statement

because it expresses a contingency; but it is not contingent that Aristotle



Kripke on Names 53

had a particular brain. He must have that brain in all possible worlds since

it is a part of his individual essence. This argument can be made with a vari-

ety of personal identity theories. Consider the following descriptions: “theperson with soul S,” “the person with consciousness C,” “the person with

memories M ,” “the person with personality P .” These all express theories

about what a person essentially is. So, we can pick whichever personal iden-

tity theory best describes the essence of a person, according to our meta-

physical views, and express it in a description. For example, if an individual

consciousness is indeed the essence of a person, then a description “the

person with consciousness C” can be chosen as constituting the meaning

of a person’s name. This type of description is not vulnerable to any ofKripke’s modal arguments. In the case of the epistemic arguments, there

is always the option to defer to those members of the community who are

authorities on the subject—the metaphysicians of personal identity. In our

example above, the people who have not met the person with brain B will

be able to defer to those who have enjoyed such acquaintance.

In summary, we can generate descriptions that determine the refer-

ence of the name, provide necessary truths concerning the bearer of the

name, give the sense of the name (thus solving Frege’s pressing problem of

informative identity statements), and can be accommodated to deal with

Kripke’s epistemic objections. The underlying thought is that descriptions

refer to objects in the word descriptively, and then names are introduced

on their backs as abbreviations of those names—and that is how names

refer. The primary way to refer is through descriptions, and names are sec-

ondarily based on descriptions. We don’t need a separate account of name

reference. There is, however, a further objection to the description theory

to consider that Kripke does not bring up at all.

2.9 Impure Descriptions

Let us return to our example of the name “Aristotle” and the definite

description “the best pupil of Plato.” Notice that this description contains

a name, “Plato.” Many of these uniquely identifying descriptions contain

such names. But according to the description theory, all names are equiva-lent to descriptions. What, then, is meant by the name “Plato”? The name

“Plato” cannot abbreviate the definite description “the teacher of Aristotle”

because that definition would be circular. To refer to Plato, we must create



54 Chapter 2

a new definite description. We could say, “the most famous philosopher of

ancient Greece,” but then the question would arise as to what the name

“Greece” means. The point is that the uniquely identifying definite descrip-tions themselves contain another name. To explain what that name means,

the descriptions continue to regress to descriptions containing other names.

This issue raises serious problems for the description theory, since names

are supposed to depend ultimately on descriptions for their reference.

One type of description that can be used embeds a demonstrative, such as

“the owner of that dog.” Here we secure reference to the owner by referring

demonstratively to her dog. No name is used. So such a description might

give the sense of a name without itself containing a name. Demonstrativessuch as “this” and “that” are very important in language and are often used

to provide descriptive reference without the use of names. Without this use

of demonstratives, reference by means of descriptions would be crippled.

So descriptive reference depends upon and presupposes demonstrative ref-

erence. That means that demonstrative reference is basic. It cannot be ana-

lyzed in terms of purely descriptive reference. Therefore, demonstratives

are not short for demonstrative-free descriptions. We will be considering

demonstratives in detail in later chapters; for now we must note that the

description theory of names is not applicable to demonstratives.

Our conclusion, then, is that though it may be true that proper names

are equivalent to descriptions, those descriptions always in the end embed

demonstratives. Since demonstratives cannot be explained in terms of

descriptions, reference is not fundamentally descriptive. Even if the descrip-

tion theory is true of names, this does not show that the way we basically

refer to things in the world is through descriptions. The basic way we refer

to the world is by means of demonstratives, which are not equivalent to

descriptions. The victory of the description theory over Kripke’s attack is

therefore a Pyrrhic one. In the end, we must accept that some referential

terms function nondescriptively.







Russell on Definite Descriptions 57

meeting relation between someone referred to as “I” and someone referred

to as “a man.” This statement would express a relational proposition relat-

ing me to the person I met. It should have the form “a R b”—but if that istrue, and “a” and “b” are names, then contrary to appearances, “a man”

should be a name. Thus we might suppose that logically “a man” is a name,

though grammatically it clearly is not. But Russell thinks that this reason-

ing is incorrect; otherwise, as he says, the statement “I met a man but it was

not Jones” would be a contradiction, on the assumption that I met Jones.

The second point Russell makes is to the same end. Consider the sen-

tence “I met a unicorn.” If we thought that indefinite descriptions were

names, then there must be something that the name names in order tomake the name meaningful. In this case, there are no unicorns to name, so

the phrase “a unicorn” cannot function in that sentence as a name of some-

thing, or else it would be meaningless instead of merely false. In the previ-

ous example (“I met a man”) there was an actual man being met who could

possibly be the bearer the name. With the unicorn example, nothing in

reality can bear that name, so it would have to be a meaningless sentence.

You could never meet a unicorn, because there aren’t any unicorns to meet.

Russell’s point here is that if “a unicorn” were a name of something then

the name could be meaningful only if something were named. Since noth-

ing is named, it would lack meaning; but it does not lack meaning. The

only way the sentence can be false is if it is meaningful. Therefore, it cannot

be that “a unicorn” is a name of something. The thing that enters into the

proposition expressed by these words is not an object named. Instead, it is

the concept of a unicorn that is the constituent of the proposition expressed

by the sentence “I met a unicorn.” With respect to the “I,” what enters into

that proposition is not a concept but an object, because I am not a con-

cept. However, sentences like “I met a unicorn” or “I met a man” bring the

concept of a unicorn or a man into the proposition, not an actual man or

unicorn. According to Russell, then, in the example “I met a man,” “a man”

refers to a general concept, not to a particular man.

Russell uses the term propositional function to describe what is left in a

proposition when part of it is deleted. If I say, “I met Jones,” this is an

ordinary proposition whose constituents are me and Jones. However, ifwe delete the name and put in its place the letter “x,” then the letter “x”

does not refer to any individual at all. It is a placeholder that indicates a

part of the sentence has been deleted and left blank. The phrase “x is a



58 Chapter 3

man” is called a propositional function, because when something specific is

added to replace “x” (usually called a variable) the entire sentence expresses

a proposition. In essence, it is the abstract form of a proposition, with-out being a particular determinate proposition. In ordinary logic, “x” here

would be referred to as a free variable. The phrase with “x” will not become

a proposition until a name is plugged in to replace the variable.

Propositional functions can be simple or complex. Russell discusses the

sentence “I met x and x is human,” and takes it to mean “I met someone

or something and that someone or something is human,” or, more simply,

“I met something and it is human.” He explains that such a propositional

function is “sometimes true” if a proper name is inserted to replace “x.”He suggests that we replace the relational form (“a R b”) with the form of

this propositional function (“I met x”). Thus the propositional function

“I met x” is said to have an instance in which the resulting sentence is

true. If I met Jones and plug “Jones” in to the propositional function, the

sentence would be true. When someone says, “I met a man,” he is not

really talking about a particular man, according to Russell. Instead, Rus-

sell thinks that when someone says, “I met a man,” he is talking about a

propositional function and saying that it has an instance—though he may

not know what this instance is. It is important to note that any name could

be plugged into this propositional function. As long as the name refers to a

real person, the function has an instance, and is therefore true. Therefore,

there are two relations that Jones can have to a proposition to make it true.

One is that Jones can be named by a name in that proposition. But in the

other relation, Jones can be an instance of a propositional function—with-

out being named by it. To put it differently, Jones can either be explicitly

named or he can fall under a general term or predicate like “man I met.”

Falling under a predicate is not the same kind of relation as being named.

If I say, “Everyone in this room is a philosopher,” I have not named anyone,

even though several people fall under the predicate “person in this room

who is a philosopher.”

If we put it in more contemporary terms, Russell’s view is that indefinite

descriptions are quantifiers. Now we realize that quantifiers and names are

semantically not at all the same. For example, take the quantifier phrase “noone”: that cannot be a name of someone! If it were, “no one is over ten feet

tall” would entail “someone is over ten feet tall.” But neither is “someone”

a name for a person—for if so, who? Even if someone is out there making




true what the speaker is saying when he says, “someone stole my bike,” the

speaker is not naming that villain; if he were, he’d know who did it.

All of this has to do with the revolution in traditional logic that stretchedall the way back to Aristotle. In the old days, everything was just terms and

predicates. Russell rejects this traditional logic, just as Frege argued that

quantifier expressions (“something,” “everything,” etc.) should not be

assimilated to names. Frege’s position is that a quantifier word is a “second-

level concept.” He thought that these words were neither names of objects

nor concept expressions like “is a man.” A second-level concept applies to a

first-level concept. When a person says, “Someone is a man,” the quantifier

word is like a second-order propositional function: it is a comment about the first-level concept expressed by “man.” If a person says, “Jack is a man,”

then he is speaking of Jack and stating that he is a man. But if someone

says, “Someone is a man,” he is now talking about a propositional function,

stating that it has an instance. He is saying this: “The first-level concept

expressed by ‘is a man’ has at least one instance.” In Russell’s example of “I

met a man,” the correct analysis is this: “the propositional function ‘I met

x and x is human’ has at least one instance.” In this there is no mention of

Jones by name, even if he is the instance in question.

This has a bearing on statements about existence. When an atheist says,

“God does not exist,” what he is really saying is “The propositional func-

tion ‘x is a god’ has no instance.” He is not saying about some individual

named “God” that he does not exist—that would be self-defeating. Russell

argues that a person cannot make a true negative existence statement about

a named individual because he was never talking about any individual in

the first place. Instead, the speaker was really talking about a propositional

function and asserting that it has no instances. By paraphrasing the state-

ment into a statement about a propositional function, we are not mis-

led into believing that terms like “a man” or “someone” or “no one” are

somehow functioning like names that require a reference. The only thing

referred to with a propositional function is a concept, about which we state

that it has, or lacks, instances. The point that Russell is ultimately building

up to is that a definite description is also a quantifier, not a name. In adopt-

ing this approach, Russell thereby resolves many puzzles that arise withdefinite descriptions, particularly when they are empty.

Russell had previously held Alexius Meinong’s view. This is the view that

in addition to the ordinary objects that exist, there are things that subsist , or



60 Chapter 3

have a peculiar quasi existence. Things that people normally do not think

exist, such as unicorns and golden mountains, have this quality of sub-

sistence. Because of this subsistent category, Meinong thinks that expres-sions like “the golden mountain” do refer to things, and because they have

reference they also have sense. This view is in contrast to Frege’s view that

such terms have sense but not reference. With Meinong’s view, “the golden

mountain” is meaningful because it refers to the golden mountain, which

is a subsistent thing. In his system, such expressions can be endowed with

reference, so long as we accept this expanded ontology of subsistent enti-

ties. Russell now avoids this view by developing a theory of descriptions

that does not postulate any Meinongian ontology in order to give meaningto empty definite descriptions. Russell believes that such phrases do not

denote anything, even when they have an existent correlate. It is the same

point he makes about the phrase “a man”—the definite description is not

a phrase that functions like a name at all. Cases where there is no entity to

denote (e.g., “the golden mountain”) do not require an extra ontology like

Meinong’s. Rather, we say that the expression is not a denoting phrase to

start with, but something completely different from that, just as “a man” is

not a denoting phrase. Russell argues that definite descriptions also express

propositional functions that do not refer to or denote or name objects. As

Frege would put it, they function as quantifiers. Therefore, since quantifiers

are different from names, definite descriptions are different from names.

Russell’s new theory is developed against the background of Meinong’s

theory, which is also a version of Frege’s theory in assuming that definite

descriptions function as proper names.

3.2 Three Theories of Definite Descriptions

Before continuing with a thorough analysis of Russell’s theory, it is impor-

tant to note that Russell does not follow the rules on when things should

be quoted or not. Indeed, Russell is notorious for his misuse of quotations.

We should be more careful.

There are three theories about definite descriptions relevant to Russell’s

“Definite Descriptions.” We can use Russell’s first example, “the king ofFrance,” to explain these three theories. The description “the king of France”

is an empty description—one with no reference—because at the time Rus-

sell used the example, France had no king. Although this description is




empty, it is just as meaningful a description as “the queen of England,”

though that description does have reference. The fact that there are mean-

ingful empty descriptions refutes the idea that the meaning of a definitedescription is identical to its reference. If reference and meaning were iden-

tical, then our first example would have no meaning.

Frege’s theory is consonant with this fact, because it allows that such

expressions have sense but no reference. Of course, the sense is where the

meaningfulness lies. As far as we can tell from Frege, he believes that every

meaningful expression has a sense, and there are no expressions whose

meaning is simply their reference. Every expression that exists in natural

language is something whose meaning consists in its sense, where the senseis independent of the reference. Russell, in his discussion, never takes into

account Frege’s view. Some readers could potentially be confused reading

this excerpt alone, because Russell is constantly making assertions that

contradict Frege’s theory. Russell presupposes that Frege’s theory is wrong

without overtly stating his rejection of the theory of sense and reference.

Instead, Russell holds a referential theory of meaning, where he believes

that the meaning of an expression must be its reference.

Meinong’s view is that “the king of France” has a reference to a peculiar,

subsistent entity. Its reference is not something that exists in the same way

that the reference of “Queen Elizabeth II” exists. In Meinong’s ontology,

the world is divided into existent things and nonexistent things, and non-

existent things also have a kind of being. Given his distinction between

existence and subsistence, it is possible for Meinong to argue that “the king

of France” refers to a subsistent being. By considering fictional characters,

Meinong’s view becomes easier to understand. According to him, the name

“Hamlet” refers not to any existent Prince of Denmark but to a fictional

character. In Meinong’s theory, such fictional characters have being with-

out existence—subsistence. Therefore, a name like “Hamlet” refers to a

subsistent entity. With this theory, a referential theory of meaning can be

maintained, without adopting Frege’s distinction between sense and refer-

ence. If an expression is meaningful because of its reference, we have no

need to bring in sense to establish meaning, because we always have subsis-

tent references when existent references are lacking.According to Russell, every proper name or singular expression has a

meaning determined by its reference. He does not accept a two-level the-

ory of reference and sense; he thinks he can get by with reference alone.



62 Chapter 3

Contrary to appearances, he argues, a definite description is not a singular

term at all and does not denote an object. Frege thinks that empty descrip-

tions like “the king of France” have no reference but that such expressionsare meaningful because they have a sense. Meinong thinks they refer to

subsistent entities and are meaningful that way. Russell thinks they are not

referential expressions, so their emptiness isn’t a problem.

As mentioned above, Russell was a Meinongian in his earlier years. But

since he liberated himself from trying to find a reference for empty descrip-

tions, he does not have to reconcile himself to accepting shady subsistent

entities. He thinks that ordinary language is logically misleading, because

it makes definite descriptions occupy the place of names. For example, inordinary language, the sentences “The king of France is bald” and “Ber-

trand Russell is bald” are both subject-predicate sentences. The first one has

a definite description as the subject while the second has a name for the

subject. Ordinary language makes it seem as though definite descriptions

function as proper names, even though logically they do not.

Quantifier expressions also illustrate this point. The sentence “Someone

is bald” appears to express a subject-predicate proposition in the same way

that “Bertrand Russell is bald” does. These two expressions look grammati-

cally and syntactically the same. However, it would be very strange to think

that “someone” is a name (“Someone, come here!”). Consider the claim

that “someone” denotes Jones in the sentence “Someone is bald,” where

Jones is in fact bald. But “someone” cannot be the name of Jones, because

the statement “Someone is bald but it’s not Jones” is not a contradiction,

even though Jones may be the only bald person. The appearance of subject-

predicate status for “Someone is bald” has to be misleading.

At the same time, it is not plausible to think that “someone” refers to

some shadowy, ideal, possible bald individual, as Meinong supposes. Rus-

sell argues that terms like “someone” are logically not singular terms. Part

of his purpose is to explain what their logical role is. Since we have seen

that these sorts of terms are not referring expressions at all, their meaning

cannot be constituted by reference. Because of the defectiveness of ordinary

language, these sorts of statements are misinterpreted as having subject-

predicate form. However, the fact that such terms lack a singular referencedoes not mean that they lack meaning.

Frege and Meinong have their own explanations as to why such terms

as “the king of France” lack an existent reference but have meaning. Frege




uses his sense–reference distinction, and Meinong postulates an existence–

subsistence distinction. Russell rejects both of those ideas. He thinks that

every expression that is referential has a meaning that is determined by itsreference, but these sorts of expressions are not referential at all. However,

Russell accepts that these sorts of expressions appear to be referential, owing

to the deceptiveness of natural language. This point about the deficiencies

of natural language was very important to Russell, because it showed that

ordinary language can be logically misleading and bears on the question of

constructing an ideal logical language. In Principia Mathematica, Russell and

Alfred North Whitehead formed an ideal language that is essentially the

same as predicate logic. The formation of this logical language led to theidea that natural language was adequate for practical purposes but deficient

for logical ones. This view was the standard one for a long time and shaped

philosophy for the first half of the twentieth century—until Ludwig Witt-

genstein came along and argued against this view, which he had also held

in his Tractatus Logico-Philosophicus. So this issue about descriptions had

wide philosophical implications.

It is important to understand the context within which Russell produced

this work. Much of the correct methodology in twentieth-century philoso-

phy and expectations about language hung on the theory of descriptions,

in addition to its contributions to pure logic. Russell’s theory practically

shaped the whole of twentieth-century analytical philosophy. The resulting

dialogue of twentieth-century philosophy revolved around whether phi-

losophers agreed with it or were against it. So, the theory was of massive

importance at the time Russell developed it.

3.3 Indefinite Descriptions and Identity

Russell’s position is that statements containing descriptions like “a man”

must be paraphrased to reveal their meaning. This will involve changing

their form quite dramatically, and also introducing logical symbols. To

paraphrase such statements, he uses propositional functions to take partic-

ular expressions out of a sentence and substitute a variable “x.” In this case,

he would insert an “x” into “a man,” creating the propositional function“I met x and x is human.” This propositional function is then said to have

at least one instance, meaning that it applies to at least one thing in the

world. Jones is the instance out of all those things in the world that might



64 Chapter 3

make that propositional function true. Although the sentence appears to

refer to a certain individual in the world with the expression “a man,” the

original sentence’s form is logically misleading. For Russell, what the sen-tence really says is that the particular propositional function has at least

one instance. He uses this apparatus of explanation to make it philosophi-

cally clear that this sentence is about a propositional function.

Today we routinely use quantifiers to express Russell’s point. Take, for

example, the long logical form:

(1) There is an x such that I met x and x is human.

The same propositional function can have several variations. It could alsobe read existentially:

(2) There exists an x such that I met x and x is human.

Different theories of the quantifiers correspond to the ways in which such a

statement is read. A useful thing to remember about interpreting existential

quantifiers is that the variable “x” can be replaced by a name. After such a

substitution, there will be at least one instance where the substitution will

make the statement true. In our particular case, “Jones” would make the

statement true. Such an analysis is often called the “substitutional interpre-

tation” of the existential quantifier, because a particular substitution into

the open sentence expressing a propositional function makes the result-

ing sentence true. Russell tends to assume the substitutional interpretation.

The best way to understand his interpretation is with the sentence “I met

something and that something is human.” The only term in this sentence

that refers to an individual is “I.” The phrase “a man” becomes a part of

the existential quantifier. Then, there is a conjunction of the predicatesgiving us the assertion about my meeting a human. The only things that

are brought in by the quantifier phrase are concepts. To better illustrate

this point, we can use a statement involving a nonexistent entity: “I met

a unicorn.” Since there are no unicorns, I could not have met a unicorn.

However, when using Russell’s apparatus to analyze this sentence, we see

that the proposition contains only me and the property of being a unicorn.

The sentence is actually saying (falsely) that there is an instance of that

property and that I met that instance. In this form, no unicorn is named.

The advantage of Russell’s theory is that we can explain how to speak

of nonexistent things without creating an entirely new ontology. In Mei-

nong’s view, we need subsistent golden mountains to analyze “I climbed




the golden mountain.” Russell avoids creating an entirely new ontology

of subsistent things, because he thinks that the statement is really about a

propositional function. Russell argues that genuine names that are emptyare indeed meaningless, but “the golden mountain” is not a genuine name.

Russell just assumes that Frege is wrong, because he assumes that the mean-

ing of a name comes from its reference if it really is a name. In contrast to

Frege, Russell also sharply distinguishes between names and descriptions.

He thinks that descriptions, definite and indefinite, do not function in the

way that names do.

Russell includes a few paragraphs on the important distinction between

the “is” of predication and the “is” of identity, which we shall briefly pauseto explicate. Although these points are not really essential to his main line

of argument, they have major significance in analytic philosophy. He says

there are two kinds of “is”: the “is” of identity and the “is” of predication.

The “is” of identity is used in sentences that could be paraphrased as “a

= b,” like “Hesperus is Phosphorous.” Russell points out that we do not

always use “is” in the sense of identity. Take the statement “This table is

brown.” The table has the color brown, but the identity of the table is not

brown. There are a great many things in the world that are brown and not

just this table. It would be absurd to claim that this table is identical to the

color brown! According to Russell, the “is” that is present in “this table is

brown” is the “is” of predication. The “is” used in the sentence “Socrates

is human” is thus entirely different from the “is” used in the sentence

“Socrates is a man.” The former is the “is” of predication, and the latter is

the “is” of identity. He gives us the following paraphrase of the sentence

with the identity “is”:

(3) There is an x such that Socrates is identical to x and x is human.His general point is that we must be aware of the two different forms of “is”

in language. Also, the ambiguity of “is” adds further evidence to his point

that ordinary language is logically misleading, because this one word—

“is”—is used in both statements of predication and statements of identity.

Russell believes that an ideal language would not have such ambiguities.

3.4 Russell’s Rejection of Meinong’s Ontology

Russell’s stalwart rejection of the Meinongian ontology can be found in the

following impassioned passage:



66 Chapter 3

For want of the apparatus of propositional functions, many logicians have been driv-

en to the conclusion that there are unreal objects. It is argued, e.g., by Meinong, that

we can speak about “the golden mountain,” “the round square,” and so on; we can

make true propositions of which these are the subjects; hence they must have some

kind of logical being, since otherwise the propositions in which they occur would

be meaningless. In such theories, it seems to me, there is a failure of that feeling for

reality which ought to be preserved even in the most abstract studies. Logic, I should

maintain, must no more admit a unicorn than zoology can; for logic is concerned

with the real world just as truly as zoology, though with its more abstract and gen-

eral features. To say that unicorns have an existence in heraldry, or in literature, or

in imagination, is a most pitiful and paltry evasion. What exists in heraldry is not an

animal, made of flesh and blood, moving and breathing of its own initiative. What

exists is a picture, or a description of words. Similarly, to maintain that Hamlet, for

example, exists in his own world, namely, in the world of Shakespeare’s imagina-

tion, just as truly as (say) Napoleon existed in the ordinary world, is to say some-

thing deliberately confusing, or else confused to a degree which is scarcely credible.

There is only one world, the “real” world: Shakespeare’s imagination is part of it, and

the thoughts that he had in writing Hamlet are real. So are the thoughts that we have

in reading the play. But it is of the very essence of fiction that only thoughts, feel-

ings, etc., in Shakespeare and his readers are real, and that there is not, in addition to

them, an objective Hamlet. When you have taken account of all the feelings roused

by Napoleon in writers and readers of history, you have not touched the actual man;

but in the case of Hamlet you have come to the end of him. If no one thought about

Hamlet, there would be nothing left of him; if no one had thought about Napoleon,

he would have soon seen it that someone did. The sense of reality is vital in logic,

and whoever juggles with it by pretending that Hamlet has another kind of reality is

doing a disservice to thought. A robust sense of reality is very necessary in framing

a correct analysis of propositions about unicorns, golden mountains, round squares,

and other such pseudo-objects.2

We can clearly see the force of Russell’s point here. To say that Hamletis an existent in Shakespeare’s imagination or our own imaginations is a

confused way of speaking. Hamlet, Russell argues, does not have the same

existence in our imaginations as you have as you are reading the text. At

most, the sentence “Hamlet has an existence in Shakespeare’s imagination”

can mean that Shakespeare invented the fictional character of Hamlet. The

sentence does not mean that we can go to a place called “Imagination” and

find Hamlet skulking there, existing like one of us does in reality. Herein

lies another misleading aspect of ordinary language. The sentence “There

exists a dog in the next room” allows the listener or reader to understand

its meaning—she will see a dog in the next room if she goes into that room.

However, the sentence “There exists a dog in my imagination” makes it




seem as though imagination is a place to which people can travel and, upon

arriving there, we will find a dog, barking and wagging its tail. This notion,

Russell argues, is ludicrous; a dog or a unicorn does not exist in one’s imagi-nation in the same way that a horse exists in a paddock.

As to whether the passage above refutes Meinong’s position, we cannot

yet say. Meinong never says that phrases like “the golden mountain” refer

to things that have an existence. His whole argument is based on the thesis

that they have only subsistence. Meinong never states that things exist in

the imagination in the way normal people exist in towns and cities. Really,

Russell is arguing against what he thinks Meinong is proposing, not what

Meinong is actually proposing. However, for the sake of understanding Rus-sell’s theory, we will assume that he is correct about how we should deal

with definite descriptions that refer to these nonexistent entities—that is,

they have no reference at all.

3.5 The Details of Russell’s Theory of Descriptions

The theory of descriptions is now quite simple. An indefinite description

like “a man” is equivalent to the existential quantifier. The reader may

now be wondering how Russell distinguishes a definite description from

an indefinite one. Suppose we start with the indefinite description in “A

present king of France is lucky.” We could paraphrase that sentence in the

following way: “There exists someone x such that x is a present king of

France and x is lucky.” Russell then asks us to consider a case where the

sentence has “the king of France” as a component. The difference lies in

whether uniqueness is implied. In the sentence “I met a man” the speaker

of the sentence does not logically imply that he met just one man. Such

descriptions with “a” can apply to many men. On the other hand, a defi-

nite description with “the” (e.g., “the king of France) can only apply to

one individual if it applies to any. Therefore, uniqueness is what is added

when “a” is replaced by “the.” Russell thus argues that definite descriptions

should be analyzed in basically the same way that indefinite descriptions

are analyzed; the only difference in the analysis of definite descriptions is

that uniqueness is added. Keeping these considerations in mind, we willfirst examine an analysis of an indefinite description; then we will examine

an analysis of a definite description. So consider “An F is G” and “The F is

G.” The former is true if and only if at least one thing is both F and G. The



68 Chapter 3

latter is true if and only if at least one thing is F and that thing is G and at

most one thing is F and that thing is G. Both imply existence, expressed by

“at least,” but only the second implies uniqueness, expressed by “at most.”If we analyze the sentence “The queen of England is happy,” we would say

that there is a queen of England, and that there is only one queen of Eng-

land, and that she is happy.

There are three conjuncts in this analysis of “The F is G”: (1) there exists

something that is F , (2) there is only one thing that is F , and (3) that thing

is G. Therefore, if you uttered the sentence “The king of France is bald,” you

would be saying that there exists something that is a king of France and

that there is at most one thing that is a king of France and that thing is bald.That is Russell’s general form for the analysis of the statement “The F

is G.” His theory is fairly straightforward. The basic idea is that the word

“the” means existence and uniqueness. Existence means at least one, but

uniqueness means at most one, and then the particular predication (“is

bald”) follows. Thus, Russell’s interpretation of definite descriptions begins

in grammatical form with the simple phrase “the F .” It is then paraphrased

as a conjunction of existence and uniqueness, thus producing a more com-

plex linguistic form. This logical form is quite different from the apparent

form in ordinary language, where “the F ” is not a conjunction at all. The

definite description disappears as a singular term in this analysis, and so it

has no reference assigned to it.

A side note on a slightly technical part of the Russellian analysis: there

are two ways of logically analyzing uniqueness. One is with this notation:

“∃!x ( Fx and Gx),” read “There is a unique x such that Fx and Gx.” This is a

very easy and convenient way to build uniqueness into the quantifier. In

that way, we have specified uniqueness without an analysis: we just use “!”

as a primitive symbol expressing uniqueness. But there is also another nice

way of analyzing uniqueness in logical vocabulary. Consider the following:

(4) There is an x such that Fx and for all y if Fy , then x = y and Gx.

In plainer language, this analysis is saying the following: “There is an x

such that x is a king of France, and for any object y , if y is a king of France

then y is identical to x, and x is bald.” This is a way of saying that someone

is uniquely king of France and bald. We are saying, intuitively, that if there

is anything else in the world that is a king of France, then it is identical to

the first thing. That implies that there is not more than one thing, because




anything else that is a king of France is just the first thing. Such is the stan-

dard way for expressing uniqueness using ordinary quantifier logic with

identity. It is not essential to understanding the theory. Rather, it is one wayto analyze what uniqueness means. Uniqueness just means “at most.” This

part of the theory, using standard logic, is not essential to Russell’s basic

idea—it is just one explanation of what uniqueness is.

As we have seen, Russell thinks that definite descriptions are not proper

names, despite the fact that in some ways they appear to be proper names.

Once the philosopher of language realizes that grammar is logically mis-

leading, he or she can then have a theory that is not logically misleading.

According to Russell, we do not need to postulate in our theory of meaninganything more than the reference of terms, once our sentences have been

fully analyzed. Russell is a kind of Millian about genuine proper names,

because he believes that ultimately expressions mean what they do in vir-

tue of referring to what they refer to.

If Russell does not believe that definite descriptions are proper names,

we may wonder what proper names are for him. Russell does think there are

proper names, but he has a peculiar set of criteria for them. As before, one

of his points is that the words that appear in language to be proper names

are not actually proper names, because language is logically misleading.

Therefore, a name like “Bertrand Russell” will occur in a language though

it is not a proper name at all. Russell advocates the description theory of

names and considers such names to be the equivalent of a description. He

takes a name and gives a paraphrase of it, turning it into a description (e.g.,

“the author of Principia Mathematica”), and then analyzes the description

by his theory of descriptions, thereby eliminating the name as a name.

According to Russell, none of the names of ordinary language is a logi-

cally proper name. They are all fake names—they all appear to be names,

but they are not actually names. His view is that all the standard words we

consider to be proper names in natural language are all disguised definite

descriptions, and those descriptions are all analyzed away by the theory

of descriptions. Following his theory, these descriptions do not have their

meaning in virtue of their reference; so neither do ordinary proper names.

Russell does believe that there are words that can have meaning in virtueof their reference, but those are what he calls logically proper names. Logi-

cally proper names are meaningful in virtue of what they refer to. Our ordi-

nary names are not logically proper names, however, because they do not






he suggests that a logically proper name is a demonstrative, because a

demonstrative can refer directly to sense data. In Russell’s view, one cannot

refer directly to material objects since the material object might not exist(the viewer could be hallucinating the object). Therefore, the only logi-

cally proper names are phrases like “that black patch you are now seeing,”

where this refers to a subjective sense datum. According to Russell, the only

logically proper names are demonstratives, and they can only refer to sense

data. This certainly seems odd; we don’t usually classify demonstratives as

names. When did you last call one of your sense data by its proper name?

Have you ever referred to a sense datum as, say, “Phil”?

If we look back at our discussion of Frege, we may have a few questionsin regard to Russell’s Millian theory. For instance, how does Russell’s idea of

logically proper names work with identity statements? Russell never talks

about that, perhaps because he is very concerned by the question of exis-

tence. Frege’s main concern is with identity. Russell does not have anything

to say here about identity statements. He assumes that two logically proper

names of the same thing have the same meaning, because the meaning of

a proper name is its bearer. Russell is committed to the position that an

identity statement linking two logically proper names must be a tautology.

Russell avoids an obvious objection here by avoiding the question of Hes-

perus and Phosphorus.

Russell’s position as to how to handle an identity statement that links

two logically proper names is that two nonsynonymous logically proper

names, in his system, cannot designate the same object. Names can dif-

fer in their meaning while referring to the same thing only if they are not

really names. If they are names, as Russell defines logically proper names

to be, then they cannot differ in their meanings while co-denoting. The

identity statement must contain demonstratives that refer to sense data.

Of course, it will be a false identity statement if the reference is to two dif-

ferent appearances. For the viewer, Hesperus elicits different sense data in

the morning than Phosphorus does in the evening. Because these represent

two entirely different pieces of sense data, they do not fit Russell’s strict

criteria for logically proper names. Thus “Hesperus” is not a name, for him,

but “this sense datum of a luminous point” is. In Russell’s system, there areno identity statements that are informative and contain ordinary names.

One important consequence of Russell’s theory that generated much

discussion is how he handles truth-values. According to Russell, the



72 Chapter 3

truth-value of the sentence “The king of France is bald” is false. It is natural

to assume that this statement would be false only if the subsistent Mei-

nongian king of France had hair. Russell does not think along these linesat all. He believes that any statement containing that description is false,

because the king of France does not exist. In his handling of truth-values,

the sentence “Sherlock Holmes is a detective” is false, because it logically

implies the real existence of Sherlock Holmes. In his famous article “On

Referring,” P. F. Strawson objected to this point, arguing that such a state-

ment is neither true nor false, because there is no king of France to be bald

or not be bald. The only way for that sentence to be true would be by the

king of France being bald, and the only way it could be false is by the kingof France having a good head of hair. Since neither of those things is the

case, the statement “The king of France is bald” must to be neither true nor

false. But Russell’s analysis implies that it straightforwardly false.

3.6 Problems with Russell

Although in the previous sections we explained Russell’s analysis, we have

not yet discussed whether or not this analysis is correct. The following pas-

sage is an excellent summary of what we discussed in the previous sections:

We may even go so far as to say that, in all such knowledge as can be expressed in

words—with the exception of “this” and “that” and a few other words of which the

meaning varies on different occasions—no names, in the strict sense, occur, but

what seem like names are really descriptions. We may inquire significantly whether

Homer existed, which we could not do if “Homer” were a name. The proposition

“the so-and-so exists” is significant, whether true or false; but if a is the so-and-so

(where “a” is a name), the words “a exists” are meaningless. It is only of descrip-tions—definite or indefinite—that existence can be significantly asserted; for, if “a”

is a name, it must name something: what does not name anything is not a name,

and therefore, if intended to be a name, is a symbol devoid of meaning, whereas a

description, like “the present king of France,” does not become incapable of occur-

ring significantly merely on the ground that it describes nothing, the reason being

that it is a complex symbol, of which the meaning is derived from that of its constitu-

ent symbols. And so, when we ask whether Homer existed, we are using the word

“Homer” as an abbreviated description: we may replace it by (say) “the author of the

Iliad and the Odyssey .” The same considerations apply to almost all uses of what looklike proper names.4

In this passage, Russell makes three major points. He defines a name as a

simple symbol whose meaning is its reference. A name without reference




would lack meaning. Calling a name “empty” is a contradiction in terms,

because a name without reference is not a name. Russell also believes that

descriptions are quantifiers and that ordinary “names” are equivalent todescriptions. The only reason why ordinary proper names appear to be

names is because of the infirmities of natural language.

Russell’s conception of genuine names has an obvious consequence for

existential statements. He believes that existential statements are highly

misleading because they appear to contain names when they do not. State-

ments like “a exists” look like they contain the proper name “a.” There are

two possibilities for this type of statement. First, if the name does refer to

something, then the meaning of the name guarantees that the name hasa reference. Therefore, adding “exists” to the name is stating a tautology,

because names in Russell’s system will refer only to things that exist. We

can create an example to illustrate this point. If someone looks up outside

and says, in reference to the color of the sky, “That shade of blue exists,”

he knows that that shade of blue exists, because it is an aspect of a sense

datum. To add that the color exists is unnecessary, since that is understood

in virtue of grasping the name alone.

The second possibility arises if the name does not refer to anything. If

the name does not refer to anything, then the statement containing it must

be a meaningless statement with a meaningless part—and hence not a real

statement. Take the sentence “a does not exist.” Since the name “a” does not

refer, we can say that it is empty . The problem with that alleged statement,

“a does not exist,” is that it cannot be true since the name lacks reference

and is therefore meaningless. According to Russell, existential statements

cannot be applied to names. On the other hand, existential statements can

be applied to descriptions, because in the case of descriptions they do not

need reference in order to have meaning. Existential statements will never

contain names. In Russell’s system, names must refer to have meaning, so it

is trivial to say that their reference exists because it will always have to exist.

Russell is making a very radical proposal. The thought behind this pro-

posal is that there are propositions that lurk behind sentences and each

proposition has a kind of intrinsic logical form. It is as if these propositions

are clothed in the sentences of ordinary language, but the clothing is verymisleading as to the real form of the proposition. The job of the philosopher

is to slip beneath the clothing and discern the real nature of the proposi-

tion. Then, he is able to devise a notation to reflect that nature. Russell’s



74 Chapter 3

proposal led to the idea that philosophers needed to devise a logically perfect

language that reveals the actual structure that is hidden behind ordinary

language. In our example of “a

exists,” the sentence looks like a subject-predicate sentence like “a is red,” but in actuality it is a quantifier sentence.

Therefore, the underlying proposition is of a completely different kind than

that expressed by the sentence “a is red.” One of the reasons why Russell’s

analysis of descriptions was so important was that it initiated discussions on

the possibility of creating a logically perfect language. Many philosophers

believed that such a logically perfect language could solve all philosophical

problems. In particular, a completely logical language could solve ontologi-

cal problems, ridding us of the shadowy ontology of Meinong.For example, consider the ontological proof for the existence of god:

God has all perfections, and one of those perfections must be existence, and

therefore God must exist. According to Russell, this presupposes that exis-

tence is a predicate. In other words, subject-predicate sentences like “God

exists” would assign a predicate to something named by “God.” According

to both Russell and Frege, that sentence is not a subject-predicate sentence at

all, because the word “exists” is not a predicate. The idea is that existence is

not a predicate or a property of things, like being red. Rather, it is a second-

order concept that is really a property of a propositional function. Therefore,

the ontological argument is unsound. To resolve philosophical problems, we

must reform language so as to reflect the hidden form of propositions.

3.7 Primary and Secondary Occurrences

So far, we have only considered sentences of the form “The F is G.” We may

wonder how Russell handles sentences of the form “The F is not G.” He

argues that such sentences are ambiguous. To understand his point, we can

consider a case where the “not” applies to a predicate, for example, “The

queen of England is not pregnant.” Here we are attributing nonpregnancy

to Her Majesty. But instead of placing the negation sign immediately before

“G,” we could place it at the beginning, creating the sentence “It is not the

case that the queen of England is pregnant.” If we translate this into Rus-

sell’s analysis, we get the negation of the existential clause: “It is not thecase that at least one thing is a queen of England.” This sentence is express-

ing the proposition that it is not the case that a queen of England exists.




Let us now consider an example where the description is empty: “It is

not the case that there is at least one king of France.” By negating the exis-

tential statement that there is a king of France, the statement becomes true.Since it is not the case that there is at least one king of France, the sentence

“The king of France is not bald” will be true when interpreted this way.

But under the first interpretation, the sentence will not be true. These two

propositions have different truth-values. Thus, the truth or falsity of the

sentence depends on at what point the negation is inserted. In the latter

case, the entire sentence was negated; in the former case, only the predicate

was negated.

Consider the sentence “It is not the case that there is a queen of Eng-land and she is pregnant.” Since there is a queen of England, this sentence

is false. On the other hand, if “not” were placed before the predicate, the

sentence would be true (since the queen of England is not pregnant). To

handle this kind of ambiguity, Russell brings in the concepts of primary and

secondary occurrence. A primary occurrence of the description happens

when the negation occurs before the predicate. A secondary occurrence of

the description happens when the negation applies to the whole sentence

including the description. To illustrate this point more clearly, we can bring

in the concept of the scope of negation from logic. In the primary occur-

rence, negation has narrow scope; in the secondary occurrence, negation

has wide scope and thus applies to the description. The scope merely tells

you what is included in the negation. Are we negating the whole proposi-

tion or just the part of it that corresponds to the predicate?

This point about negation also applies to necessity. Like negation, neces-

sity has a similar kind of ambiguity. One might wonder how to read the

sentence “The queen of England is necessarily pregnant.” It could be read

either as “Necessarily there exists a queen of England and only one and

she’s pregnant” or as “There exists a queen of England and only one and

she’s necessarily pregnant.” In the former case the modal operator has wide

scope; in the latter, narrow scope. These can have different truth-values.

When these sorts of operators like negation, necessity, or possibility occur

in sentences containing descriptions, the scope determines the logical

interaction between the operator and the description. This interaction canget quite complex if the sentence contains multiple operators.

This concludes our discussion of Russell’s theory of descriptions. In the

next chapter we will look at some possible criticisms of Russell.





4 Donnellan’s Distinction

4.1 Introduction

To summarize our progress so far, we have examined two major theories of

descriptions: Frege’s theory and Russell’s theory. In Frege’s theory, descrip-

tions are proper names referring to things. In Russell’s theory, logically

proper names refer, but descriptions do not—they are analyzed in terms of

quantifiers. In a case in which a description fails to apply to anything, these

two theories have different consequences. For Russell, statements made

using descriptions without reference (e.g., “The king of France is bald”) are

always false, since they assert existence. Since the sentence expresses in part

the proposition that there exists a king of France, and there is no king of

France, the truth-value of the sentence is false. According to Frege, such a

sentence would be neither true nor false. If the description refers to some-

thing, the sentence is true if the predicate applies to the object to which the

description refers. The condition for it to be false is that the thing referred

to by the description does not satisfy the predicate. However, if the descrip-tion does not refer to anything, it can be neither true nor false. Therefore,

it is not true of every proposition that it is either true or false. P. F. Strawson

is famous for making this idea of “truth-value gaps” explicit in his paper

“On Referring.” The point becomes clearer when we consider an example

involving names. Take an ordinary proper name used in a statement. If that

name refers to nothing at all, we would not conclude that the statement

is false, because there is no reference to fail to satisfy the predicate. It is

neither true nor false. These two theories are intended to give a uniform

account of the meaning of definite descriptions whenever they occur. They

are theories of the “inner logic” of descriptions.



78 Chapter 4

We shall see that Keith Donnellan disagrees with both of these camps.

According to Donnellan, these uniform accounts of the semantics of defi-

nite descriptions cannot give an analysis of definite descriptions as they areused in every statement. He proposes that definite descriptions can func-

tion in two different ways. In some statements, they function in the way

Russell claims, and in other statements they function in the way Frege and

Strawson claim. Donnellan does not reject their views completely, but he

thinks that neither one nor the other theory covers the semantics of all

definite descriptions.

According to Donnellan, there is a third possibility as far as truth-values

are concerned. Russell thinks that an empty description gives rise to a falsesentence. Frege thinks it gives rise to a sentence that is neither true nor

false. Donnellan thinks that an empty description can give rise to a true

statement. Thus, he offers a third possibility. His reasons will emerge as we

go on.

The general point Donnellan makes with his examples is that descrip-

tions can work in more ways than the uniform ones recognized either by

Russell or Frege/Strawson. The theories we have examined so far analyze

only the semantics of language. Donnellan believes that to have a more

complete theory of language, we must include the pragmatics of language.

Semantics is about the abstract analysis of language independently of

speakers, whereas pragmatics examines language in relation to speakers

in concrete speech situations. Donnellan’s critique forms part of a more

general movement toward the analysis of speech acts in the understand-

ing of language. We have to look at what speakers do with words and not

just at the words themselves. Donnellan believes that our views about how

descriptions function in acts of communication will change if we examine

the role of descriptions in speech acts.

4.2 Referential and Attributive Uses

Donnellan calls the view of Strawson and Frege the referential view of

descriptions, because they take the position that descriptions are referen-

tial, namelike devices. Since Russell’s stance is that a definite description isa quantifier, we could label Russell’s theory the quantifier view of descrip-

tions. But Donnellan chooses to call it the attributive view. The following

passage outlines his understanding of these terms:



Donnellan’s Distinction 79

I will call the two uses of definite descriptions I have in mind the attributive use and

the referential use. A speaker who uses a definite description attributively in an as-

sertion states something about whoever or whatever is the so-and-so. A speaker who

uses a definite description referentially in an assertion, on the other hand, uses the

description to enable his audience to pick out whom or what he is talking about and

states something about that person or thing. In the first case the definite description

might be said to occur essentially, for the speaker wishes to assert something about

whatever or whoever fits that description; but in the referential use the definite de-

scription is merely one tool for doing a certain job—calling attention to a person or

thing—and in general any other device for doing the same job, another description

or a name, would do as well. In the attributive use, the attribute of being the so-and-

so is all-important, while it is not in the referential use.1

The attributive use is shown in sentences where the predicate “F” in the

description is used to apply to whatever satisfies it, not to a particular

thing. The fact that a thing in the world actually satisfies the predicate

is essential and all-important. With Donnellan’s notion of attributive use,

we could paraphrase the sentence “The king of France is bald” as “Who-

ever is uniquely king of France is bald”—perhaps asserted in the belief that

being king of France induces baldness in whoever occupies that position.

To determine if this sentence is true, we would have to find whoever in theworld satisfies the description “the king of France” and then determine

whether that person is bald. This is clearly along the lines of Russell’s analy-

sis of the semantics of descriptions.

The referential use occurs when the description picks out a particular

object in order to identify something for an audience, where the descrip-

tion is just a tool for directing the audience’s attention in the right way. In

the simplest case, the object of interest is right in front of the speaker and

in plain sight of the audience. The description is used to show the audi-

ence the particular object the speaker has in mind. The description here is

not essential and all-important, because many other modes of identifica-

tion would work as well. Imagine a classroom full of students in which

one of the male students is wearing a green shirt. A female student could

make a statement about him in the following ways: “The guy in the green

shirt has a pensive look,” “He [pointing] has a pensive look,” “Billy has

a pensive look.” The speaker then chooses one but could have used the

others, depending on what she thinks will direct the audience’s attention

to the right person most effectively. Her purpose was to identify a certain

individual and make a remark about him—she couldn’t care less about the



80 Chapter 4

description itself. She wanted to point out the guy’s pensive look and any

mode of designation would do.

Donnellan’s point is that these are very different speech situations inwhich the speaker has very different communicative intentions. According

to him, the description functions differently depending on the intention

behind the speech act. He uses a thought experiment to illustrate this point

more clearly. Imagine that a detective on a crime scene has encountered the

body of a dead man, Smith. The condition of the body is so mutilated that

the detective says, “Smith’s murderer is insane!” When he said that, he did

not know the identity of the murderer. That statement could be rephrased

as “Whoever Smith’s murderer is, he is assuredly insane.” This is an excel-lent example of the attributive use. For that statement to be established

as true, the detective would have to find the person who murdered Smith

and determine whether or not he is insane. He certainly had no specific

individual in mind; hence the use of the quantifier phrase “whoever is the

murderer of Smith.”

The same description could also have a referential use. Suppose Jones is

being tried for the murder of Smith and one of the jurors notices that Jones

is behaving erratically the whole time. The juror then points at Jones and

says, “Smith’s murderer is insane.” The juror has thus succeeded in identify-

ing Jones. He wanted to single that man out and make a remark about him;

here the quantifier phrase would be quite inappropriate.

Now consider a case where Jones is not in fact Smith’s murderer though he

is on trial and behaving erratically. Donnellan thinks the juror has still iden-

tified that individual even though he is not Smith’s murderer, because the

audience understands that he intends to refer to Jones and to say he is insane.

It could be the case that Jones is insane but not Smith’s murderer. In that case,

the juror has still said something true about Jones because he is insane and the

speaker has singled him out. Regardless of the situation and the truth or fal-

sity of the juror’s description, the juror has succeeded in identifying the indi-

vidual in question by using that definite description. The description itself

is not all-important in the reference that the juror has achieved and it is not

essential that the referent actually satisfy it. Although the description may be

defective if it does not apply to Jones (depending on the situation), the jurorhas still managed to identify a particular individual using that description.

It is as though the description can function either as a quantifier phrase or

as a demonstrative that points out someone. The juror has succeeded in his




referential intention by identifying an individual and making a statement

about him. The detective, on the other hand, is best interpreted as saying

something analyzable along the lines of Russell’s theory.There is another thought experiment Donnellan uses that illustrates the

same point. Imagine that you are at a party and there is a man apparently

drinking a martini who is a famous philosopher. Seeing this man, you say,

“The man drinking a martini is a famous philosopher.” However, suppose

that although the man is a famous philosopher, he is drinking water from

a martini glass, not a martini. You have said something true about him, but

your identifying description does not apply to him. Nevertheless, it can still

perform its function of indicating to whom you meant to refer.Now we can consider a similar case that illustrates the attributive use.

Imagine that the woman running the party does not want people drink-

ing alcohol and says, “Who’s the man drinking the martini?” She is not

intending to identify a particular individual as you were in the previous

example—indeed she is trying to discover who the martini drinker is. If

it turns out that the man apparently drinking a martini is not drinking a

martini, she will not be concerned. Her speech act requires that there be

somebody who satisfies that description. If there is somebody at the party

who fits that description, she would have accomplished her aim by using

that description. She is using the description to mean “whoever is drinking

a martini”—she has no particular individual in mind.

It is also possible that there is in fact another man at the party who is

drinking a martini, is in another room, but is not a famous philosopher.

The statement “The man drinking a martini is a famous philosopher”

would then be false if the description is interpreted attributively. Although

the man drinking the martini was not your intended reference, he hap-

pened to fit your description. Your reference was to the person you incor-

rectly described as a martini drinker, though you also said something true

about him.

The best way to understand both of these examples is to determine the

intention of the speaker. Ask yourself—does the speaker intend to iden-

tify a particular individual or just to speak of whoever satisfies a particular

description? Sometimes the use of the definite description will have a gen-eral (attributive) intention behind it, and sometimes it will have a singular

(referential) intention behind it. It all depends on what the speaker intends

to communicate.



82 Chapter 4

Donnellan continues the article by reiterating the main line of argument.

Each of his subsequent examples illustrates the difference in intention in

the attributive and referential uses. That is Donnellan’s fundamental wayto understand any of these cases. If it does not matter that the description

fits the object, it is a referential use. If it does matter, then it is an attribu-

tive use. Therefore, we can actually refer to something using a description

without truly describing what we are referring to. Referential success does

not depend on accurate description.

To sum up: The core of Donnellan’s argument is the distinction between

referential and attributive uses. He illustrates the distinction by means

of the thought experiments we have just described. A speaker uses thedescription attributively when she says “Smith’s murderer” with a general

intention. The speaker has no particular person in mind when using this

description. She could have equally said, “The murderer, whoever he may

be, must be insane.” The referential use occurs when the speaker has a par-

ticular individual in mind and uses her description to pick out the indi-

vidual she has in mind. Donnellan’s main argument deals with these two

uses of descriptions—the generality of the attributive use and the particu-

larity of the referential use. A consequence of the distinction, according to

Donnellan, is that in the referential use the speech act can be successful

regardless of the truth or falsity of the description. Referring back to the

Smith’s murderer case, Jones might not be the murderer but a juror can still

identify Jones by saying, “Smith’s murderer is insane.” Unlike the attribu-

tive use, the descriptive content is not all-important in the referential use.

The description in the referential use is incidental, a mere instrument to

identify an individual. Donnellan thinks that the theories of Russell and

Frege/Strawson are incorrect because they do not acknowledge the duality

of uses of descriptions.

In the rest of his paper, he brings up various consequences of this basic

point. By understanding the distinction between these two uses, we can

understand his main line of argument. One obvious point is that the refer-

ential use occurs when a particular thing is pointed out, and in the attribu-

tive use a remark is made involving some general notion. It is the difference

between a quantified proposition (as in “whoever”) and a particular proposi-tion (as in “this individual”). The distinction is analogous to the distinction

Russell discusses when he talks about the difference between a name and a

description. Using our understanding of Russell is another way to explain




Donnellan’s distinction. He thinks that some uses of definite descriptions

are namelike in the Russellian sense, but others are propositional-function-

like. Yet the expressions themselves look uniform from use to use.One of the consequences of this distinction is that even though in both

uses the speaker presupposes the individual she is referring to (or trying to

refer to) satisfies the description, there are different outcomes of the indi-

vidual not satisfying the description. If the description is attributive and

no one satisfies it, then the statement cannot be true. According to Russell,

the statement would be simply false. For example, according to the theory

of descriptions, “The king of France is bald” is false because there exists no

such thing as a king of France. If we use this description attributively, andthe implication of there being something that fits the description turns

out to be false, then the statement cannot be true and must be false. On

the other hand, according to Donnellan, in the case of referential use the

statement can still say something true regardless of whether the intended

referent satisfies the description. It might be that Jones really is insane even

though Jones is not in fact Smith’s murderer.

There may also be instances where the speaker does not even believe

the description she uses to refer to the individual is true of that individual.

In most instances, the speaker will think that the description applies (e.g.,

that Jones in the dock is a murderer or that the man over there is drink-

ing a martini). However, Donnellan suggests that there can be instances

where the speaker knows the description is not true but uses it to identify

the individual anyway. Consider the example he gives of a spurious king.

The speaker may believe that this alleged king is a usurper and so not really

king, but because everyone else in the country thinks that the man is the

rightful king, the speaker still refers to him as such (e.g., “Is the king in the

counting house?”). The speaker does not believe that the individual she

wishes to speak of is the king but uses the royal description anyway. She

makes a successful referential use out of a piece of false description.

The hearer of the sentence may or may not believe the description as

well. For example, instead of all of the people around the alleged king

thinking that he is the king, they could all think he is a usurper. They may

still refer to him as “the king” to avoid any trouble. Everyone in the courtwill refer to the usurper with the description “the king” and know he is not

the king but still use that description anyway. In this case, if our original

speaker asks, “Is the king in the counting house?” everyone in the court will



84 Chapter 4

understand to whom the speaker refers, even though they do not believe

that impostor to be the king. The description can still refer even if it is

false— even if the speaker and hearer know it to be false.

4.3 Denoting and Referring

Having said all this, Donnellan makes a further distinction between denot-

ing and referring . He does not deny that there may be a sense in which

the description “Smith’s murderer” denotes somebody other than Jones,

assuming Jones is innocent. The juror is referring to Jones with that false

description, but Donnellan accepts that the description may have a deno-tation that is other than Jones. If we suppose that Brown is the man who

actually killed Smith, then “Smith’s murderer” denotes Brown. In that case,

the juror refers to Jones by saying “Smith’s murderer,” but his description

denotes Brown. Donnellan borrows this idea of denotation from Russell.

According to Donnellan, the speaker can refer to somebody with a descrip-

tion who is not the person denoted by that description. Thus referring is to

be distinguished from denoting.

Denotation is a semantic notion—a strict and literal interpretation of the

phrase “Smith’s murderer.” It is not the pragmatic notion of what or whom

the speaker is referring to in using that phrase. This marks the distinction

between a pragmatic question and a semantic question. In effect, Donnel-

lan is admitting that he is primarily interested in the pragmatic question

of how individual speakers convey a message to hearers on particular occa-

sions. He accepts that the description, considered in itself, denotes (semanti-

cally) whatever fits the description, and so functions “attributively.” So a

speaker can use a description that semantically denotes a particular indi-

vidual (Brown) to pragmatically refer to another individual (Jones). Donnel-

lan is therefore not arguing that there are two interpretations of semantic

denoting. He thinks that denoting follows Russell’s theory, but there are

pragmatic uses in which a speaker refers to something other than the

denotation.

As a matter of fact, at one point in “Reference and Definite Descriptions”

Donnellan clearly states that he is not arguing against Russell’s semantictheory:

It does not seem possible to say categorically of a definite description in a particular

sentence that it is a referring expression (of course, one could say this if he meant




that it might be used to refer). In general, whether or not a definite description is used

referentially or attributively is a function of the speaker’s intentions in a particular

case. “The murderer of Smith” may be used either way in the sentence “The murder-

er of Smith is insane.” It does not appear plausible to account for this, either, as an

ambiguity in the sentence. The grammatical structure of the sentence seems to me to

be the same whether the description is used referentially or attributively: that is, it is

not syntactically ambiguous. Nor does it seem at all attractive to suppose an ambigu-

ity in the meaning of the words; it does not appear to be semantically ambiguous.

(Perhaps we could say that the sentence is pragmatically ambiguous: the distinction

between roles that the description plays is a function of the speaker’s intentions.)2

This is a very important passage in terms of the significance of Donnellan’s

arguments. He claims here that there is no semantic ambiguity in descrip-tions. By semantic ambiguity he means what the words actually mean in

the language—their logical analysis. There is no semantic ambiguity in

descriptions even though speakers may use those descriptions in two differ-

ent ways. He thus in effect admits that descriptions are always semantically

attributive, that is, Russellian. One of the major criticisms of Donnellan,

to be considered later, is that his critique of Russell’s theory is ineffective

because he tries to apply a pragmatic distinction to a semantic question.

Therefore, understanding the import of this passage is particularly impor-

tant to this discussion.

4.4 Truth-Value Gaps

Donnellan makes some of his main points against Strawson toward the end

of his article. He argues that Strawson is wrong to suggest that when using

an empty description referentially the speaker says something neither true

nor false. According to Donnellan, the speaker can say something true by

using a description that fails to refer. If there was no murderer of Smith at

all, but just a gruesome accident, and the speaker shouts, “Smith’s murderer

is insane!” referring to Jones, Strawson thinks that the utterance would be

neither true nor false. But Donnellan argues that the speaker would have

said something true of Jones, assuming that he is in fact insane.

He goes on to say that in certain instances he agrees with Strawson.

There might be cases where you do fail altogether to refer to an object usinga description. Imagine first a case where an onlooker sees a man apparently

carrying a stick and says, “The man carrying a stick is out of breath.” Now

suppose a man is there, but he is carrying a rifle instead of a stick. Donnellan



86 Chapter 4

thinks the onlooker is still referring to the man there even though the man

carrying a rifle does not fit the speaker’s description. However, there could

be a case where the onlooker has completely hallucinated a walking man.The onlooker could also have mistaken a tree or a rock for a man with

a stick, in which case Donnellan believes the onlooker would have still

successfully referred to something. But this referential ability comes to an

end at a certain point. If the onlooker has completely hallucinated a man

with a stick and there is only empty space there, not even a tree or a rock,

then Donnellan thinks he has failed to refer to anything whatever—man,

rock, tree, chunk of space. He is totally out of luck, referentially speaking.

Strawson would have been right to say that in such a case the utterance isneither true nor false. Here the speaker’s intention to refer would have been

completely nullified. The question of truth-value would not arise in that

sort of case.

Therefore, Donnellan does think that there are cases of intended refer-

ence to something where it turns out that no reference takes place. The

consequence of such radical reference failure is that the speaker will be stat-

ing something neither true nor false. Of course, in Russell’s theory, such a

statement would express a straightforwardly false proposition. Donnellan

takes a middle ground. He does not think that it is always the case that the

speaker says something true or false, but he also thinks that Strawson over-

states the frequency of such truth-value gaps. He thinks both Russell and

Strawson are wrong about certain cases of reference failure, though they are

right about others.

As in his concluding points on Strawson, Donnellan sees some common-

ality between his views and Russell’s. Although Donnellan believes that

Russell’s theory is incomplete because it does not recognize the referential

use of descriptions, he still thinks that his conception of descriptions is

analogous to Russell’s conception of names. Russell regards genuine names

as labels for particular objects and not as descriptions of objects. He thus

makes a sharp distinction between names and descriptions. In Russell’s

system, a genuine name acts merely as a tag for an object and does not

describe the object at all. Donnellan suggests that he can map his distinc-

tion onto Russell’s distinction, because he thinks that descriptive contentdoes not play a role in the referential use of descriptions. Donnellan views

descriptions used referentially as mere labels for object—they are namelike.

Whether or not the description correctly describes the object is irrelevant,




because the object has been successfully labeled. In his system, these

descriptions only appear to be descriptions, because they do not refer by

describing. They only label or point. Descriptions can thus act like namesin the Russellian sense, and so it is not important if the object satisfies the

description, since they succeed in referring even if inaccurate. For Donnel-

lan, the descriptive content of the description is incidental and dispensable

to the role it plays in referring, in the case of referential uses.

There is another class of examples that Donnellan does not cover in his

paper, but which illustrates this point well. In these examples, the descrip-

tions function as names, and it is entirely obvious that they do not accu-

rately describe the things to which they refer. Consider “the Holy RomanEmpire”—a description that notoriously refers to something that was nei-

ther holy, nor Roman, nor an empire. The description in that example is

not referring by means of its descriptive content. These words stand for

something that is entirely removed from their actual predicative meaning.

Compare “the European Community” or even “the United States” or “the

Grand Order of Pig Farmers” (this one is invented!). The combinations of

words in these descriptions have become labels that refer but the descrip-

tive meaning is beside the point. These are like Donnellan’s referential uses.

4.5 Evaluating Donnellan’s Distinction

When assessing the cogency of Donnellan’s arguments, it is important to

consider certain cases that may arise when using other types of expres-

sion in sentences. Consider a situation similar to his thought experiment

about the famous philosopher at the party apparently drinking a martini.

This time imagine that the famous philosopher at the party is a certain

individual, say, Jerry Fodor. Let’s suppose that the hostess of the party has

heard of the philosopher Saul Kripke and heard him described; moreover,

she has reason to believe he is at the party. Suppose now that she sees Fodor

chatting to a group of people about philosophy. She forms the belief that

this must be Kripke and says, “Kripke is very animated.” Of course, she is

wrong about who is in front of her, but the question arises—whom did she

refer to with that name? We might be tempted to say that she succeeded inreferring to Fodor with “Kripke” and made a true remark about him, even

though her referent does not “fit” the name she used. Kripke himself might

be in another room passed out, so not at all animated—did she refer to him



88 Chapter 4

and make a false statement about him? Following Donnellan, we might say

that such an example illustrates a referential use of names, in which accu-

racy counts for little: hasn’t she in some sense referred to the man in frontof her, that is, Fodor? Semantically, the name denotes Kripke, but prag-

matically our hostess seems to have referred to Fodor. She has referred to

somebody who is not Kripke with the name “Kripke,” which has a specific

meaning that makes it denote only Kripke. In other words, our hostess has

used the name in a way that does not fit its actual conventional meaning.

It thus seems that Donnellan could have written a paper called “Refer-

ence and Names” and said all of the same things about names that he says

about descriptions. There are two uses of names, referential and attributive,and referring must be distinguished from denoting, and so on. But some-

thing seems to be going wrong with this argument if the ways in which

speakers misuse words show that particular semantic theories of names are

false. If we wonder whether Donnellan’s objections apply to theories of

proper names, then we must also consider if they apply to demonstratives.

Suppose a tourist in front of an animal at a zoo says, “That antelope is

brown.” However, the animal is not an antelope but a different species of

deer. Although the speaker has in some sense succeeded in referring, the

animal he has pointed out does not fit the demonstrative he used. The

speaker’s misuse of the demonstrative is exactly like the hostess’s misuse of

the name “Kripke.” The intended reference of the tourist is the animal in

front of him, but it is not an antelope as he supposed. It is possible, then,

to use a demonstrative to refer to something other than the denotation of

that demonstrative, if it has one. According to Russell and Strawson, such

a demonstrative would be empty since it lacks denotation. But the tourist

could still succeed in saying something true about the animal in front of

him, even if the animal were not an antelope.

Since it holds of names and demonstratives, it seems as though we could

apply the Donnellan treatment to any expression. There are myriad exam-

ples in popular culture of linguistic misuse, particularly when speakers use

certain terms to try to sound sophisticated but instead demonstrate their

ignorance. Certain speakers treat the words “disinterested” and “uninter-

ested” as though they are interchangeable. But “uninterested” means to lackinterest in something whereas “disinterested” means to be impartial about

something. A disinterested observer of a tennis match need not be an unin-

terested observer. On the contrary, a disinterested observer could be a very




interested observer—she is just impartial. Someone may say, “I’m just com-

pletely disinterested in that subject matter,” and the hearer, knowing this

common error, can infer from the speaker’s misuse of the word the point sheis trying to make, namely, that she lacks interest in that subject matter. True

things can thus be conveyed by the misuse of words. If we were very inge-

nious about it we could create Donnellan cases for quantifier words, or even

words like “and” or “not”—almost anything. All you need to do is produce a

case where a speaker utters a word that has a certain conventional meaning

(denotation) and uses that word in an incorrect way. Even though the word

will not apply to the thing the speaker is applying it to, the audience under-

stands what the speaker meant to convey and so the speech act is successful.Any expression of language at all could be used in that deviant way. If you

know that I have a tendency to confuse my quantifier words (perhaps I am

new to the language I am speaking), you can on occasion interpret my use

of “someone” to mean “no one”—so when I say “someone is in that room”

you interpret me to be intending to convey my impression that no one is in

that room (especially if the room is palpably empty).

The significance of that point concerns whether or not the production

of Donnellan cases can undermine semantic theories of certain classes of

expressions. If there is an established, semantic definition of a word, cap-

tured by a particular theory, can that theory be undermined by pointing

out that people sometimes misuse the word? No. The misuse of a word does

not change its semantic status, or show that a particular theory of its mean-

ing is incorrect. People can misuse words in the way Donnellan describes,

but that does not mean that these misuses establish an interesting linguistic

duality. If a foreign speaker of English does not understand the language

and he uses the word “and” when he means “all,” his misuse of the word

“and” would not change the meaning of “and,” or show that the theory

of “and” as a truth-functional sentence connective is mistaken or oversim-

plified. Would we say that the meaning of “and” is ambiguous because a

foreigner used the word incorrectly? No. Nor would we say that “and” has

two uses, as a sentence connective and as a universal quantifier. As Donnel-

lan admits in the passage cited earlier, he is not pointing out any semantic

ambiguity. But then Donnellan’s considerations could not even be relevant to a question of semantics, because they are purely a matter of pragmatics.

The pragmatic point he is making is that it is possible for speakers to use

words to convey something that is completely divorced from what those



90 Chapter 4

words actually mean. Thus a speaker can express a belief about Jones by

using words that denote Brown (“Smith’s murderer”). Donnellan’s point is

a purely pragmatic one and does not undermine any semantic theory. SinceRussell’s and Strawson’s theories were offered as semantic theories, Donnel-

lan’s point is irrelevant to those theories. For all Donnellan has said, Russell

is completely correct about the semantics of descriptions. Descriptions will

always semantically denote what fits them. Speakers can use such descrip-

tions incorrectly to make singular reference, but that does not show Russell

is wrong in the semantic theory he provides.

4.6 Implication and Implicature

To further evaluate Donnellan’s position, we will now bring in some points

made in an excerpt from Stephen Neale’s book Descriptions.3 In this excerpt,

Neale makes use of some ideas developed by Paul Grice. Since these ideas

are independently important, we will spend a little time explaining them.

Perhaps the most well-known idea covered in his article is that of “conver-

sational implicature.” To explain the notion of conversational implicature,

we can use the example of a professor who is asked to write a reference let-

ter for one of his graduate students:

To whom it may concern,

John Smith has very good handwriting.

Sincerely,

Professor Horatio Handwavy, PhD

The committee reviewing John Smith’s application would not infer thathe has outstanding philosophical ability from this recommendation letter.

They would infer that Professor Handwavy thinks poorly of Smith. Suppose

that after reviewing Smith’s entire application and interviewing him the

committee decides that Smith is an excellent candidate. Then one of the

committee members asks the letter writer why he said that John was a poor

student. Handwavy indignantly replies, “I did not say that he was a poor

student, I just said that he has good handwriting. In fact I think Smith is

a brilliant student.” And it is true that he never said anything false aboutSmith’s philosophical ability. In fact, he said something quite true, because

John is also an excellent calligrapher. But he certainly implied something

false, irresponsibly so. He didn’t straightforwardly lie, but he certainly gave




an extremely false impression. He is morally at fault, even if not factually

at fault.

This example illustrates conversational implicature, which relates towhat a statement suggests given its context. Nothing said in the letter

as written logically implies that John Smith is a bad philosophy student.

However, given the context of the recommendation letter, the professor

did conversationally imply it. We could reasonably paraphrase the original

sentence into its conversational implicature: in that context, saying “John

Smith has good handwriting” is tantamount to saying “John Smith is a bad

philosophy student.” The notion of conversational implicature reveals the

distinction between what a speaker strictly says when uttering a sentenceand what he or she implies in uttering it. What the speaker means, and can

be reasonably taken to mean, can depart quite dramatically from the literal

meaning of the sentence that is uttered. When a speaker utters a sentence,

therefore, there is the proposition conversationally meant and the proposi-

tion literally expressed. These may coincide, but they may not.

Neale outlines this distinction in his book. The proposition expressed

is closely connected to the meaning of that sentence in a particular lan-

guage, whereas the proposition meant depends on the context and expecta-

tions about the speech act. The proposition expressed and the proposition

meant can be completely different propositions that are not logically

related to each other. Therefore, in conversational implicature, proposi-

tions are implied conversationally that are not directly expressed by words.

The point is very important philosophically because it undermines various

philosophical claims made about various subjects. It is really important to

distinguish between whether an utterance of a sentence is strictly speaking

false and whether it is misleading to say it in a certain context. The fact

that something is misleading to say in a certain context does not show that

it is false. It is misleading to say, “It looks to me as if there is a dog in the

doorway” if you are not in any doubt about there being a dog there, but it

might still be true that this is just how things look to you.

Neale’s main problem with Donnellan is that he disregards this distinc-

tion. Donnellan is suggesting that Russell’s analysis of definite descriptions

is inadequate because it does not handle his cases of referential use. Nealerejects that form of argument, because he does not see Donnellan’s prag-

matic points as having any implications for semantics. Although Neale

never mentions it, we have discussed a passage from Donnellan’s original



92 Chapter 4

article that admits this very distinction. In that passage, Donnellan clearly

states that there is no syntactic or semantic ambiguity in sentences contain-

ing definite descriptions. Nevertheless, he still thinks there is somethingwrong with Russell’s account of the meaning of definite descriptions. The

question is how he can make the admission and persist with the argument.

He thinks that his two pragmatic uses somehow show that something is

wrong with Russell’s semantic analysis, but he himself accepts that his con-

siderations do not bear on semantics.

Let us suppose that Russell’s account is correct for attributive uses, so

that descriptions are quantifier expressions when used attributively. Accord-

ing to Donnellan, there is no semantic ambiguity in definite descriptions.Therefore, when definite descriptions are used referentially, they have

exactly the same meaning as when they are used attributively. If that is the

case, then we must suppose that Russell’s theory gives the correct meaning

in both cases. We have seen how the misuse of words cannot undermine

an analysis of their semantics. So Donnellan has not pointed to anything

that could threaten Russell’s semantic theory. If Russell is correct about the

attributive use, then he must be correct about the referential use as well.

The curious thing is that Donnellan already admits the point that Neale

is urging against him—that there is no semantic ambiguity. Yet he doesn’t

seem to appreciate the significance of this admission.

Neale believes that Donnellan’s arguments show the necessity of bring-

ing to bear Grice’s distinction between the proposition expressed and the

proposition meant. To understand why this distinction matters, we shall

return to Donnellan’s examples. Let us consider again the “Smith’s mur-

derer” case where Jones is the man in the dock. The juror sees Jones’s erratic

behavior and wants to express his belief that Jones is insane, so he says,

“Smith’s murderer is insane!” The proposition meant is that Jones, the man

in the dock, is insane, even though Jones did not in objective fact murder

Smith. The proposition meant is in line with Donnellan’s referential use.

However, the proposition expressed by the sentence itself (“Smith’s murder

is insane!”) is that Smith’s murderer is insane, which may or may not be

true. In the case that Jones is insane, the proposition meant (that Jones is

insane) would be true, but the proposition expressed would be false, assum-ing the real murderer (Brown) is not insane. The analysis of Donnellan’s

examples with this Gricean distinction allows us to see that there are two

different propositions “associated with” the utterance of the sentence in




this instance. These propositions are about different individuals and can

differ in truth-value.

The handwriting case also illustrates the distinction between propositionexpressed and proposition meant. In that case, the proposition expressed

is that John Smith has good handwriting and the proposition meant (or

apparently meant) is that John Smith is not a good philosopher. One prop-

osition is entirely different from the other. Although the words can be used

by a speaker to convey a particular proposition, the actual words spoken

may not mean that proposition. What Donnellan really showed was that

sentences can be used by speakers to mean propositions that those sen-

tences do not express—and hence to convey information that the words donot themselves contain.

Thinking about this idea in a more general way, we can see many uses

of language have much the same character. Take irony, for example. If a

speaker says something ironically, the proposition expressed is the opposite

of the proposition meant—for example, “You are so smart,” said sarcasti-

cally. However, it would be strange to claim that the possibility of irony

somehow changes the semantic analysis of the sentence. Irony depends

on the fact that the proposition expressed is not the same thing as the

proposition meant. Irony, then, is another example of this type of distinc-

tion working itself out, where the relationship between literal meaning and

speaker meaning is complex. In this case, one proposition is actually the

negation of the other.

Hyperbole and exaggeration also illustrate these distinctions. Hyperbole

uses exaggeration to convey a particular point. It would be misguided for

someone to interpret a hyperbolic statement as literal. If you were to describe

someone as extremely tall by saying, “That guy is like twenty feet tall,” most

listeners would not think that the man is actually twenty feet tall. There is

a difference between what a sentence means and what the speaker means

by using that sentence in a particular way. Metaphors also demonstrate this

point. If Romeo says, “Juliet is the sun,” it would be strange for someone to

claim he has discovered a hidden semantic ambiguity in the word “sun.” We

must not conflate the message conveyed in using language with what the

words themselves literally mean. It is indeed of the essence of language thatwe can sometimes use words to mean what they do not in fact mean.

This concludes our discussion of Donnellan, but not of Russell’s the-

ory. Even though Donnellan’s critique of Russell seems misguided for the



94 Chapter 4

reasons given, other objections to Russell’s theory persist. Let us quickly

survey these objections.

4.7 Further Objections to Russell’s Theory

The first objection is Strawson’s: empty descriptions give rise to statements

that are neither true nor false. According to Russell’s theory, “The F is G”

expresses an existential proposition—that there exists an F. If there is no F,

the sentence expresses a false proposition. Strawson’s point is that Russell’s

assignment of truth-value is intuitively wrong—it sounds more natural to

say that the sentence fails to express a proposition that has a truth-value.We do not want to say that “The king of France is bald” is false when there

is no such king. It could be false only if there were a king of France and he

had a decent head of hair. Thus Strawson contends that when the descrip-

tion is empty, the statement is neither true nor false.

Another type of example makes the criticism even clearer: “The golden

mountain is golden.” This statement looks to be true a priori, but according

to Russell’s theory it would be simply false, because of the lack of golden

mountains. Such a statement does not seem to fit Russell’s theory at all.

Russell might reply that this is just a matter of ordinary language—he has

shown that contrary to appearances the statement is false. There is some-

thing to be said for Russell’s response. It is always possible to insist that

sentences like “The golden mountain is golden” are strictly speaking false.

We do not ordinarily say they are false, but they are false. Skepticism tries

to show that nobody knows anything. According to skepticism, it would

be false to say, “I know I am reading these words.” It seems fairly strange to

say that sentence is false, but it is possible to argue that it is strictly speak-

ing false. In the same way, with statements like “The golden mountain is

golden,” we might insist that the statement really is false though to com-

mon sense it seems true. But still, Russell’s position does strike one as hard

to accept and makes one wonder if his analysis is correct.

The second objection is that “the golden mountain” and “the king of

France” are phrases, not sentences. They are parts of sentences, not actual

sentences. Grammatically, these phrases constitute the same part of speechas a name or demonstrative. If a speaker only says, “that dog,” or “Saul

Kripke,” he has only uttered a fragment of a sentence and hence hasn’t

said anything. According to Russell, however, descriptions are complete




sentences because they expand into assertions of existence and unique-

ness. If a speaker just says “the man outside,” we would think he has yet

to express a complete proposition; but on Russell’s theory, he has alreadysaid that there exists a man outside and only one man outside. This seems

strange since the speaker never actually completed a sentence. Note, in

addition, that if we apply the description theory to names and then ana-

lyze the description Russell’s way, then simply uttering a name is already

expressing a complete proposition—to the effect that an F uniquely exists.

But did I say anything with truth-value just by saying “Eric Clapton”?

Both these objections suggest that definite descriptions are more name-

like than Russell allows. They are used as subject terms to identify some-thing of which an attribute is predicated. Whether the statement is true or

false depends on whether the thing identified with the descriptive term

has the attributed property. The description looks more namelike than sen-

tencelike. The description looks like part of a sentence—the subject part—

not a whole sentence. Again, this makes us wonder if Russell has the correct

analysis.

Nonindicative sentences raise more worries about Russell’s theory. Con-

sider an example of an imperative sentence: “Kill the King of France!” Using

Russell’s theory, we would have to paraphrase this sentence as “Kill there is

a unique king of France!” The first thing to be said about this paraphrase is

that it is meaningless, ill formed, and ungrammatical. If the definite descrip-

tion is simply replaced with Russell’s paraphrase, the sentence comes out as

nonsense. Russell’s theory cannot be applied mechanically in this case. Rus-

sell never discusses how he would handle cases where descriptions occur in

imperatives. It doesn’t help to render the imperative as “Make it the case

that the king of France is dead!” because then the imperative will order the

addressee to make it the case that a unique king of France exists—which is

rather contrary to the command to kill him.

A closely related problem to the problem posed by imperatives is illus-

trated by the sentence “George IV wondered whether the author of Waver-

ley smoked cigars.” Replacing the description with Russell’s paraphrase, we

have it that George IV wondered whether there exists an author of Waver-

ley , and only one author of Waverley , and whether he smoked cigars. ButGeorge IV may have never wondered if there existed an author of Waver-

ley and only one such. He only wondered whether or not the author of

Waverley smoked cigars—he took it for granted that said author exists. If



96 Chapter 4

the definite description occurs in a propositional attitude context (in this

case “wonder whether”), then we obtain the wrong analysis when applying

Russell’s theory. So not all occurrences of descriptions fit Russell’s theory.A third objection stems from the fact that descriptions can function and

yet be radically incomplete. Take the description “the table” and consider

the sentence “The table is bare.” Now if we analyze this sentence according

to Russell’s theory, there is a problem in the second conjunct, “There exists

only one table.” The original statement surely did not imply that there

is only one table in the world! If it did, it would be false. When incom-

plete descriptions are analyzed by Russell’s theory, the uniqueness clause is

shown to be dramatically false.Certain maneuvers could help Russell escape these problems. Some peo-

ple have suggested that a phrase such as “the table” is actually a demon-

strative. Thus “The table is bare” means “That table is bare.” If we use this

paraphrase, the problem of uniqueness falls away because the context sin-

gles out the object of reference. Such descriptions turn out to be demonstra-

tives and so are not analyzed by Russell’s theory. But then we have admitted

that not all descriptive phrases fall under Russell’s analysis. Demonstratives

are singular referential devices that pick out an object; they are not quanti-

fier phrases. Since some grammatical definite descriptions are not quanti-

fier-like, Russell is wrong to have claimed that all definite descriptions are

quantifier-like.

Then we have degenerate namelike descriptions like “the Fonz,” “the

Ace,” and “the Situation.” Presumably Russell would deny that these are

descriptions at all, but they look like descriptions—and they are clearly

namelike. What about “the GOP”?

A final problem for Russell concerns “the former” and “the latter.” How

do we analyze these as quantifier phrases? “Jack and Jill went up the hill

and the former fell over while the latter sat down”: it is quite impossible to

paraphrase these “the” phrases using Russell’s theory. Try it and see.

Russell’s theory seems to contain a strong element of truth, but difficul-

ties emerge if we try to apply it across the board. How to deal with these

difficulties is an unsolved problem in the philosophy of language.





98 Chapter 5

Kaplan employs a slightly more theoretical way of explaining intension

and extension. He characterizes the intension of a sentence as a function

from possible worlds to truth-values. Thus intensions act like mathematicalfunctions, taking worlds as arguments and giving truth-values as values. For

example, in an addition equation like 2 + 3 = 5, 2 and 3 are the arguments for

the addition function and the value of the function for those arguments is

5. In the same way, the value of the function that is the intension of “Nadal

was the number one tennis player in the world in 2010” is True for the actual

world as argument, but for other worlds as argument the value of the func-

tion is False. Sentence meanings are thought of as functions from worlds to

truth-values. Intensions determine extensions with respect to worlds.In specifying the function a given sentence expresses from worlds to

truth-values, we determine the truth conditions of the sentence. The truth

conditions of a sentence are the set of worlds in which it is true: so our

sample sentence is true in just those worlds in which Nadal is number one.

A possible world semantics theorist explicates meanings as functions from

worlds to truth-values, which is to say in terms of truth conditions. This

idea can be extended to parts of sentences, such as definite descriptions.

Take the following definite description: “the inventor of bifocals.” This

description, like a whole sentence, has a particular intension and an exten-

sion, which is the reference of the description. In the actual world, Benja-

min Franklin is the reference (extension) of that description. However, in a

different possible world he might not be the extension, because he might

not have invented bifocals and someone else did. The intension of the

description determines a different object as extension in different worlds,

just as the intension of the sentence determines different truth-values in

different worlds. The meaning of the definite description is a function from

worlds to extensions in the same way that the meaning of a sentence is a

function from worlds to extensions. The difference lies in the fact that for

a sentence, the extension is its truth-value, whereas for a description, the

extension is an object. In the case of this particular definite description, the

extension corresponding to the intension with respect to the actual world

is Benjamin Franklin, but that same intension with respect to a different

world could give Thomas Jefferson as the extension. The extension variesfrom world to world, while the intension stays fixed. This is a way of talking

about contingency: it is only contingent that the inventor of bifocals was

Benjamin Franklin.



Kaplan on Demonstratives 99

But there is also necessity to consider. The sentence “2 + 2 = 4” expresses

an intension that has the same extension with respect to every world,

because the proposition is necessarily true. There are no worlds where 2 + 2equals anything but 4. The function gives the same value as output regard-

less of what world goes into it as input. Any world you go to, you will see

that 2 +2 = 4 in that world. The intension here is a constant function from

worlds to truth-values, because it never varies in the output of the function

from world to world. On the other hand, if we had written “2 + 2 = 5,” that

would have the truth-value of being false in every world, because there is

no world where 2 + 2 = 5.

There are also cases in which definite descriptions are necessarily true oftheir bearers. We talked about one of these when we were discussing Kripke

in chapter 2. For example, “the successor of 3” refers to only one number

from world to world because in every possible world the successor of 3 has

to be 4. To put it in Kripke’s vocabulary, that description is a rigid designa-

tor, because it has the same designation in every world. Using that termi-

nology, we could say “Nadal is number one” is a non-rigid designator of

the truth-value True, and “2 + 2 = 4” is a rigid designator of the truth-value

True. Thus there are definite descriptions that are rigid designators, and

they function in essentially the same way that the non-rigid ones do, that

is, they are associated with intensions that operate as functions from worlds

to extensions. The difference is that the rigid designators express constant

functions, whereas the non-rigid designators express variable functions.

Suppose we make a representation of the proposition expressed by a sen-

tence containing a definite description. The thought expressed by the sen-

tence, the proposition, will consist of the intensions of the various terms

of the sentence. The intension for the description will be something like

the concept of being F. So the component of the proposition correspond-

ing to “the F ” will be the concept of being uniquely F , and then there

will be other components for other expressions in the sentence. Such a

proposition would be in accordance with possible world semantics. The

extension is determined by determining the object that uniquely satisfies

the concept F in a world, which in our example will be Benjamin Frank-

lin in the actual world. However, Benjamin Franklin is not a component ofthat proposition, only the concept F is; the man himself is a component of

the world. The proposition is made up of concepts or intensions or senses,

not references or extensions. References exist in the objective world, not



100 Chapter 5

inside propositions—there is no room for them inside propositions. Propo-

sitions, for the Fregean possible worlds theorist, are made of intensions, not

extensions.

5.2 Kaplan on Indexicals

Kaplan disagrees with the picture of meaning painted by possible world

semantics because of the presence of indexicals in language. He thinks

indexicals must be analyzed in a different way and that a very different

conception of meaning is needed to represent the meaning of indexicals. In

the very beginning of the article, Kaplan introduces the idea of the seman-tics of direct reference:

If there are such terms, then the proposition expressed by a sentence containing

such a term would involve individuals directly rather than by way of the “individual

concepts” or “manners of presentation” I had been taught to expect. Let us call such

putative singular terms (if there are any) directly referential terms and such putative

propositions (if there are any) singular propositions. Even if English contained no

singular terms whose proper semantics was one of direct reference, could we deter-

mine to introduce such terms? And even if we had no directly referential terms andintroduced none, is there a need or use for singular propositions?

1

Kaplan defines a singular proposition in contrast to the traditional defi-

nition. A singular proposition will not contain the concept or intension

corresponding to “Benjamin Franklin.” It will contain the actual man Ben-

jamin Franklin. The real Benjamin Franklin is a constituent of the singular

proposition in the same kind of way that a concept can be a constituent

of a general proposition. This is very much opposed to the classic Fregean

model, because there are now actual concrete individuals in the propo-

sition. The notion is more in line with Russell’s view that certain terms

(genuine names) introduce into the proposition the reference of the term.

Russell made a sharp distinction between a term that introduces a con-

cept (e.g., a description) and a term that introduces an object (e.g., a logi-

cally proper name). Kaplan is advocating a return to Russellian semantics

as against Fregean semantics, because he thinks of a singular proposition as

one that contains concrete individuals. If a directly referential term occurs

in a sentence, the singular proposition contains the object of reference

without the intermediary of a Fregean sense. Kaplan thinks that when it

comes to indexicals, this view is the right one.




In the Fregean story, the word expresses the sense and that sense deter-

mines the reference, which is a particular individual. Therefore, when the

word refers to the individual it does so indirectly by expressing the sense.The sense is the propositional component, the thing that enters into the

proposition. The sense determines the reference by being the concept of

a certain individual, but that individual is not a component of the prop-

osition. As an indirect result of this relationship of expressing, the word

denotes the individual. The direct reference story is very different. There

is the word and the referring relation and the individual, and that’s all.

The expressing relation and the sense, with sense determining reference,

are now cut out of the story. Kaplan does bring in more linguistic machin-ery later, but the propositional component is constituted simply by the

individual. The individual is the propositional component, which is why

he writes that the relationship is identity. The individual thing referred to

is literally identical to the propositional component. The word does not

refer in a mediated way via the sense; it refers directly to the individual. The

propositional component is the meaning, and so the meaning turns out to

be an individual inhabiting the world outside language.

One very big difference between the Fregean model and the direct refer-

ence model is that in the Fregean model many senses can correspond to

the same reference. This cannot happen in Kaplan’s model, because the

individual determines the sense, not the other way around. The proposi-

tional component is the meaning, which is determined by the reference,

and that relation is simply identity. Therefore, there can only be one sense

for each reference, so that coreferential terms must have the same sense.

Kaplan’s model does not recognize Fregean cases with two terms having the

same reference but different senses. However, as we have discussed several

times, this account of the meaning of names runs into Frege’s problem of

identity. Although the direct reference model is attractive in some ways,

Frege thought that his apparatus of sense and reference is needed to solve

the problem of identity. Kaplan does not attempt to confront Frege’s prob-

lem in this paper, concentrating on other questions, but we must keep it in

mind as we proceed. On the face of it, it seems impossible to deal with cases

like “Hesperus” and “Phosphorous” in terms of reference alone. At any rate,this is a challenge to direct reference theories.

What is an indexical? Demonstratives are a subclass of indexicals. A

demonstrative is a word like “that” or “this” that is typically accompanied



102 Chapter 5

by a pointing gesture. Indexical words include ones like “here,” “there,”

“you,” “he,” “I,” and “now.” The basic idea of an indexical is that it is a

word that is used in a particular context and which depends for its refer-ence on that context. Accordingly, we can call an indexical a context-depen-

dent expression. Indexical words therefore differ from names and definite

descriptions, even though some definite descriptions contain indexicals.

Kaplan also makes the qualification that he does not mean to include

indexicals that are used anaphorically, as in “John went to the shops, and

he bought a sandwich there.” Instead, he is interested in indexicals as they

are used without borrowing their reference from a previous singular refer-

ence (as with “he” and “John”). His concern is to understand the semanticsof these words. The notion of direct reference will play a big part in this.

5.3 The Two Principles of Indexicals

Kaplan tells us that two basic principles about indexicals will guide his dis-

cussion. First, indexicals are context dependent: the reference of an index-

ical depends on the context in which it is uttered. If Rafael Nadal says,

“I’m hot,” he is referring to himself because the context of that utterance

includes the speaker. If you, the reader, say, “I’m hot,” the context is a dif-

ferent one, and so it refers to you, the reader. Definite descriptions and

proper names do not have this property of context dependence: if you say

“Rafael Nadal,” you refer to the same person that Nadal does when he says

that name—you don’t refer to yourself!

The second principle is that indexicals are directly referential. A directly

referential term is one in which the proposition expressed by the indexical

sentence is a singular proposition. If a speaker says, “I’m hot,” the proposi-

tion expressed by that sentence will consist of the speaker (the person “I”

refers to) and the property of being hot. Kaplan thinks that indexicals are

directly referential in the way that Russell and Mill thought that names are

directly referential. The reference is not mediated by a descriptive concept

that identifies an object uniquely.

Kaplan’s view of indexicals is rather like Kripke’s view of names: both

go against description theories of what determines the reference of theseexpressions. Kaplan thinks both names and indexicals are directly referen-

tial. So semantically, indexicals are namelike in the Russellian sense. Since

names are rigid designators, it would follow that indexicals are also rigid




designators. Kaplan does think that indexicals are rigid designators. How-

ever, he thinks using that terminology confuses two distinct notions that

he believes need to be kept separate.A description that is a rigid designator does not differ semantically from

a description that is a non-rigid designator. It is not directly referential. The

propositional component is the same as it was before—a concept. The com-

ponent of the proposition expressed by the rigid designator “the successor

of 3” will be the concept of the successor of 3. It will not be the number 4

itself. In the case of the rigid description, the propositional component is

a concept (not an individual), so a rigid description is not a directly refer-

ential device. It does not create a singular proposition, but rather creates ageneral proposition. Its components consist of a general concept (the mean-

ing of the description) and whatever is predicated. This becomes clearer if

we consider Kripke’s necessity of origin example. Consider a person with

origin O. The propositional component corresponding to “the person with

origin O” is just the general concept of having origin O. Semantically, the

description functions the way it does when the description is not rigid—the

propositional component is a general concept. Therefore, it does not follow

from the fact that the expression is a rigid designator that it is also directly

referential. Descriptions can be rigid without being namelike. Another pas-

sage from Kaplan’s text explains this point:

For me, the intuitive idea is not that of an expression which turns out to designate

the same object in all possible circumstances, but an expression whose semantic rules

provide directly that the referent in all possible circumstances is fixed to be the actual

referent. In typical cases the semantic rules will do this only implicitly, by provid-

ing a way of determining the actual referent and no way of determining any other

propositional component.2

Kaplan’s idea of direct reference is not the idea that the term designates the

same thing in all possible circumstances. Rigid designation can arise from

individual essence independently of the rules of language. It can result

from the facts of metaphysics. Origins are metaphysically necessary, but

that is not a semantic idea. Rather, individual essence is something that

comes from the nature of numbers and the nature of people. Direct refer-

ence is meant to be a property of an expression that results from its statusas a piece of language. The semantic rules that are part of the very meaning

of the expression determine that it is directly referential.



104 Chapter 5

Kripke uses some terminology in Naming and Necessity that is relevant

to our present discussion. A de facto rigid designator is said to be one that

designates the same object in every possible world as a matter of metaphysi-cal fact (e.g., “the successor of 3” or “the person with origin O”). On the

other hand, a de jure rigid designator is one that designates the same object

in every possible world because of its meaning or the semantic rules that

govern it. For Kripke, names are de jure rigid designators, but rigid descrip-

tions are de facto rigid designators. Kaplan believes in a similar distinction

between rigidity and direct reference. Rigidity is not the same notion as

direct reference because there are rigid descriptions without direct refer-

ence. Here is Kaplan again:

If I may wax metaphysical in order to fix an image, let us think of the vehicles

of evaluation—the what-is-said in a given context—as propositions. Don’t think of

propositions as sets of possible worlds, but rather as structured entities looking some-

thing like the sentences which express them. For each occurrence of a singular term

in a sentence there will be a corresponding constituent in the proposition expressed.

The constituent of the proposition determines, for each circumstance of evaluation,

the object relevant to evaluating the proposition in that circumstance. In general,

the constituent of the proposition will be some sort of complex, constructed from

various attributes by logical composition. But in the case of a singular term, which

is directly referential, the constituent of the proposition is just the object itself. Thus

it is that it does not just turn out that the constituent determines the same object in

every circumstance, the constituent (corresponding to a rigid designator) just is the

object. There is no determining to do at all.3

This passage clearly distinguishes between rigidity and direct reference.

The proposition that corresponds to a directly referential term is a singular

proposition. The proposition that corresponds to a rigid description is a

general proposition, because descriptions are not directly referential. The

terminology Kaplan uses is similar to Russell’s. Russell would say that a

sentence that contains a definite description expresses a general proposi-

tion because it is equivalent to a quantifier sentence. The general proposi-

tion that is expressed by that sentence may look like a singular proposition,

because it is a grammatically singular sentence; but that is a grammatical

illusion, because it is logically a general proposition. But there is also a

class of expressions Russell calls names (Kaplan calls them directly referen-tial), where the proposition expressed is a singular proposition instead of

a general proposition. The idea of singularity of propositions is captured

in the representation of propositions as containing individual things as




constituents. Rigidity is just the idea of having the same reference in every

world, but direct reference is the idea of what constitutes the corresponding

proposition. Rigidity is a modal notion, but direct reference is a semanticnotion.

If we consider this issue from the perspective of the speaker, we can ask

what he understands when he grasps propositions of different types. In

the case of descriptions, rigid or non-rigid, the speaker is grasping some-

thing general constituted by concepts. In the case of the directly referential

term, though, the speaker grasps an individual, and that individual occurs

in the very content of the proposition that has been grasped. If a speaker

says, “This room is nice,” the proposition before his mind at that momentcontains a certain actual room. There is a sense in which that room is a

part of his mind, a part of the proposition he grasps. One consequence of

this is that if there is no such room (e.g., he is hallucinating), then there

is no such proposition. Since the speaker has used a demonstrative, he has

directly referred (ostensibly) to a room that does not exist. There is then no

singular proposition that he has succeeded in expressing. Thus it is possible

to think that one is expressing a singular proposition when one is not really

expressing such a proposition—as when one hallucinates a particular object

and says, “That is F.” For instance, you might hallucinate a tiger and say,

“That tiger looks fierce.” But since there is no tiger there, you have failed to

express a proposition containing a particular existent tiger. Singular propo-

sitions are object dependent, so they fail to exist if the intended object fails

to exist. Direct reference thus can give rise to illusions of propositions. But

that can’t happen in the case of purely general propositions.

5.4 Context of Use and Conditions of Evaluation

To further distinguish rigid designation from direct reference, Kaplan out-

lines the distinction between context of use and conditions of evaluation.

This is a very important distinction. Context of use consists of the person,

the time, and the place in which a given sentence can be uttered. A circum-

stance of evaluation is a possible world where a given proposition can be

either true or false. We have to distinguish quite clearly between the twoconcepts. The reason we might fail to see the distinction is that different

contexts of use produce different references. When I say “I,” I refer to me,

and when you say “I,” you refer to you. Therefore, different contexts of



106 Chapter 5

use with the same indexical term produce different references. Accordingly,

they can produce different truth-values, because I might be what I say I am

while you might not be what you say you are.We may wonder if this is the same as what happens in the case of a

description (e.g., “the inventor of bifocals”) with different references in

different possible worlds. Don’t we have variation in extension with con-

stancy of intension in both cases? Kaplan’s point is that we should not con-

flate those two kinds of dependence of extension. We should not confuse

dependence on context and dependence on world. Consider the sentence

“I do not exist.” If a speaker says, “I do not exist,” that can never be truly

uttered, because it can’t be uttered by someone without that someone exist-ing. Take any context of use and it will always be false, because the context

includes the speaker. In any context in which somebody says, “I exist,” it

is always true (cf. Descartes’s point in the Cogito). It is necessarily true in

the sense that in any context in which the sentence is uttered it must be

true. However, it is not a necessarily true proposition that the speaker who

uttered the sentence “I exist” exists. Along with anyone else who utters that

sentence, he might not have been born. There are possible worlds in which

that speaker would not be alive to utter the sentence “I exist.” No one

exists necessarily (except perhaps God, if he does exist). Therefore, there is

a big difference between the context of use and the circumstances of evalu-

ation. The circumstances of evaluation concern the extension of the propo-

sition expressed once it has been expressed, but the context of use concerns

which proposition gets expressed to start with. Thus context determines

which proposition is expressed by a use of “I,” but circumstance determines

whether a particular proposition so expressed is true in a world.

For this reason, Kaplan makes a firm distinction between context of use

and circumstances of evaluation. And the first point that Kaplan makes

against possible world semantics is that it blurs that distinction. It does not

recognize the difference between circumstances of evaluation and contexts

of use because it just talks about descriptions and intensions in relation to

possible worlds. All we have in possible world semantics are circumstances

of evaluation, where different circumstances give different extensions for

a given intension. It does not have the notion of a context of use. It dealswith the modal notion of variation of extension with possible circum-

stances, not the notion of a context that fixes what was said on an occa-

sion. In effect, it treats all of language as context independent (this is not




surprising, given that it deals with languages modeled on standard formal

logic and these languages contain no indexicals).

This discussion of context dependence leads up to a distinction thatKaplan draws between what he calls character and content . This is the heart

of his theory. All the points we have just made can be formulated by means

of these notions of character and content. Fortunately, this distinction is

easier to understand than some of the earlier points Kaplan makes. Con-

sider a word like “I” or “here” or “now” and look at its meaning. The mean-

ing such words have whenever they are uttered is called “character.” It is

what the word means in the language—its lexical meaning. Roughly, this

meaning or character specifies that if one utters the word “I,” it refers to thespeaker, whoever he or she is. “Here” is the word that you use to refer to the

place you are in, wherever that place is. Similar definitions hold of “there”

and “now.” Character captures the meaning of those indexical expressions,

because it determines what is referred to by those expressions when they

are uttered in a particular context. Essentially, character is the dictionary

meaning of the word. It is important to note that the word has the same

character no matter what context it is used in. If Jack says the word “I” and

John says the word “I,” there are two different contexts of utterance, but

the word “I” has the same meaning in both and thus the same character.

Character seems close to the Fregean sense of a word, because the sense

of a word corresponds to its linguistic meaning. But there is a big difference

between character and Fregean sense. Character does not by itself deter-

mine reference, whereas Fregean sense does. Character does not determine

reference because when John says “I” and Jack says “I” they utter the word

with the same character but not the same reference. Therefore, the mean-

ing of an indexical is not a sense in the Fregean understanding of the term.

The context in which an indexical is used also works to determine its refer-

ence. It cannot be done by the character alone. Obviously, a speaker can-

not say the word “I” and succeed in referring to a certain place. He must

use the word with the correct linguistic meaning. But the character is too

general and unspecific to tie down a unique reference without contextual

supplementation. Consequently, both character and context determine ref-

erence. Both of those factors operating together fix what the speaker refersto. Character, then, is very different from sense. In the case of sense, sense

determines reference, and there is no need to bring in the context of use.

With Frege, we learned that sense determines reference regardless of the



108 Chapter 5

context of use. Unlike sense, however, character requires interaction with

the context of use to determine reference.

The full meaning of an indexical utterance cannot consist of its char-acter alone; if it did, the full meaning of a sentence would not determine

the proposition it expresses. The proposition that is expressed is something

separate from the character. Kaplan calls the proposition expressed by the

sentence its content . If I say, “I’m hot,” and you say, “I’m hot,” we express

different contents, because we are talking about different people. The sen-

tence we both uttered has exactly the same character, because the same

character is expressed by the sentence no matter what the context is. But

a different propositional content is expressed by the sentence in the twocontexts. The content is a product of both the character and the context.

Content includes reference, but character does not. Content is what has

truth-value in different possible worlds, but character is what interacts with

context to produce content—character by itself cannot have truth-value.

Another reason why the content is separate from the character is that

the same content can be expressed by a sentence with a different character.

An utterance of the sentence “I’m hot” expresses a content with a certain

character, but the same content could be expressed by someone who utters

the sentence “you are hot” in reference to the person who uttered the first

sentence. In those two utterances, there is the same proposition and the

same content, but a different character. Therefore, character does not deter-

mine content and content does not determine character. They are two inde-

pendent semantic dimensions of an indexical utterance.

Thus the total meaning of an indexical utterance has two parts or

aspects: the character and the content. There is no single straightforward

entity called “meaning” or “sense” because the indexical utterance has two

different semantic dimensions. In Kaplan’s picture, indexicals have two

sides to their meaning, whereas in Frege’s picture there is only one side,

the Fregean sense. The reason for this is that Fregean sense is supposed to

determine reference. However, in the case of indexicals, their ordinary lexi-

cal meaning does not determine their reference, because their reference is

context dependent.

Context dependence is the central pillar of Kaplan’s theory of indexicals.All the other aspects of his theory stem from that one major point. Kaplan

is saying that Frege is wrong to suppose that the linguistic meaning of an

expression is always a reference-determining sense. Frege’s theory works




perfectly when applied to context-independent definite descriptions. The

thing that determines the reference of a definite description is the very

thing that constitutes the linguistic meaning of it. But in the case of indexi-cals, they do not coincide. Fregean sense, and its descendant, possible world

intensions, cannot accommodate indexical expressions. These are modeled

on the case of the pure definite description, but indexical terms are nothing

like pure descriptions. They are directly referential and context dependent,

whereas descriptions are neither of these things.

5.5 Possible Worlds, Meaning, and Indexicals

Consider the two sentences “The queen of England is pregnant” and “I

am pregnant.” To better understand the semantics of these two sentences,

imagine that the second sentence is uttered by Queen Elizabeth II. She

refers to herself with the word “I,” and she is also the denotation of “the

queen of England,” so we have coincidence of reference. We have already

talked about the many reasons why these two sentences are not synony-

mous. Now we are interested in what Kaplan believes to be the essential dif-

ference between the two sentences. The first sentence expresses a meaning

and that meaning is an intension. The intension is a function from possible

worlds to truth-values. If we consider just the definite description, it will

express a function from possible worlds to objects. In the actual world, that

function gives us the object Queen Elizabeth II. But in other possible worlds

the description might designate a different individual. It is not necessar-

ily the case that Elizabeth II is England’s present queen. Since “the queen

of England” is a non-rigid designator, the intension corresponding to the

meaning of that description will determine a different object in different

possible worlds. Notice that this description is completely context indepen-

dent. It does not matter in what context it is said; it will have the same ref-

erence. All that matters is that the intension is something that determines

a certain object when given a certain possible world as argument. To use

Kaplan’s terminology, certain circumstances of evaluation fix which object

that description refers to, and these can vary.

Kaplan is arguing that this model applies only to certain types of expres-sions. Indexicals are a class of words to which this model does not apply.

If we return to our example above, Kaplan believes that the description

“the queen of England” is a non-rigid designator that does not directly



110 Chapter 5

refer. The propositional component corresponding to the description is an

individual concept, not a particular object (an actual thing in the world).

It is not directly referential (in the Russellian sense). Kaplan suggests thatindexicals cannot express such intensions, which are context independent,

so their meaning cannot be understood as functions from possible worlds

to extensions. The meaning of the sentence “I am pregnant” is a character

(in the technical sense that Kaplan gives to “character”). Character is not an

intension from possible worlds to extensions—it is not something that can

be applied to a world to determine what the extension is of that term in the

world. The meaning of the word “I,” for example, is common to everybody

who uses the word “I,” so it is impossible to look into a possible world anddetermine what the reference of the word “I” will be in that world. It has

none, considered out of context.

Character is nothing like a classic intension in possible world seman-

tics. The sentence “I am pregnant” considered by itself does not express

a proposition at all. A proposition must be something that is true or false.

That sentence by itself is not true or false but must be uttered in a context

first. If a man says, “I am pregnant,” that would certainly not be true. If a

woman who is pregnant says that sentence, then it is true. The character

alone fails to determine a proposition. The character is thus not a function

from worlds to extensions. An indexical sentence can of course express a

proposition on an occasion, but the context must be added to the character

to produce anything propositional. The combination of the character and

context determines the proposition. Kaplan gives us the equation:

Character + Context = Content

Content is what is said, stated, asserted—and that is a proposition. The con-

tent is not the same as the character. It is something produced by the char-

acter when it is combined with the context. The content is what the speaker

states when he uses the particular sentence in a specific context. This con-

tent corresponds to the classic notion of intension, but the character that

produces it does not. Rather, character is best conceived as a function from

context to content . The function here is not from worlds to truth-values.

Rather, it is something that expresses a relationship between the context

and what is said when the expression is uttered. Character determines

(with context) what you say; it does not determine whether what you say

is true—that depends on the circumstance of evaluation. In the case of




character, the function takes contexts as arguments and produces contents

as values—whereas contents are functions that take worlds as arguments

and produce truth-values as values.So two very different functions are involved in an indexical utterance,

and Kaplan’s entire point in his paper is that we should not confuse these

two functions. In one case (“The queen of England is pregnant”), a fixed

intension for the description combines with different circumstances to

give a particular extension (e.g., whoever “the queen of England” refers

to in a particular world). In the other case (“I am pregnant”), there is no

fixed intension, and the reference of “I” can vary as different propositions

are expressed in different contexts. We must not confuse the way contextcontributes to extension with the way circumstance contributes to exten-

sion. Definite descriptions like “the queen of England” are context inde-

pendent, but indexicals like “I” are context dependent. Therefore, what is

said when using an indexical depends on the context, but this is not so for

descriptions. Descriptions float free of context, but indexicals are steeped

in context.

A number of consequences flow from the distinction between charac-

ter and content. One is that not all meanings are intensions. There can-

not then be a complete theory of meaning based on classic possible world

semantics. There are two kinds of lexical meaning: character-type meaning

and content-type meaning. There is only one type of meaning in the classic

intension-based semantic theory—Fregean sense. But according to Kaplan,

there are two irreducibly distinct types of meanings. Thus the meaning of

an utterance of the sentence “I am pregnant” is given in two stages. One

stage gives the character, which is a function from contexts to contents,

and the other stage gives the content, which is a function from worlds to

truth-values. This type of theory is sometimes called dual-aspect semantics.

It is a rejection of the one-dimensional Fregean picture of things. Frege

did not consider indexicals when writing “On Sense and Reference.” In a

later essay called “The Thought,” Frege does discuss indexicals and picks

up on some of these issues. But he did not originally design the theory of

sense and reference in “On Sense and Reference” by looking carefully at

indexicals. He was mainly interested in mathematical language, which isa context-independent language. Hence his examples are all context-inde-

pendent names and descriptions, for which a one-dimensional semantics

is adequate.



112 Chapter 5

There are now two kinds of semantic compositionality, as Kaplan points

out. The two ways that the meaning of a complex expression can depend

on its parts are through character compositionality and content compositional-ity . An example will illustrate this point. If the queen of England says, “I’m

pregnant,” and another speaker says, “She’s pregnant,” the indexical has

been changed. The character of “I am pregnant” is different from the char-

acter of “She is pregnant.” However, the content is the same, so the content

of the whole thing, the proposition expressed by it, does not depend on the

specific character of the words. Here there is the same content but different

character, but there are also cases where the same character can have differ-

ent content. The two are not connected to each other in any simple way,at least not in the way Frege had supposed. There are two sorts of composi-

tionality because there are two different levels of meaning. Different sorts

of semantic unit are combined together to produce complex expressions.

A terminological issue arises here: one might assume that the Fregean

theory of meaning is two leveled. Relative to Russell’s theory, Frege’s theory

is two leveled, because Russell thinks there is only one level, the level of

reference. Russell handles everything concerning meaning beyond the sim-

ple level of name reference with the theory of descriptions. To him, every

primitive expression means what it does by virtue of denoting something.

In Russell’s system, predicate expressions denote universals (e.g., the predi-

cate “red” denotes the universal of redness). Russellian semantics is one-

dimensional because ultimately there are only references. In Frege’s view,

there is the sense and the reference, so it seems right to suppose that his

theory is two leveled. However, such an assumption is ill founded, because

in Frege’s view reference is not constitutive of meaning. In Frege’s theory,

the sense is the meaning, and only the sense. Reference is outside of mean-

ing, which is why words can be meaningful even if there is no reference.

Although Frege’s theory recognizes a level of meaning above reference, his

theory of meaning is still one-dimensional, because sense does all the work.

Kaplan’s theory can be characterized as having two levels or three levels,

depending on how each level is understood. Kaplan’s theory of meaning

has two levels—character and content—and both of these correspond to

the intuitive idea of what somebody meant when he uttered a sentence;but there is also the level of reference. So we might speak of three levels in

the same spirit as thinking of Frege’s theory as having two levels. What is

important is that Kaplan breaks Fregean sense into two and hence intro-

duces an extra semantic level.




5.6 Kaplan on “Today” and “Yesterday”

Finally, Kaplan talks a bit about the words “today” and “yesterday.” Thisdiscussion raises a tricky problem for Kaplan in the end. Suppose I say on

a certain day, “Today it’s raining.” How do I say the same thing as I said

today tomorrow? Suppose I say tomorrow “Today it’s raining”—have I said

the same thing as I said the day before when I said, “Today it’s raining”?

Suppose the first day was Tuesday: then the first use of “today” referred to

Tuesday and the second to Wednesday. So I have not said the same thing.

I have referred to Tuesday in the first case and Wednesday in the second

case. The same indexical word cannot refer to the same day on consecutivedays. To say the same thing on Wednesday as I said on Tuesday, I have to

say, “Yesterday it was raining.”

Clearly the words “today” and “yesterday” are not synonyms of one

another. They have different meanings even while referring to the same

thing. And yet, in a very intuitive sense, these two sentences manage to

say the same thing. They do not say the same thing in the sense that they

have the same linguistic meaning, because the sentence “Today it’s raining”

and the sentence “Yesterday it was raining” do not have the same linguistic

meaning. However, each can say the same thing as the other depending

on the speaker’s context. So, in Kaplan’s terminology, two sentences with

different character can say the same thing. What makes this thing said the

same? Kaplan might suggest that it is the identical reference of the two

terms. But as we have repeatedly seen, just because the reference of two

terms is the same does not mean they have the same propositional con-

tent. We know, for example, from “Hesperus” and “Phosphorus” that these

names do not say the same thing. If somebody says, “Hesperus is a planet,”

it would be wrong to report him as having said that Phosphorus is a planet.

But in the case of indexicals for days, it is necessary to use a word (“yester-

day”) that has a different meaning from the first word used (“today”) in

order to say the same thing. We have to change the meaning to keep what

is said the same! Something strange is going on because the meaning of the

word is being pulled apart quite radically from what is said by using the

word. The question is whether Kaplan has the resources to capture this ideaof what is said: is it character or is it content? It can’t be character because

the characters are different; but how can it be content if content is just a

matter of reference? We will go into this in more detail in the next chapter.





6 Evans on Understanding Demonstratives

6.1 The Fregean Theory of Indexicals

Kaplan takes indexicals to refute Frege’s theory of meaning, at least for their

case. In particular, the Fregean notion of sense does not apply to indexicals.

Gareth Evans, however, questions this conclusion, arguing that it is pos-

sible to develop a Fregean interpretation of indexicals. In such a theory, we

would be able to apply the theory of sense and reference to indexicals. We

know that it is not possible to do this by equating the sense of the indexi-

cal with the conventional linguistic meaning (character) of the indexical,

because then sense would not determine reference. Different people can use

the same indexical word with the same meaning and yet make different ref-

erences. Sense cannot be identified with the standard conventional mean-

ing of an indexical word if we are to create a theory of indexicals where

sense determines reference. To establish a Fregean theory of indexicals, we

must find a sense for them that goes beyond their conventional meaning,

that is, their Kaplanian character. What would this sense be like?It can’t be character, but can it be content? No, the sense can’t be identi-

cal to the content either, in Kaplan’s sense, because in Frege’s system senses

are never identical to references and there are always many senses corre-

sponding to the same reference. Content for Kaplan is just a singular propo-

sition, constituted by reference alone. Therefore, it would be impossible

for the sense to be identical to the reference, because then there would be

only one sense for a given reference. The sense of the indexical as uttered

by someone will be identical neither to its character nor to its content. And

there does not seem to be anything left in Kaplan’s scheme for Evans to

equate with Fregean sense.



116 Chapter 6

A possible answer is that the sense of an indexical is neither the charac-

ter nor the content but rather the description a speaker has in mind when

she uses an indexical. This suggestion borrows from the description theoryof names. When a name is used it is held to be synonymous with a descrip-

tion that the speaker has in mind that uniquely applies to the bearer of the

name. Analogously, we might propose a description theory for indexicals,

suggesting that when a speaker uses an indexical, she has a description in

mind that is synonymous with the indexical as then used, and that this

description uniquely applies to the object of reference.

Suppose I say, “I’m a philosopher.” Then we might suggest that the

description I have in mind is “the author of The Subjective View ,” since Iwrote that book. So on this Fregean description theory of indexicals, when

I use the word “I” the sense of it is expressed by “the author of The Subjective

View .” When you, the reader, use the word “I” you have a description in

mind that uniquely applies you, and thus you refer to yourself with “I” in

virtue of this mediating description. Just as with the description theory of

names, the proposition expressed by a sentence of the form “I am F ” would

be represented using a general concept expressed by a particular definite

description. This indexical sense would function just like a classical inten-

sion in possible world semantics.

We might go on to apply Russell’s theory of descriptions to the descrip-

tion associated with the indexical, thereby combining Frege’s view with

Russell’s. We then have a description theory of the meaning of individual

occurrences of the word “I” that takes these occurrences to be equivalent

to quantified propositions of the Russellian form. When I say, “I’m a phi-

losopher,” what I’m saying is “There exists an author of The Subjective View

and there is only one such and he is a philosopher.” There is no Kaplanian

direct reference in this paraphrase, just quantifiers and predicates.

Evans uses some terminology that might not be familiar to some readers.

The word “I” as uttered on a particular occasion is called a token of the word.

The word “I” that is common to all tokens of it is called the word type. You

use the same word type when you say “I” as when I say “I,” but you utter

a different token of that type. If I say “I” at a given time, that is a differ-

ent token from my saying “I” at a later time. Nevertheless, each utteranceconsists of a token of the same type. Tokens are events that occur at par-

ticular times and places, but types are more abstract. The Fregean theory of

indexicals claims that we should analyze tokens of indexicals as expressing



Evans on Understanding Demonstratives 117

Fregean senses, and then equate each of those tokens with descriptions (at

least according to this first style of Fregean theory). The description might

be constant from token to token, as it is with token utterances of names.However, we are trying to accommodate the point that someone else can

use the word “I” and refer to a different person, not me, so we will need a

different description with a different reference. We can conclude from this

that the word “I” is ambiguous, according to this theory, because it has dif-

ferent senses on different occasions. It would be rather like having a room

full of people all named “John Smith.” No John Smith would be identical

to any other—hence “John Smith” would have variable sense and reference

across this room of people. The name “John Smith” in that case would beambiguous. Similarly, “I” would be ambiguous, having different sense and

reference in different contexts. The type is ambiguous, that is, though the

tokens would all have a specific sense and reference. A definite description

for each of them would give the sense of the token, but as a type the word

would be ambiguous.

That is one possible idea for how to handle indexicals Frege-style—by

proposing a description theory of the sense of indexical tokens. The seman-

tics of indexicals would consist of three elements: character, content, and

a description that captures the sense on a particular occasion of utterance

(the token sense). In this picture, indexicals are not directly referential.

The word is synonymous with a description, and the description has an

intension that is context independent. The role of context is just that dif-

ferent individuals use the same (type) word and they associate different

descriptions with it, with these descriptions determining what they refer

to. It is necessary here to distinguish between the descriptive sense and the

character. The word has the same conventional meaning (character) in its

different uses, but the sense varies from context to context. So it’s not that

once sense is admitted we can do without character: we will have character,

sense, and reference in our final semantic theory.

The author that Evans is mainly criticizing, John Perry, assumes that

the theory we have just outlined is the correct Fregean model, because

he thought it must be some kind of description theory of sense. Evans’s

reply is that Perry has overlooked a different kind of Fregean theory, onenot built around definite descriptions. There could be other ways to think

about sense than descriptively, he believes, and these other ways are equally

Fregean. Not all sense has to be descriptive sense, he contends. He agrees



118 Chapter 6

with Perry that a description theory of indexical sense is a very implausible

idea. It does not seem at all attractive to suppose that people have uniquely

identifying descriptions in mind when they use these terms. It is unappeal-ing to think that context plays no substantive reference-determining role.

Perry gives us a very neat argument against this type of position, as follows.

6.2 The Point of Indexicality

The point and essence of indexicality is best understood by considering

two types of examples: mirror examples and amnesia examples. Let us first

examine mirror examples. Suppose you are sitting in a restaurant and yousee the reflection of a man or woman in the mirror in front of you, and

you have the following thought about the individual in the mirror: “That

person is very good-looking.” You could have certain other beliefs about

the person in the mirror too—say, that he (or she) seems rather pleased

with himself (herself). Though this is improbable, it is perfectly conceivable

that the person in the mirror is you, but you didn’t realize it for a second or

two. Suddenly, you are thunderstruck with the realization, “Oh, it’s me I’m

seeing!” You had referred to yourself without realizing it. This tells us that

when you refer to yourself with “I” it cannot be via the kinds of descrip-

tions that truly apply to you in the mirror reflection, because then you

would have to realize the truth of “I am the person reflected in the mirror.”

The word “I” cannot mean these descriptions. It is informative to discover

the truth of “I am the person in the mirror,” so this cannot be a tautology,

which it would if “I” (that token) were synonymous with “the person in the

mirror.” Almost any description is such that it is potentially a discovery that

you are the person so described.

Another, more extreme example that makes this point even more clearly

is the amnesia case. Imagine a person who sustains a trauma to the head,

and when he wakes up can’t remember anything. I am going to suppose

that this unfortunate individual is myself. The doctor asks me, “Where

do you live?” and “What’s your name?” and I don’t know because I can’t

remember. I can’t remember any facts about myself. I might say, “I can’t

remember anything about myself”—yet I successfully refer to myself. Sothere I am in the hospital and I don’t know about my past history, and

I start reading a book called The Subjective View . As I read I say to myself,

“The author of The Subjective View is not much of a philosopher.” I report




my opinion to the doctor, who smiles indulgently and replies, “You are the

author of The Subjective View.” I thereby make a substantial discovery, thus

showing that “I” in my mouth does not mean “the author of The SubjectiveView .” And we could have anticipated that because I succeed in referring to

myself with “I” even though I have amnesia. So it can’t be that I succeed in

making this kind of first-person reference in virtue of knowing true descrip-

tions about myself. I certainly don’t refer to myself with “I” by knowing my

famous deeds and well-known facts about myself.

Perry gives this argument, and Evans agrees with him. The upshot is

sometimes called the indispensability of the indexical “I” or the essential index-

ical. The idea is that the word “I” cannot be removed from the language andreplaced by descriptions, because indexical sentences express a different

kind of proposition from nonindexical sentences (e.g., sentences involving

the descriptions we used in the mirror and amnesia examples). This poses a

serious problem for the description theory version of a Fregean account of

the meaning of indexicals. Evans agrees that descriptions don’t work to give

the meaning of an indexical because of this kind of argument. If indexicals

have sense, it cannot be descriptive sense. But what other kind is there?

6.3 Evans’s Theory of Sense and Reference for Indexicals

Since Evans agrees with this point, we may wonder how it would be at all

possible to have a Fregean theory of the meaning of indexicals. There seems

to be nothing else the sense could be except some sort of descriptive concept.

We have already explained how the sense of an indexical can be neither a

character nor a reference, and now we see it cannot be a description either.

To approach the question, Evans tells us what he thinks a theory of mean-

ing should look like. That is, he tells us how sense is related to reference. He

spends the first half of his paper talking about this relationship. First, we will

examine his conception of a theory of reference, then outline his theory of

sense, and finally explain how he thinks the two are related. Then we can

discuss whether or not this overall theory applies to indexicals.

First of all, a semantic theory is founded on a theory of reference. A

theory of reference is an assignment of reference to every meaningfulexpression in a language. And we know that Frege’s position on assigning

reference has two main parts. One part is that if the expression is a proper

name, it will be assigned an object as reference. Proper names, for Frege,



120 Chapter 6

can be ordinary names or definite descriptions or even whole sentences.

Ordinary objects are assigned to ordinary singular terms and truth-values

are assigned to sentences as their reference. In the second part of the theory,Frege also assigned concepts to predicate expressions. In Frege’s system, a

concept is a function from objects to truth-values. In the sentence “Socrates

is a man” the concept corresponds to the word “man” and the argument is

the reference of “Socrates.” When you apply that concept to the argument,

the value of the function for that argument is the True (an object for Frege).

The value of the function would have been the False if we had inserted the

argument Cleopatra into the function, because Cleopatra is not a man. A

truth function is a function from truth-values to truth-values. Connectivesand predicates are logically the same because they both map objects into

truth-values. Since truth-values are objects, they function as arguments for

functions into truth-values. Thus in Frege’s system there is an assignment

of objects to complete singular terms, where complete singular terms can

be proper names, definite description, and whole sentences; but there is

also an assignment of reference to incomplete expressions, like predicates

or sentence connectives, which are assigned concepts. Quantifier expres-

sions are all that is left, and these are classified as denoting second-level

concepts, since they map first-level concepts onto truth-values. The general

point is that the theory of reference in the Fregean model is an assignment

to every expression in the language of a semantic value that is its reference.

The notion of reference is taken very broadly. It is correlative with the truth

conditions of a sentence.

But Frege’s system is meant to be a theory of a speaker’s understanding,

not merely of the truth conditions of his sentences. A theory of sense is

then needed to account for how we apprehend references. It is a theory of

how references come before the mind, and how they are represented in the

mind. A sense, as Frege tells us, is a mode of presentation, and the mode of

presentation is the relationship between an object in the world and the per-

son who makes reference—it is the mode in which the object is presented

to the person’s mind. The way Evans explains this idea is that a sense is a

“way of thinking” of a reference: not so much how it presents itself to me,

but how I think of it—how it enters my thoughts.Evans’s point in regard to this specific part of Frege’s theory of sense is

that it has not stipulated anything to the effect that senses must be descrip-

tions. We have just stated in a very abstract way the idea that senses are ways




we have of apprehending things. Whether or not these ways are descrip-

tions is a completely different question. What is built into the notion of

sense is just that the sense is something that presents the reference.The next question is how to specify what the sense is. From our explo-

ration of Frege’s work, we know that senses are different from references,

but we have not established how to specify them. Frege himself says little

about this question. Fregean senses seem rather elusive in and of them-

selves (can you point to them, stub your toe on them, examine them from

different angles?). Evans’s view is that the sense of an expression is speci-

fied by saying what the reference of the expression is. Suppose we want

to give the sense of the word “Hesperus.” Evans thinks we can give thesense of this word by saying “The reference of ‘Hesperus’ = Hesperus.” This

certainly gives the reference of the name, and certainly the statement is

true. Compare that sentence with the following sentence: “The reference of

‘Hesperus’ is Phosphorus.” Is that sentence true or not? Yes, that sentence

is also true, since Hesperus is Phosphorus. Evans’s claim is that these two

sentences both correctly say what the reference is of “Hesperus” but only

one also specifies the sense. The sentence “The reference of ‘Hesperus’ =

Hesperus” specifies the sense, while the sentence “The reference of ‘Hespe-

rus’ is Phosphorus” does not—though both state the reference. In the first

sentence we have an example of what Evans calls a sense-specifying refer-

ence assignment. The sense is given by stating the reference, but only some

statements of reference succeed in giving the sense.

Evans’s idea is that we can specify the sense of a name by saying what

its reference is, as long as we do it using the right kind of ascription of ref-

erence. In the second sentence, we have stated the reference but have not

specified the sense. Though he does not explicitly say so, the right way to

state reference if we want to specify sense is by using a synonym of the name

we are talking about. The reference can be stated in two different ways, by

using a name with the same sense as the name mentioned or by using a

name with a different sense, that is, by using a synonymous name or a non-

synonymous one. In the first way, the sense has been specified, but in the

second way it has not. Evans’s position, then, is that senses can be specified

only by assigning references, but not all ways of assigning reference conveysense. Here we have said nothing about senses as descriptive concepts. A

sense is a way of thinking of an object, but there is no way to specify a sense

except by talking about the object.



122 Chapter 6

Notice that in this way of formulating specifications of sense, nothing

is said like “The sense of ‘Hesperus’ is so-and-so.” In specifying what the

sense is, we must say what the reference is—there is no way to specify sense“directly.” We don’t speak about senses in specifying them. If we say, “The

reference of ‘Hesperus’ is Hesperus,” intending to convey the name’s sense,

nothing has been said directly about the sense of “Hesperus” itself. This

is different from saying that the sense of the word “bachelor” is given by

the sense of the words “unmarried male.” In Evans’s theory, the sense of

a word is not specified by giving the sense of another word. At this point

Evans invokes a suggestion of Dummett’s that involves using a distinction

of Wittgenstein’s—the distinction between “saying” and “showing.” InWittgenstein this distinction is a matter of obscurity, and we will not cover

it here in detail. Basically, there is an intuitive idea of saying versus showing

that we can illustrate in the following example.

6.4 Saying versus Showing

Imagine someone holding a pen behind his back. He can say, “I have a

pen in my hand,” or he can just reveal his hand and show you the pen.

Either way you come to know that he has a pen in his hand. In the show-

ing gesture he does not say anything about a pen—he just shows it to you.

As a result of showing the pen, the observer gains knowledge without the

mediation of language. Evans is using Wittgenstein’s general intuitive idea

of saying and showing as illustrated in this simple example. The claim is

that reference clauses say what the reference is and show what the sense

is—without stating it directly. In the pen example, an individual came to

know something without it being communicated to her verbally. In the

same way, these reference clauses are supposed to show the sense of “Hespe-

rus” without actually saying what the sense of “Hesperus” is. It is a bit like

my wishing to convey to you that I am English by opening my mouth and

speaking in an English accent, without ever saying “I’m English.” I get the

point across without stating it in so many words.

Evans’s claim is that it is not possible to say directly what senses are; it is

only possible to show what they are. Evans claims this for good reason. Itis difficult to see how Frege can ever specify what a sense is independently

of the reference of a particular expression. Invoking the saying–showing

distinction promises to get Frege out of a tight theoretical corner. It makes




sense of the elusiveness of sense—at least it attempts to do that. Senses

belong to the realm of that which cannot be said but can nevertheless be

shown.The second point Evans wants to make about sense, which follows from

the first point, is that the sense of an expression is reference dependent . Given

the reference-stating way of thinking about sense, an expression’s having a

sense will require that it have a reference. According to Evans, it is not pos-

sible to give a clause specifying the sense of “Hesperus” if there is no such

thing as Hesperus. By asserting “The reference of ‘Hesperus’ = Hesperus” we

presuppose that there is such a thing as Hesperus. We are using the name

“Hesperus” to refer to Hesperus, so we must be assuming that Hesperusexists. Thus, Evans’s mode of specification of sense presupposes the exis-

tence of the reference. For this reason, he thinks, there cannot be senses

without references. Senses are ontologically dependent on references. This

idea of reference dependency, we will recall, derives from Russell. It is the

idea that some expressions have a meaning that depends on the fact that

the expression actually refers to something. In Russell’s theory, the mean-

ing of a name is the actual object denoted. If there is no such object, there is

no such meaning. Evans argues, like Russell, that the sense of names is ref-

erence dependent. He accordingly calls such terms “Russellian.” For these

Russellian terms, there cannot be sense without reference. Names have a

meaning, or sense, which depends on their having an existent reference.

The next point Evans makes is that even though there are reference-

dependent senses, as Russell conceives them, names can have the same

reference and different senses. The sense can be reference dependent, but

that does not mean it is strictly identical to the reference. We can have a

divergence of sense between two coreferring names that are nevertheless

Russellian. Frege would say that “Hesperus” and “Phosphorus” have dif-

ferent senses, and that the sense is something that does not depend on

reference. By contrast, Evans believes that even though those two names

have different senses the senses are reference dependent. There cannot be

a sense without reference (hence they are Russellian), but sense is some-

thing over and above reference and not identical to reference (hence they

are Fregean). In Evans’s semantics, names can be both Fregean and Russel-lian at the same time. The meaning is not reducible to the bearer, but the

meaning depends on the bearer. Evans is attempting to absorb the insights



124 Chapter 6

of Russell about names while still trying to answer Frege’s worries about

identity statements.

6.5 Mock Sense

If names do not have sense unless they have reference, then what about

empty names? Evans argues that, despite appearances, Frege does not really

believe that there can be sense without reference. Evans attributes this posi-

tion to Frege on the basis of what he says about fictional names. A fictional

name like “Sherlock Holmes” apparently has sense and can therefore occur

in meaningful sentences. But such a fictional name has no reference. Soit seems that its sense is not reference dependent. Evans does not accept

that conclusion. He tries to give textual evidence to support his interpreta-

tion of Frege. Frege says: “The logician does not have to bother with mock

thoughts, just as a physicist, who sets out to investigate thunder, will not

pay any attention to stage-thunder. When we speak of thoughts in what

follows we mean thoughts proper, thoughts that are either true or false.”

Evans defends the idea that the sense of an empty fictional name is defec-

tive because such names have a merely quasi sense, a mock sense. He sug-

gests the comparison of empty names with vagueness. Frege himself made

this defectiveness point about vagueness. The predicate “bald” says that

someone is lacking in hair, but it is not precise about a specific threshold

of hairs one must not have to qualify as bald. Frege held that such vague

predicates lack genuine sense. Since there are borderline cases of baldness,

there are sentences containing the word “bald” that are neither true nor

false. However, in Frege’s system sentences cannot express a thought that

is neither true nor false. Therefore, Frege was prepared to insist that vague

predicates lack sense. Vague sentences express merely a quasi sense, not a

proper scientific sense. There can be no vague predicates in science (e.g.,

math and physics). Vagueness is a defect of natural languages.

Frege thus makes a distinction between words that have a proper sci-

entific sense and words that lack such a scientific sense. He was ready to

say that a vague predicate may appear to have a robust sense but not to

possess such a sense when logically examined. Evans argues, analogously,that a fictional name may have a kind of degraded sense but does not have

a strict proper sense. He takes the position that all proper senses are refer-

ence dependent, but the improper mock senses are not reference dependent




(so fictional names don’t have real sense). There is thus a class distinction

between two kinds of sense. There is authentic non-nonsense sense and

there is specious phony sense. Evans thinks that Frege has the resources tomaintain both that upper-class sense is reference dependent and that there

are lower-class expressions whose sense is independent of reference. In the

case of empty names, the putative sense is always low-class, irresponsible,

reference-indifferent sense.

6.6 Empty Names

Philosophers have taken several views about empty names; the question isvexed. Accept it as a given that there is no such god as Zeus, that is, “Zeus

does not exist” is true. What should we say about the sense of that name?

The strict Millian view is that a name has sense only if it has reference, so in

this case the name “Zeus” would have no sense. Indeed, it could not even

be a name if it lacked reference, because it would be rendered meaningless.

But if a name lacks sense, sentences containing it must be meaningless,

which would make “Zeus does not exist” meaningless, instead of true.

Another view is that “Zeus” does have a sense and that sense is con-

tained within a synonymous definite description. The sense of an empty

name is thus no different from that of a name of something that does exist.

We could give Zeus the description “the most powerful of the Greek gods,”

and then the sense of the name would be no more lacking than a name

defined by “the most powerful man on Wall Street.”

A third possibility, noted above, is that the empty name has a kind of

sense, but it is only has a mock or apparent sense. This would be rather like

an impostor pretending to be a big shot: he isn’t really one, but he puts on

a good show. The name has pretend-sense, make-believe sense.

A fourth possibility is that “Zeus” lacks an existent reference but instead

has a Meinongian subsistent reference. The name denotes the most power-

ful of the Greek gods—and though this being does not exist, he neverthe-

less subsists. The sense of the name can be constituted by this shadowy

subsistent reference. This theory of empty names is Mill meets Meinong.

Each of these theories has its attractions and drawbacks. The Millianview, though nice and simple, makes some true sentences come out mean-

ingless. The description theory saves meaning for empty names but runs

into objections as a general theory of names. Meinong’s view gives a smooth



126 Chapter 6

and comprehensive theory, but the ontology has struck many as hard to

stomach. The pretend-sense theory looks reasonable for fictional sentences

like “Zeus smote the Cyclops,” which aren’t part of factual discourse; butisn’t it a plain scientific fact that “Zeus does not exist” is true? The thought

expressed here is not a kind of mock phony thought lacking in truth-value

but a straightforwardly true thought—and how can this be if “Zeus” has

only a mock sense? Evans has described another approach to empty names,

but it is hard to see how it captures the linguistic data adequately.

6.7 Evans’s View of Names

In the next part of his paper, Evans sets out to defend the thesis that names

are Russellian. He writes:

Therefore, on the present conception, the sense of a singular term is a way of think-

ing about a particular object: something that obviously could not exist if that object

did not exist to be thought about.1

He asserts here that if a sense is a way of thinking about an object, there

could not be a sense without the object existing. Let us first consider this

assertion in application to perceiving. Suppose I visually perceive a certain

object, for example, a pen. My perceptual state could be specified by say-

ing which thing I perceive: “CMG is seeing that pen.” In this case, the

perceived object is referred to in the course of characterizing my perceptual

state. My perceptual state is a way of seeing that pen. You might have a dif-

ferent way of seeing the pen because you have a different perspective, but

we all see the same pen. But is it strictly necessary for the pen to be there in

order for me to have a way of seeing it? What if I am hallucinating a pen?Don’t I still enjoy a perceptual state—a way of seeing—even though there

is nothing there?

How can we characterize the perceptual state of someone hallucinating

a pen? Not by saying “He sees that pen,” which presupposes there is a pen.

Rather, we say something like “It appears to him that there is a pen in front

of him.” This kind of sentence does not commit us to the proposition that

there really is a pen in front of the person hallucinating. There is no refer-

ence to any particular pen here. We can thus ascribe a perceptual content

to him without specifying a reference for that perceptual content. This is

fortunate, because there is no such reference.




In general, it is not true that a way of seeing an object can exist only if

the object exists. There can be modes of presentation of objects without the

objects existing. So Evans’s argument that senses are reference dependentbegins to run into trouble. Consider an ordinary definite description, say,

“the queen of England.” The meaning of that description can be charac-

terized as a way of referring to an object. In one’s thoughts it is a way of

thinking of an object (thinking of Elizabeth II as the queen of England).

But Evans does not think that descriptions are reference dependent, because

there can clearly be meaningful expressions like “the queen of England”

without there being a queen of England. For instance, Evans would agree

that “the king of France” is a meaningful description fully endowed withsense, even though there is no reference for that description. Evans’s gen-

eral argument here would imply that since the sense of the description is a

way of thinking of an object, there must be an existing object of which it is

a way of thinking. But it is just a non sequitur to suggest that where there

is a way of thinking there must be an object thought about. There are obvi-

ously ways of painting mythical beasts, but that does not imply that there

are mythical beasts that are painted. So Evans has not shown that senses are

reference dependent and hence Russellian.

Evans also contends that Russellian terms can be Fregean. That is, he

thinks that coreferring terms can have reference-dependent sense and still

differ in their sense. However, that raises the question: What is the differ-

ence between two Russellian terms that differ in their sense? What does

that difference consist in? It certainly cannot consist in their having dif-

ferent references, because they have the same reference. There has to be

something that goes beyond reference to generate the distinction of sense.

Whatever this is, it cannot consist in reference alone. But if there is some

semantic dimension to the name that goes beyond its reference, it should

be possible to have some conception of what that difference is. Is it perhaps

the way the reference is conceptualized? But now we are moving in the

direction of a description theory, and descriptive concepts are not reference

dependent. The semantic difference cannot be explained in purely Russel-

lian terms, because in Russell’s theory that would be just the reference. If

you say there is no difference, then the terms are not Fregean after all.If there is a Fregean distinction between them, it must float free of refer-

ence—as general concepts do. The extra ingredient of sense cannot itself be

reference dependent.



128 Chapter 6

The upshot of all this is that Evans has not succeeded in describing a

coherent alternative to the description theory of sense that might provide

a viable Fregean treatment of demonstratives. He accepts that a descriptiontheory of indexical sense has to be wrong, and then he tries to construct

an alternative nondescriptive Fregean theory. But it remains unclear that

there is any such nondescriptive Fregean alternative, so indexicals appear

to refute Frege’s general semantic principles after all.

6.8 Evans on “Today” and “Yesterday”

Evans makes an important point later in his paper about the words “today”and “yesterday.” Suppose on a certain day D1 I say truly, “Today is cold.”

Now the next day D2 I want to express the same thought I expressed on

D1. I can’t do this by uttering, “Today is cold” on D2, because that will

refer to D2. To express the same proposition as I did on D1 requires the use

of the word “yesterday”—I have to say, “Yesterday was cold.” Intuitively, I

have expressed the same thing on D2 as I did on D1 by using that sentence.

The same thought is expressed on two different days by using these differ-

ent sets of words. These forms of words are related in a certain systematic

way—there are rules for which word to use in a different context to express

the same thing. When we understand these words, we grasp these rules.

There is a very similar linguistic structure in the case of spatial indexicals (as

well as personal indexicals). For example, if I say, “Here is cold” and I then

move away from that place, I must say, “There is cold” in order to say the

same thing: the same proposition is expressed about the original location

from different locations using different words. The indexical used must be

changed when the context of utterance changes.

Evans’s point about these cases is that they apparently require a Fregean

notion of sense, because the sense of the word “today” when it is used on

D1 is the same as the sense of the word “yesterday” when it is used on D2.

As noted at the end of the previous chapter, “today” certainly does not have

the same character (or conventional meaning) as “yesterday.” To capture

what the two words have semantically in common, Evans thinks we need

to invoke Fregean sense. We clearly need a piece of semantic machinery tocapture the commonality when those two different indexicals are used to

express the same thing in two different contexts. Character is not suitable,

because the character is different in the two cases. We might suppose that




although the character is different, the Kaplanian content is the same. In

other words, the reference is the same. The reference of “today” on D1 is

D1 and the reference of “yesterday” on D2 is also D1. The sense in whichthe same thing has been said on those two successive days is captured (it

may be said) by the fact that those two tokens of the indexicals have the

same reference. Notice that this view is a non-Fregean view of having the

same thought, because it makes no distinction between sense and refer-

ence. Frege does not think that having the same reference is ever the same

thing as expressing the same sense. But at least the content is the same on

both days, unlike the character.

Suppose that D1 is a Tuesday and so “today” is tied down to a specificTuesday. D2 would then be a Wednesday. Now there is a relationship

between those two names of days and the two indexical terms. We can

say that Tuesday is identical to the reference of “today” when said on D1,

which is identical to the reference of “yesterday” when said on D2. So D1

can be referred to with “Tuesday,” “today,” and “yesterday.” Now consider

the relationship between saying “Today is cold” on Tuesday and saying

“Tuesday is cold.” The word “Tuesday” here refers to the same day to which

“today” refers. We have the true identity statement “Today is Tuesday.”

There is a truth-value relationship between “Today is cold” and “Tuesday

is cold,” such that if one statement is true the other is also. Each of these

statements has the same Kaplanian content, because “Tuesday” refers to

the same day as “today.” But, intuitively, “Today is cold” does not say the

same thing as “Tuesday is cold.” Each word refers to the same day, but they

have different senses. We can see this from the fact that someone might not

actually know that today is Tuesday when he uses the word “today” to refer

to Tuesday. He might assent to “Today is cold” but dissent from “Tuesday is

cold” because he disbelieves that today is Tuesday. If he later discovers that

today is Tuesday, he would have learned an a posteriori synthetic truth. So

“Today is cold” cannot express the same thought as “Tuesday is cold” even

though the reference is to the same day.

Those two statements (“Tuesday is cold,” “Today is cold”) do not say

the same thing according to Frege’s test for the identity of thoughts. Also,

they intuitively do not say the same thing. However, they have the samecontent in the Kaplanian sense. This case is different from where “today” is

said on D1 and “yesterday” is said on D2. In that case, each of the sentences

does say the same thing, because no new information is acquired when



130 Chapter 6

one discovers that those sentences are related in the way they are. There

is an analytical, logical connection between those two indexicals, written

into the rules for their use. We know that if “Today is cold” is true on D1then “Yesterday was cold” has to be true on D2. But we do not know that if

“Today is cold” is true on D1 then “Tuesday is cold” has to be true, because

“Today is cold” can be truly uttered on days that are not Tuesday. These

two sentences are not synonymous in the ordinary sense of making the

same statement. The word “yesterday” said on D2 captures the same sense

as “today” said on D1, but “today” and “Tuesday” do not express the same

sense. Therefore, the identity of sense between the former pair is not cap-

tured by Kaplanian content, because that content is in common betweenthe latter pair of statements too. Sameness of content is not enough for

sameness of sense. So we need an extra semantic ingredient to capture what

is common to “today” and “yesterday” but not to “today” and “Tuesday.”

We are thus driven to accept a third level, beyond Kaplan’s character and

content, which is closer to Frege’s idea of sense.

6.9 Character, Content, and Information

Now we can combine three semantic elements to explicate the full mean-

ing of an indexical sentence as used on an occasion. The first is character,

the second is content, and the third corresponds to the sameness of sense

that exists between “today” and “yesterday.” Let us call this third layer

information. The same information is conveyed by saying “Today is cold”

on D1 as would be conveyed by saying on D2 “Yesterday was cold.” The

speaker acquires the information from his sense experience on D1 that the

day is cold and that information is stored in his memory. On D2 when he

says “Yesterday was cold,” he is merely referring back to the information he

acquired from the previous day that is stored in his memory. The speaker

has the same piece of knowledge acquired the previous day but he expresses

it by using different words. Therefore, the same information is in the speak-

er’s mind over the two days and he expresses it using these two different

sentences. This notion of information is not reducible to either character

or content. Content is too wide a notion and does not capture the exactmeaning of what the speaker says. To avoid confusion, we might rename

Kaplanian content real-world correlate. The real-world correlate of the index-

ical is the object to which the speaker refers. We can still regard this as a




component of the proposition expressed. We could also naturally rename

character perspective. Perspective incorporates the two different temporal

perspectives the speaker has on the given day—as present and as past. Letus insert this into the proposition too. The same information is expressed

from two different perspectives. It is information about the same real-world

correlate. We should not say that there is only the real-world correlate and

the perspective, because then we could not understand the relationship

between “today” and “yesterday” in the right way. The information is pre-

served over time and then expressed from two different perspectives, but

the information is more like a cognitive state than a real-world correlate.

This can be bundled into the proposition along with the other two ele-ments. None of these propositional components determines any of the oth-

ers, so none is redundant. If we think of the informational component as

descriptive, which is natural, then we shall not insist that the descriptive

information determines a particular day—it might be available on other

days too (so it is not equivalent to reference-determining Fregean sense).

We have three irreducibly distinct and indispensable semantic ingredients:

real-world correlate, perspective, and information.

According to this triple-layer semantics, it turns out that everybody is a

little bit right and little bit wrong about this subject. Kaplan is right to intro-

duce character and content but wrong to think that character and content

are all that is required. Evans believes that only Fregean sense is needed. He

is right to think that there is something common to “today” and “yester-

day” but wrong to suppose that nothing separates them (character). Evans

leaves no room in his theory for this semantic difference: he needs charac-

ter in the full meaning of an indexical utterance as well as sense. The same

information is indeed expressed by these two words on successive days, but

each term has a different conventional meaning. Kaplan and Evans both

offer incomplete theories because they each need something from the oth-

er’s arsenal to fill out the full account of indexical meaning. We need both

character and content, but we also need to recognize that indexicals with

different character can have something in common (what we have been

calling information) that is not reducible to content. The next task would

be to inquire more closely into what this notion of information amounts to(a task we shall leave for homework). All we can say now is that information

is an epistemic notion: it relates to what someone knows. What is clear by

now is that the topic of indexical semantics bristles with complexity and

difficulty, and no current theory has all the pieces worked out.





7 Putnam on Semantic Externalism

7.1 Background

Our earlier discussions of indexicality will help us to understand the force

of Hilary Putnam’s arguments in “Meaning and Reference.” For indexical

expressions, the classic theory of descriptive intensions that determine

extensions looks highly implausible—as Kaplan argues. The meaning of an

indexical when used on an occasion is not equivalent to a definite descrip-

tion of the object or type of object referred to. As Putnam notes toward the

end of his paper, two people can use the word “I” to refer to themselves

even if they don’t differ in the descriptions they would ascribe to them-

selves; so the difference of reference cannot stem from uniquely identifying

knowledge possessed by the two speakers. Here context plays an indispens-

able reference-determining role—and not simply what occurs descriptively

inside the speaker’s mind. What you refer to can depend on who and where

you are, not just on what you think—it depends on external context, not

internal descriptions. That is, indexical reference is fixed externally by thespeaker’s objective context, not by what he has in his mind subjectively.

This is in contrast with descriptive reference, which is context indepen-

dent, because here the speaker’s internal concepts do suffice to fix what he

refers to. Thus externalism is correct for indexical reference, but internalism

is correct for (pure) descriptive reference—as with “the first dog to be born

at sea.” In the case of “I” we just need to know who is uttering the word to

determine its reference, not what that person thinks about his reference.

Putnam’s focus is on natural kind terms like “water,” “aluminum,” and

“tiger”—words that stand for types of object found in nature—not words for

human artifacts like “table,” “computer,” or “president.” He wants to know

what these words mean, particularly how they determine their reference. At



134 Chapter 7

the end of his paper he says, “Our theory can be summarized as saying that

words like ‘water’ have an unnoticed indexical component: ‘water’ is stuff

that that bears a certain similarity relation to the wateraround here

” (notethe indexical “here”). In other words, the semantics of natural kind terms

mirrors the semantics of indexical terms. Such terms do not conform to the

classic Fregean model of the definite description and its reference. Putnam

tells us that it used to be thought that for any meaningful expression there is

an intension that determines the extension in every possible world, and that

when a speaker understands the term she grasps the intension of the term.

But he argues that this is not true of natural kind terms—we do not under-

stand them by grasping such intensions. We understand them in the sameway we understand indexicals, where context plays an indispensable role.

Putnam puts this by saying that the psychological state of the speaker is not

the sole determinant of the reference of her terms—that is, internal psychol-

ogy does not determine a speaker’s reference. He thus rejects the old view to

the effect that a speaker’s reference can be extracted from what is in her mind

as she speaks. We shall now examine his arguments for this conclusion.

7.2 Twin Earth and “Water”

Putnam proceeds by constructing his Twin Earth thought experiment.

Imagine a time on Earth before chemistry fully developed but people still

used the word “water.” Because of the lack of development in chemistry,

people did not know the actual chemical composition of water, which is

H2O. The word “water” as it is used on Earth refers to water. Now imagine

an exact duplicate of Earth, Twin Earth, where there is no water. However,

there is a liquid on Twin Earth with many of the same apparent features

that water has even though it is not water. Putnam stipulates that it has a

chemical composition of XYZ. Of course, it is possible for liquids to have

the same appearance without having the same chemical composition. The

thought experiment is all perfectly metaphysically possible. Now suppose

that there are people on Twin Earth that are exactly like us—they are in fact

identical molecular duplicates of us, perfect twins. They speak a language

we would call English and one of the words they use is “water.” However,in Twin Earth English, “water” refers to the liquid on Twin Earth (XYZ), not

the liquid on Earth (H2O). The term has a different extension on the two

planets. However, since the time period we are considering is before the





136 Chapter 7

words. Putnam thinks that what is in your head does not determine your

meaning because it does not determine reference. The people on Earth and

Twin Earth have the same things going on in their heads, but they don’tmean the same thing by their term “water” because they don’t refer to the

same thing. The meaning of the word cannot be deduced from the speaker’s

psychological state. It depends on other extrinsic factors, and we will soon

see what those are. The speaker’s inner state of understanding does not

necessarily fix what he is referring to, so the meaning of his term cannot be

“read off” his state of understanding. Therefore, Putnam concludes, mean-

ing is not “in the head.” His point is that meaning is not a psychological

phenomenon.Let us restate Putnam’s argument with all the pieces now in place. The

essential point of the Twin Earth thought experiment is that we would

be right to say that “water” in Twin Earth English refers to XYZ and that

“water” in Earth English refers to H2O. Since the inhabitants of Twin Earth

are molecular duplicates of us, this point has serious philosophical conse-

quences for what constitutes meaning. As molecular duplicates, their brain

states are exactly the same as ours. Consequently, if we could gaze into the

minds of our molecular duplicates while they say the word “water,” we

would see exactly the same experiences, beliefs, emotions, and desires that

we would see if we gazed into our own minds when we said that word. So

we can see that on these two different planets the word “water” has a dif-

ferent reference and therefore a different meaning, despite the fact that the

speakers who use that word are in the same psychological state when they

use it. Since the same descriptions are associated with the word by the two

sets of speakers (“the colorless, tasteless liquid that flows in rivers” etc.),

the speakers are in the same psychological state, even though the word

“water” has different reference in the two cases. Since meaning determines

reference, as Putnam assumes, following Frege, the two words must have

different meanings. Thus “water” on Twin Earth does not have the same

meaning (or sense) as “water” on Earth. Nevertheless, the speakers are in

the same psychological state when using this word.

Another easy way to see how the argument works is to consider the case

of ordinary names. Take the name “Aristotle” and suppose that on TwinEarth there is no Aristotle, because it is too far away for Aristotle to have

ever visited Twin Earth. However, there is a person on Twin Earth who looks

and behaves exactly the same way as Aristotle, without being him. On Twin



Putnam on Semantic Externalism 137

Earth, when the speakers use the name “Aristotle” they refer to their Aris-

totle—not to our Aristotle. To avoid ambiguity and confusion, we can call

their Aristotle “Albert.” When they use the name “Aristotle” they refer toAlbert (as we call him), because the name “Albert” is our name for the man

they refer to with “Aristotle.” Putnam’s point holds here because the speak-

ers on Twin Earth are exact psychological and physical duplicates of us, but

they refer to a different person when they use the name “Aristotle” from

the person we refer to when we use that name. They refer to Albert (though

their name for him is “Aristotle”), whereas we refer to Aristotle. Since mean-

ing determines reference, it cannot be that the meaning of our word “Aris-

totle” is in our heads. The Twin Earthlings’ psychological state is exactlythe same as ours and yet they do not refer to Aristotle but to Albert with

“Aristotle.” There is a different reference yet the same internal psychology.

It is important to note that when the speakers either on Earth or Twin

Earth say the word “water,” there are no experts on these planets as to what

water is. We are assuming in this first example that the thought experiment

concerns a time before the rise of chemistry. No one on Twin Earth or Earth

knows the molecular composition of the liquid they refer to with the word

“water.” So the case is not like the contemporary world.

In addition to the example of the word “water,” Putnam also gives us

the case of molybdenum and aluminum. It is essentially the same situation

as with Twin Earth “water” except Putnam assumes that there are some

experts who can tell the difference between aluminum and molybdenum.

He supposes that there are some metallurgists who can determine this fairly

easily (on Twin Earth pots and pans are made of molybdenum and on Earth

they are made of aluminum, and the metallurgists can tell the difference by

a simple test). They look very similar and are used for the same purposes,

but a metallurgist would quickly be able to determine what type of metals

they were. There is nothing really new in this second example—it is exactly

the same as the first one, but Putnam just happens to bring in some experts.

In this case we also have duplicate speakers referring to different things

with the same term, so it is not a matter of what is going on inside you that

determines what you refer to but what kind of environment you are in.

The third example Putnam mentions is the use of the words “elm” and“beech” to refer to different species of trees. This example does add some-

thing to the original story because Twin Earth is not required to see this

point. It is a point about Hilary Putnam himself, stuck here on Earth. In his



138 Chapter 7

idiolect, when he uses the word “elm” he does not associate any descrip-

tions with that word that he does not associate with the term “beech,”

because he confesses that he cannot tell the difference between elms andbeeches. Since most of us are similarly (and shamefully) ignorant of the

differences between elms and beeches, we could not supply a descrip-

tion to distinguish one from the other. Nevertheless, the words “elm” and

“beech” mean something different in our idiolects as we use them—they

do not have the same reference or extension. Although there is nothing

in our minds that would allow us to make a distinction between the two,

one term refers to a tree that is an elm, and the other refers to a tree that

is a beech. This should remind us of Kripke’s example of Feynman andGellman (see chapter 2). Not familiar with the particulars of their work, a

speaker’s description of each of these physicists might be that they are both

famous twentieth-century physicists. Even though the speaker possesses no

descriptions to distinguish Feynman from Gellman, he still refers to a dif-

ferent person when he uses “Feynman” from when he uses “Gellman.”

We may wonder how we can use these words to refer to different natural

kinds of trees, even though what is in our head may be the same in respect

to these two words. The speaker means something different by “elm” and

“beech,” even if the stuff in his head is the same. This is a question about a

speaker’s idiolect in a specific linguistic community, as opposed to compar-

ing two linguistic communities (Twin Earth and Earth). Earlier in the book

(chapter 2) we talked about the division of linguistic labor in connection

with Kripke and names. That same division of linguistic labor, in which

experts determine what particular words refer to, is present here. When

we ignoramuses use “elm” or “beech,” we really intend that our reference

with those words should depend on our relation to the arboreal experts in

our midst. When we use the words, we intend to refer to what the experts

refer to when they use the words “elm” and “beech.” In this case, too, the

individual speaker’s meaning cannot be read off his psychological state,

but can only be gleaned from his context—specifically, the experts in his

linguistic community.

There are a few other examples that Putnam does not go into detail about

that are useful to our discussion. Toward the end of the article, he begins totalk about indexicality and that notion seems to play a central role in these

cases. Many of them directly involve indexicals. Imagine somebody point-

ing to an elephant. When the speaker says “that elephant,” imagine that his




brain is in a certain state and that he perceives the elephant in a certain way

(as big, gray, etc.). Now imagine on Twin Earth or some other place on Earth

that there is an exact twin of that person who says “that elephant,” pointingto a different elephant. He is a molecular twin of the first speaker, so every-

thing is internally the same in both of their minds. However, when the first

speaker says “that elephant” he refers to a different elephant from the one

his twin refers to. They refer to different animals even though they are in

exactly the same psychological state, because they are pointing at different

elephants. The context fixes the reference, not the perceptions and ideas in

their minds. They refer to what they see, and they see different elephants.

Another example is the word “I.” I say, “I am hungry”; now consideran exact duplicate of me who says, “I am hungry.” He does not refer to

me, he refers to him, but he is in exactly the same psychological state as

me because he is a molecular duplicate of me. By uttering the word “I” he

refers to an object a, whereas I refer to an object b, but we are both in the

same internal psychological state. Therefore, if meaning determines refer-

ence, our meanings are not in our heads—what we say cannot be extracted

from an examination of what is happening inside us. The context, that is,

who is actually uttering the word on that occasion, determines what we say.

Putnam’s recipe for producing these outside-the-head cases is very straight-

forward: we just vary the speaker’s environment while keeping his head the

same, and we find that the semantics varies. It is not difficult to generate

similar cases for “now” and “here.” The trick of the examples is simply that

context can vary while internal states stay constant.

Let us make something explicit. Toward the end of the article, Putnam

hints at this point, but it has a much greater significance than he recog-

nizes. He is really arguing for a disjunction: either meaning is not in the

head or meaning does not determine reference. His thought experiments

are neutral between these two propositions—we can interpret them either

way. However, Putnam assumes that meaning determines reference, and

hence concludes that meaning is not in the head. If meaning determines

reference, then meanings are not in the head. But what if meaning does not

determine reference? Then meaning can stay in the head, while failing to

determine reference. He has shown that meaning does not determine refer-ence, on this alternative interpretation.

We could therefore accept Putnam’s Twin Earth cases but question

whether they prove that meaning is not in the head and hence not



140 Chapter 7

determined by psychological state. Could it be that meaning is in the head

and hence is determined by psychological state, but meaning does not

determine reference? There are thus two theoretical possibilities: (1) mean-ings are not in the head and hence are independent of psychological state,

or (2) meanings are in the head and hence are dependent on psychological

state—but meaning is not sufficient to fix reference. Why does Putnam

choose one of these interpretations over the other?

We could interpret the Twin Earth case as showing that the people on

Twin Earth mean the same thing by the word “water” as we mean by the

word “water” but that their reference for that word is different from our ref-

erence for that word. What they mean is what is in their heads, the descrip-tions they would give. But what they mean does not uniquely determine

what they refer to. It is only on the assumption that intension determines

extension—that sense determines reference—that Putnam’s cases establish

that meaning is not in the head.

To better illustrate this point, let us return to the indexical examples.

In the elephant case, when each speaker says “that elephant” he refers to

a different animal when pointing to each elephant. It is indisputable that

they refer to something different, but it does not follow that they mean

something different. It largely depends on our definition of “meaning.”

There is a lot of complexity in the notion of meaning, particularly in the

case of indexicals. We learned in earlier chapters that we need at least a two-

dimensional theory of indexical meaning. Using Kaplan’s idea of character

as the meaning of “meaning,” the words “that elephant” have the same

character and hence the same linguistic meaning for the first speaker as for

the second. That character, however, does not determine reference. What

determines reference is character plus context, not character by itself. So

meaning, construed as character, does not determine reference. For this rea-

son, it would be a strange interpretation to say that this example shows that

meaning is not in the head—instead, it shows that meaning (character)

does not determine reference. As Kaplan would say, it shows that character

does not determine content. We will come back to this point later, but first

we must cover Putnam’s view of what his examples show.

One thing he concludes is expressed in the following passage:

HYPOTHESIS OF THE UNIVERSALITY OF THE DIVISION OF LINGUISTIC LABOR:

Every linguistic community exemplifies the sort of division of linguistic labor just

described; that is, it possesses at least some terms whose associated “criteria” are





142 Chapter 7

psychological states. Speakers are embedded in a world and they interact

with various things: it is the fact that they have those interactions that

determines what their words mean. What words mean is not just a functionof what is in someone’s head, individually or socially. Rather, meaning is

a function of the speaker’s actual external environment. The environment

itself determines what words mean. So, Putnam concludes, meaning is not

in the head—it emerges from interactions with the environment. This doc-

trine came to be known as semantic externalism, because it says that mean-

ings are externally determined.

As noted earlier, Putnam thinks the case of natural kind terms is similar

to the case of indexicals. In the case of indexicals, we can clearly see thatthe reference depends on the speaker’s way of being embedded in his or her

environment. We can see the operation of context. What determines the

thing I refer to when I say “that woman,” pointing to a particular woman

in front of me? Not what is lurking in my head, but the fact that a certain

woman is in my environment right now in front of me and I am pointing

straight at her. In the case of indexicals, it is very clear that reference is fixed

by the speaker’s location in the world. Here externalism seems obvious,

because indexicals are so clearly context dependent.

Putnam now makes a direct link between indexicals and natural kind

terms like “water.” He suggests that there is an indexical element in natural

kind terms. We can explain the reference of our word “water” by using a

demonstrative—as in “‘Water’ refers to that liquid,” said while pointing at

H2O—and that is how we tie down the reference of the word. As we have

discussed earlier, indexicals play a crucial role in determining the reference

of words that are not themselves indexical (e.g., proper names and definite

descriptions like “the father of that baby”). On Earth when we say “water”

the reference is determined by the indexical “that liquid.” On Twin Earth

when they say “water,” the reference is also determined by “that liquid,”

but here the demonstrative picks out a different natural kind. Hence the

word “water” has different reference on the two planets. Given this refer-

ential link between indexicals and natural kind terms, we would expect to

find natural kind terms functioning in the same way indexicals do. The

meaning of indexicals is not in the head, nor is the meaning of the naturalkind terms that are linked to them. Externalism thus holds for terms like

“water” because they have an indexical component.




7.4 Criticisms of Putnam

What is the best way to characterize the upshot of Putnam’s examples?What do they show about meaning? Putnam says they show that meaning

is not in the head, but we can (as noted earlier) equally conclude that they

show that meaning does not determine reference. Which characterization

is better? If we start with an indexical case like “I,” then on any reasonable

notion of meaning, the word “I” has the same meaning for whoever uses

it. The reference is not the same, to be sure, but the meaning clearly is. A

speaker refers to a certain individual when he uses the word “I” on an occa-

sion, but that fact is not reflected in what the word means, because the refer-ence depends on the meaning plus the context (character plus context). It

thus makes perfect sense to say that the meaning (character) of “I” is in the

head, because what is present in the speaker’s mind determines what char-

acter the indexical has. However, the conventional meaning of “I” is not

enough to determine its reference on an occasion. If we get fixated on the

description case, we will think that meaning has to determine reference,

because meaning does determine reference for definite descriptions. But this

is not so for indexicals. Indexicals require a more complex semantics, as

Kaplan demonstrates, in which we must distinguish different dimensions

of semantic significance. To say simply “Meaning is not in the head” is

therefore incomplete and ambiguous. Do we mean meaning as character or

as content—as conventional linguistic meaning or propositional content?

Nothing in Putnam has shown us that character is not in the head, so one

sort of meaning is in the head; all we have shown is that propositional con-

tent is not in the head. Given Putnam’s own indexical interpretation of his

cases, he should have concluded that part of meaning (character) is in the

head, though part is not in the head (content).

A further question concerns Putnam’s notion of a psychological state. He

simply assumes from the outset that psychological states are in the head.

This enables him to conclude that meaning is not psychological, because

the former is not in the head while the latter is. He thus takes it for granted

that the psychological state of the molecular duplicates on Twin Earth is the

same as the psychological state of people on Earth. He takes it that peoplecannot have different psychological states if they are physically identical.

But is that so obvious? Some have questioned this assumption of Putnam’s,

wondering whether we should conclude instead that psychological states



144 Chapter 7

are not in the head either. Let us ask ourselves what people believe on Earth

and Twin Earth. What do I believe when I say “This water is warm”? Obvi-

ously I believe that this water warm. On Twin Earth my molecular duplicatewould also say “This water is warm,” referring to some XYZ. Does he believe

that this water is warm? Well, he clearly does not believe that this water is

warm, because this water is here on Earth, not there on Twin Earth. But

does he have any belief involving the concept of water at all? No, he does

not: he has no beliefs about water at all. He has beliefs about another liquid,

not water. Let’s call his XYZ liquid “retaw”; then we can say he has beliefs

about retaw. What he believes is that some retaw is warm. His belief is about

something different from what my belief is about. He has the concept retaw whereas I have the concept water . Clearly he perceives something different

from what I perceive, because I am in the perceptual state of seeing water

and he is not in that perceptual state. He is never in that perceptual state,

because he never sees any water; he sees only retaw. We cannot report his

perceptual state with the words “He sees that water is in the well.”

The psychological state of seeing water is not a psychological state any-

one on Twin Earth is ever in. Neither does anyone there have the concept

water , nor the belief that there is water about. The psychological states

associated with the word “water” on Twin Earth are therefore not the

same as the psychological states we have on Earth. They have different

psychological states from us. To be sure, they share some psychological

states with us, namely the descriptive beliefs they apply to the liquid on

their planet. But it does not follow that they share all their psychologi-

cal states with us, and it seems plainly false to use our word “water” to

describe their psychological states. Have you ever had any beliefs involv-

ing the concept retaw before hearing about Twin Earth? Hardly; all your

beliefs involve the concept water . They no more think about the natural

kind water than they think about the particular pool of water I referred to

here on Earth with “this water.” So there are psychological states correlated

with the use of “water” on Earth and Twin Earth that differ in their con-

tent, even though the speakers are molecular duplicates. Therefore, these

psychological states are not in the head. When Putnam says that meanings

are not in the head, he should have added that psychological states are notin the head either—and for essentially the same reason. The content of

psychological states is also fixed by the person’s actual environment. That

is, the full propositional content of psychological states is partly fixed by




interactions with the environment. Thus we have externalism about mind

as well as meaning.

But this changes the whole picture. If the psychological states of Earth-lings and Twin Earthlings are different, then those psychological states do

determine the meaning of the terms used, even when this meaning is taken

to incorporate something like Kaplanian content. The psychological state

of my counterpart involves the concept retaw , while the psychological state

that I am in involves the concept water . Those two concepts are not deter-

mined purely by our internal states but by our embedding in the world.

These externally determined psychological states accordingly do determine

what we mean by the term “water.” Thus there is no divorce of semanticsfrom psychology, only a divorce of psychology from neurophysiology. Nei-

ther meaning nor mind is reducible to internal neurophysiology.

To return to the indexical case involving the elephant: one speaker says

“That elephant is big” while pointing to elephant A, while the other speaker

points to elephant B and says “That elephant is big.” The first speaker

believes that A is big, while the second believes that B is big. A and B could

be animals on different continents. Each speaker only has beliefs about the

elephant in front of him—to the effect that that elephant is big. The content

of the belief a person has when he uses an indexical term like this is deter-

mined by his environment, so his beliefs are not in his head. This is just to

apply the lessons of direct reference to beliefs as well as meanings. Belief

and meaning march in parallel, as we would expect.

Thus, psychological states are not in the head and meanings are not in

the head. Or better, an aspect of both is not in the head—because there is

also an aspect that is in the head (the aspect corresponding to character).

If psychological states are not in the head, they can determine meaning,

even assuming that meaning determines reference. My psychological state

could determine the reference of my terms even if we accept Twin Earth

cases, because the psychological states of the people on the two planets dif-

fer, despite their molecular identity. The psychological state mirrors what is

in the person’s environment too. As soon as we realize that psychological

states are not in the head, we see that Putnam is misstating his conclusion.

He is right to say that what is internal to us cannot determine our reference,but that does not entail that our psychology does not determine our refer-

ence. Rather, our psychology is not (purely) internal. We need to accept

psychological externalism too.



146 Chapter 7

In sum: Putnam is wrong to claim that meaning is totally outside the

head, because of the existence of an internal component to meaning,

namely character; and he is wrong to claim that meaning is not fixed bypsychological state, because his own arguments imply that psychological

states are as externally determined as meaning. What he is right about is

that external context plays a vital role in fixing reference. This may not

sound like the resounding and revolutionary conclusion he originally

announced, especially once we have properly investigated the semantics of

indexicals, but it does contain an important truth.



8 Tarski’s Theory of Truth

8.1 Background

We have invoked the concept of truth at several points, but we have said

nothing much about how this concept is to be understood. What is truth?

The theory of truth we are about to examine was originally proposed in 1933

in a very long and difficult article called “The Concept of Truth in Formal-

ized Languages” by Alfred Tarski, a Polish mathematical logician. The article

we will examine, however, is “The Semantic Conception of Truth,” pub-

lished in 1944, and meant as an easier presentation of the same ideas as the

original, more daunting article. This one is quite daunting enough. As Tarski

remarks at the beginning, it is a return to the subject matter of his earlier

article, which is really a treatise in formal logic. The original article is tough

reading if one does not have a solid grounding in mathematical logic. It was

a very important contribution to pure logic. It has also been philosophi-

cally important. Historically, people thought it was a great breakthrough in

the philosophical theory of truth. It finally made the study of truth rigor-ous and subject to logical treatment. It made philosophy into mathematics!

Many philosophers felt that we no longer needed to have any qualms about

employing the notion of truth, because Tarski had given us such a tight,

precise definition. Donald Davidson later took up Tarski’s theory to propose

a theory of meaning for natural languages, as we will see in the next chapter.

Tarski had tamed truth, rendered it “scientific”—quite a feat. The adjective

“Tarskian” took on the canonical stature of the adjective “Fregean”—as in

“the Tarskian theory of truth” and “the Fregean theory of meaning.”

Still, there is controversy about what Tarski’s theory really accomplishes,

both as a theory of truth and a theory of meaning. But before we get to

that, we need a good understanding of what the theory actually says, and



148 Chapter 8

for that we cannot do better than to attend to Tarski’s own words. This will

be our procedure in what follows.

Let us first talk a bit about the background to Tarski’s own proposal. Inthe history of philosophy, a number of different theories have been pro-

posed about truth: the coherence theory of truth, the correspondence theory

of truth, and the pragmatic theory of truth. The coherence theory states that

a proposition is true if and only if that proposition coheres with the other

propositions that one believes. By the standards of the coherence theory, a

belief is true if and only if that belief is consistent with one’s other beliefs.

Truth is then a matter of the logical relations between believed propositions.

The correspondence theory states that a belief is true if and only if thatbelief corresponds to the facts. A formulation Tarski uses for the correspon-

dence theory is that a proposition is true if it designates a particular state of

affairs: that is, it refers to the actual state of reality. It is called the correspon-

dence theory because it talks about the relationship between a proposition

and something in the world outside of the proposition—facts or states of

affairs or something of the sort. There are these things out there in the

world and a true proposition is one that corresponds to them. The notion

here is not coherence among beliefs but correspondence to something that

lies outside of them.

The third theory is usually associated with American pragmatism: hence

the pragmatic theory of truth. It is that a proposition is true if and only it is

useful to believe that proposition. That is, a proposition is true if and only

if one’s plans and projects go better by believing it than by not believing it.

Truth is utility. A true belief increases utility and a false belief decreases util-

ity. For example, if I hold the false belief that I can jump off a tall building

and fly away into the sky, that will likely result in a decrease of utility as I

plummet to the ground. Therefore, true beliefs are the ones that maximize

well-being.

Let’s quickly go over the standard objections to each of these theories.

The problem with the coherence theory is that a belief could be consistent

with my other beliefs and yet the whole lot could be false. Consistency

alone will never make a belief true, because false propositions can be mutu-

ally consistent (the belief that the Earth is flat is consistent with the beliefthat you will drop off the edge if you travel far enough, but neither belief

is true). Coherence is just about the relationship between one belief and

another, not about whether any of them fit objective reality. A person could



Tarski’s Theory of Truth 149

have a perfectly coherent belief set and yet all her beliefs are false. To get

truth we need to bring in something external to belief.

The pragmatic theory of truth has a very similar problem. I could havea belief about something that is useful to me but that belief is false. We

can imagine a case where someone lives in a society where having certain

beliefs is rewarded and having other beliefs is punished. In Communist

Russia, for instance, if you had the belief that the bourgeoisie were wicked,

that belief was likely to be rewarded. If you had the belief that the bourgeoi-

sie are meritorious, you would hold a belief that is likely to be punished.

It is more beneficial to hold the former belief than the latter, but it doesn’t

follow that the former belief is true and the latter false. Therefore utilitydoes not always coincide with truth. At best the two properties are gener-

ally correlated.

Most philosophers think that the correspondence theory is the best the-

ory. It captures the idea that truth depends on objective reality—not on

us. The problem the correspondence theory has concerns more technical

issues such as what a fact is and what the correspondence relation amounts

to. Are facts complexes of objects and properties? How do we count them?

How exactly do they differ from true propositions? Are there general facts

and negative facts? It is also difficult to find a clear and correct formulation

of the underlying notion of correspondence to reality. Is it a kind of denot-

ing, or some sort of isomorphism? To clarify the correspondence theory is

largely the task Tarski sets himself, so let us proceed directly to that.

8.2 Tarski’s Criteria of Acceptability

Tarski’s theory is supposed to sweep away all of this confusion and obscu-

rity about truth and replace it with a logically hygienic theory with none

of the above-mentioned problems. It is meant to be a nice, clean logical

definition about truth—which is why everyone liked it (almost everyone).

He says at the very beginning of the article that to arrive at a satisfactory

definition of truth we first need to know what the definition is meant to

achieve—only then can we properly judge the definition. He then immedi-

ately dives into his approach to defining truth. We need to determine whatwe want the theory to do and what conditions make it acceptable.

Here he distinguishes two tests for whether the theory of truth is accept-

able. He calls them material adequacy and formal correctness. A good theory



150 Chapter 8

of truth must be materially adequate and formally correct. To start with,

material adequacy means simply that the definition should (in Tarski’s own

words) “catch hold of the actual meaning” of the word “true.” In otherwords, it should not be a theory that stipulates a new meaning for the word

“true,” or seeks to reform its meaning; rather, the definition should actually

capture what the word “true” means when we all use that word. You might

think that this is a trivial requirement, because surely if we are trying to

define a word of ordinary language we should try to capture what it actu-

ally means. And you would be right: if we are trying to define “know,” for

example, we also want to catch hold of the actual meaning of that word.

Doesn’t every philosopher interested in defining a word want his definitionto be “materially adequate,” that is, correspond to what the word actually

means? Sometimes people think that there is a mysterious technical aura

surrounding Tarski’s concept of material adequacy, but really he just means

capturing the concept of truth that we actually have. Later we will see that

he has a more technical formulation of material adequacy, but to begin

with he just means that the definition should be accurate.

By “formally correct” Tarski means that there should be no logical errors

in the definition and that we must specify the formal structure of the lan-

guage we are using. For example, the definition must not commit use–

mention confusions. The theory must be formulated in such a way that

it is not guilty of any logical infelicities or lack of clarity. Again, this is a

familiar requirement that we should place on any philosophical definition

of any concept. No definition can be permitted to be formally incorrect!

In the case of truth, Tarski is particularly concerned with the paradoxes

that can arise with the word “true” (as with the paradox of the Liar who

announces “Nothing I say is true”), so he is especially concerned to avoid

logical pitfalls.

The next point he brings up concerns the application of the word “true.”

We have the predicate “is true” and from the point of view of grammati-

cal form it looks exactly the same as a predicate like “is red.” The predicate

“is red” ascribes the property of being red to an object. Similarly “is true”

appears to ascribe a property to the thing being referred to. Thus truth is

a property expressed by a predicate just as redness is a property expressedby a predicate. But what is it a property of? As Tarski says, it can be applied

to different things, three of which he mentions. First, it can be applied to

beliefs, which are psychological states: we can say that our beliefs are true




(or false). It can also apply to propositions, which are the abstract contents

of beliefs. For example, we can say that the proposition that snow is white

is true; here we say nothing about anybody’s beliefs. If we apply “true” toa proposition we apply it to something that is not dependent on any par-

ticular language or on any believer. The same proposition can be expressed

by different sentences in different languages—the sentences that are syn-

onyms or exact translations. A proposition is a kind of abstract entity and

we can ascribe truth to that. But we can also ascribe truth, Tarski says, to

sentences, which are concrete linguistic entities. We can say that the sen-

tence “Snow is white” is true, where the sentence is conceived as a series of

marks or sounds, that is, a perceptible physical entity.The previous sentence contains a reference to a sentence, unlike the one

before it. Using quotation marks, we refer to the English sentence “Snow is

white.” When we apply the predicate “is true” to a sentence we have to put

that sentence in quotation marks. We thereby create a name of a sentence,

to which we attach the predicate “is true.” Tarski does a lot of naming of

sentences in his theory. The thing about sentences is that unlike proposi-

tions they are dependent on language—they are not common between lan-

guages like propositions. It therefore changes the logic of the word “true”

when you apply it to sentences instead of propositions. Now you are apply-

ing it to the tangible vehicle of propositions, not the shadowy propositions

themselves. We can also apply “true” to speech acts performed by uttering

sentences, as with statements and assertions. All these things can be said

to be true or false, despite their variety. Tarski announces that he will take

“true” to apply to sentences, so that he is defining truth as it applies to

sentences. Thus the extension of the predicate “true” will be the class of

true sentences. This affects the form of his definition, as we shall see, par-

ticularly with regard to the use of quotation.

8.3 Aristotle and the Redundancy Theory

Tarski explains how he came up with the inspiration for his own theory by

tracing it back to Aristotle:

We should like our definition to do justice to the intuitions which adhere to the clas-

sical Aristotelian conception of truth—intuitions which find their expression in the well-

known words of Aristotle’s Metaphysics: To say of what is that it is not, or of what is not

that it is, is false, while to say of what is that it is, or of what is not that it is not, is true.1



152 Chapter 8

For the sake of simplification, we can cut out the negation part of Aristotle’s

formulation and still express the essence of Tarski’s view. Truth is saying

of what is so that it is so—this is Aristotle’s fundamental idea. If it is sothat this table is brown, then it is true to say that the table is brown. That

sounds right and it is the basis of what is now called the redundancy theory

of truth. To say that a sentence is true is just to say that things are as the

sentence says—and that’s it. We could equally have repeated the sentence.

Tarski himself never mentions this type of theory by name even though

the theory that he proposes is clearly a version of the redundancy theory.

Suppose a speaker says, “Snow is white,” and his audience replies, “Yes, that

is true.’” What does the second speaker mean when he says this? He couldhave also said, “Yes, snow is white.” It is bit more long-winded to say, “Yes,

snow is white,” because then he would have to repeat exactly what the

speaker says. It is much easier to say, “That’s true.” By saying “That’s true”

he can reassert everything that the first speaker said in a shorthand form.

Thus we can abbreviate our agreement with what someone said by using

the simple predicate “is true.” We don’t need to go to the trouble of assert-

ing the whole thing over again. This piece of linguistic machinery prevents

us from needing to repeat everything that someone else says. It can be very

useful in making a statement like “Einstein’s theory of relativity is true”—

sparing us the need to recite the whole theory of relativity! Tarski takes it

that sentences containing “true” are equivalent to the sentences to which it

is applied. The word adds nothing to the content of the sentences to which

it applies. The idea is that the word “true” is strictly speaking redundant.

We have it in our language, and we use it for practical purposes, but we

could in principle do without it.

And so we come to Tarski’s celebrated biconditional:

(1) “Snow is white” is true if and only if snow is white.

The predicate “is true” is strictly speaking redundant because the result of

applying it to a sentence produces something equivalent to that sentence

itself. We could say “The sentence ‘Snow is white’ is true” or simply “Snow

is white.” Either way we have said the same thing. The sentence “The sen-

tence ‘Snow is white’ is true” means the same thing as the sentence “Snow

is white.”

That is the insight behind the redundancy theory—which is sometimes

also called the disappearance theory or the disquotational theory . It is as if the




predicate “is true” strips a sentence of the quotation marks that enclose it,

and then disappears into the night. We disquote the sentence and write it

again after “if and only if,” thereby obtaining the definition of “true” as itapplies to “Snow is white.” But before getting to the Tarskian technicalities

involving these disquotational biconditionals, let us talk a bit more about

the Aristotelian view of truth, as Tarski understands it. Actually, the view

is very often attributed to Frege, on the strength of this passage from “On

Sense and Reference”:

“The thought that 5 is a prime number is true” contains only a thought, and indeed

the same thought as the simple “5 is a prime number.” It follows that the relation

of the thought to the True may not be compared with that of subject to predicate. 2

Frege is claiming that a sentence of the form “S is true” expresses the same

thought as S. Of course, saying they express the same thought is another

way of saying they are synonymous. Thus the sense of the sentence “The

sentence ‘Snow is white’ is true” is identical to the sense of the sentence

“Snow is white,” because they express exactly the same thought and are

exact synonyms of one another. Tarski’s truth biconditionals are just a regi-

mented expression of this Fregean insight.

By contrast, the correspondence theory tells us that “Snow is white” is

true if and only if it corresponds to the fact that snow is white. Here we

invoke, in addition to snow and whiteness, entities called “facts” and a

relation called “correspondence.” These raise tricky philosophical and logi-

cal questions. With Tarski’s theory we do not have to bother with such

questions. There is no need to bring in the concepts of correspondence

and facts—we just repeat, “Snow is white” after “if and only if.” And snow

being white is not philosophically problematic, because we know what thatis—there is no particular philosophical problem about snow being white!

This is a very simple streamlined account of what truth is, using no murky

notions. We have boiled truth down to the basics. The only real question is

a technical one about how we would apply this definition to the full range

of sentences. There is in the end no more to the concept of truth than ordi-

nary sentences and what they are ordinarily about.

The beauty of this theory lies in its triviality. It does not involve us in

any complicated conceptual analysis or controversial notions. Tarski actu-

ally rather misstates this aspect of his theory. He seems to think that his

theory is a form of the correspondence theory. He explains himself in the

following passage:



154 Chapter 8

If we wished to adapt ourselves to modern philosophical terminology, we could

perhaps express this conception [Aristotle’s] by means of the familiar formula: The

truth of a sentence consists in its agreement with (or correspondence to) reality.3

Most philosophers want firmly to distinguish the Aristotle/Frege concep-

tion of truth from the correspondence theory just stated. The view he here

describes is aptly called a correspondence theory, because it talks about a

relation of “agreement” between sentences and something called “reality,”

but his own theory makes no use of such notions. The idea is to avoid all of

that by adopting a redundancy view of truth. Tarski seems to be confusing

the classic correspondence theory with the redundancy theory. The latter

theory treats “true” as an essentially redundant device, but the former takestruth to be a substantial relation of correspondence between sentences, on

the one hand, and facts/existing states of affairs/reality, on the other. Tar-

ski’s actual theory has a very different form, as we shall see.

To get started on the details of Tarski’s theory, we must first analyze the

basic logical form of his truth biconditionals. Their abstract logical form is

the following:

x is true if and only if p.

The letter “x” is typically reserved in logic for individual variables. Indi-

vidual variables are what stand in place of names or descriptions or pro-

nouns. The letter “x” is thus a variable that stands in place of a singular

term. Of course, the singular term is a part of the sentence and not the

whole sentence. Looking at only the left-hand side of the biconditional, for

example “The sentence ‘Snow is white’ is true,” we see that it has the form

“x is T.” The part where we quote the sentence is a singular term and hence

can be replaced by a variable. If we wanted we could give that sentence a

name—say, “Burt.” We could stipulate, “Burt is the English sentence ‘Snow

is white.’” Then we could formulate the biconditional as “Burt is true if and

only if snow is white.” Logically, quotation converts a sentence into a sin-

gular term designating itself. Therefore, the logical form of “‘Snow is white’

is true” is “x is T.” In standard logical notation that would be “ Fa,” where

“a” is a name and “ F ’ is a predicate (as in “John is bald”). In other words, it

is a subject-predicate sentence.

However, the sentence on the other side of “if and only if” does not con-

tain a singular term for a sentence—it is just a sentence in use, referring to

snow and whiteness. For this reason, in logic the variables conventionally





156 Chapter 8

language and the metalanguage. The object language is the language we

are speaking about when we formulate our definition of truth for that lan-

guage. So far, the object language has been English, because the sentence“Snow is white” is an English sentence. But it could be French or Italian

or Chinese. It is just whatever language we are talking about, to whose

sentences “true” can apply. We typically refer to sentences of the object

language by using quotation marks, though that is not the only way.

The metalanguage, on the other hand, is the language that we are using

to talk about another language. So far, that has also been English, but it

could be any language. A Frenchman interested in defining truth for English

will take English as his object language but use French as his metalanguage.The distinction is simply between the language that we talk about and the

language that we use to talk about that language. So far, our metalanguage

and our object language have been the same language, namely English.

But that is not always the case. We could have an object language that is

French and a metalanguage that is English. For example, we could say “‘La

neige est blanche’ is true if and only if snow is white.” We could equally talk

about a Martian language in Swahili when formulating our Tarskian truth

theory for Martian. This terminology helps us keep straight what language

we are talking about (note that we can also talk about the metalanguage,

though we are now using a meta-metalanguage). Just because we use Eng-

lish as both our object language and our metalanguage does not mean that

we can ignore the distinction.

Most philosophers call the Tarskian biconditionals “T-sentences.” Adopt-

ing this terminology, we can say that a T-sentence is a sentence of the meta-

language that mentions (on the left-side) a sentence of the object language.

Thus we use the metalanguage to mention the object language when we

write down a T-sentence. One other point Tarski makes in this connection

is that since we apply the word “true” to sentences, and not to propositions

or statements or beliefs, we have to relativize the truth predicate. The sen-

tence “Snow is white” can in principle be true in one language but not true

in another language, because that string of marks or sounds could mean

different things in different languages. In English, the sentence “Snow is

white” means that snow is white, and snow is white, so that sentence istrue in English. But suppose some other language contains exactly the same

sentence either acoustically or graphically, yet with a different meaning,

say, that snow is black. Then in that language the sentence “Snow is white”




means that snow is black, but snow is not black, so that sentence is false in

that language. Strictly, then, we need to write our T-sentences as follows:

“x is T in L if and only if p.” Now we are logically pucker! The T-sentencefor the second language (call it Twenglish) will then read: “‘Snow is white’ is

true in Twenglish if and only if snow is black.”

We don’t have to relativize truth when applying it to statements or

beliefs or propositions, because they are not language dependent. The

proposition that snow is white is true if and only if snow is white, period.

Here meaning is built in. A proposition does not vary in meaning between

languages, because it is not a piece of language (similarly for statements

and beliefs—the proposition is built in). But if we are defining “true” as itapplies to sentences, construed as marks or sounds, then we need to rela-

tivize the truth predicate, because of potential variations of meaning from

language to language. This is simply because sentences in themselves are

just meaningless scribbles or noises.

8.5 How to Derive the T-Sentences

What we have on the table so far are two things: a philosophical motiva-

tion, deriving from Aristotle and Frege, for focusing on the T-sentences;

and some clarification of the logical status of the T-sentences and how they

should be analyzed. But we don’t yet have a theory of truth. Here in a nut-

shell is what Tarski proposes: a definition of the word “true” for a language

is materially adequate and formally correct if it entails all the T-sentences

for that language. In other words, take all the (indicative) sentences of Eng-

lish and write out a T-sentence for each of these sentences. We then have all

of the T-sentences corresponding to all of the sentences in English. A satis-

factory definition of truth, Tarski proposes, is a theory that entails all these

T-sentences. Here is where he introduces the idea of a “partial definition.”

What he is saying is that a T-sentence for (say) “Snow is white” defines the

word “true” partially with respect to that sentence. We have given a partial

definition of the word “true” for the sentence “Snow is white.” If we now

take the sentence “Grass is green” and write out its T-sentence, then we

have partially defined “true” for that sentence. And so on. Each of these isa partial definition, and the totality of them is a complete definition of the

word “true” for English. If we had such a complete compilation, we would

have shown what it means for every sentence in English to be true. That is



158 Chapter 8

the ultimate goal of Tarski’s theory. The correct and complete definition of

the word “true” is something that would entail all those partial definitions.

We just need to add them all up to get what we seek.A logically adept student might jump up at this point and say that there

is an easy way to get the desired result. We could simply form the logical

conjunction of all of these T-sentences. We take all of the individual T-sen-

tences for English and join them together with “and” (“‘Snow is white’ is

true if and only if snow is white and ‘Grass is green’ is true if and only if

grass is green and …”). The conjunction of sentences entails each conjoined

sentence, because in elementary logic “ p and q” entails “ p” (and also entails

“q”). If we had a conjunction of the full range of T-sentences, that conjunc-tion would entail each of those T-sentences. The conjunction would thus

entail all the partial definitions, so it would be the complete definition. So

start conjoining! The conjunction of all the T-sentences would meet all of

Tarski’s requirements, as so far stated.

That would be a perfectly adequate definition of truth by Tarski’s stan-

dards, except for one small point. There are infinitely many sentences in

English. We can generate an infinite number of sentences in a natural lan-

guage like English, because these languages contain certain devices that

enable the speaker to form ever more complex sentences. The most obvious

one is “and.” Whenever we have a sentence we can always add another sen-

tence by conjoining it to the previous sentence. If we start with a conjunc-

tion, no matter how long the conjunction is, we can always create another

sentence by conjoining something else to it. It is the same with negation.

We can negate “ p” to get “not- p,” and then we can negate that again to

get “not-not- p,” and so on. The rules of the English language allow us to

negate as many times as we like and therefore produce as many sentences

as we wish. Therefore, a conjunction of all the English sentences would be

an infinite conjunction—and so in consequence would be the conjunc-

tion of all the T-sentences. To use a more precise logical terminology, the

resulting theory of truth would not be finitely axiomatized, which means

it could not be written down (or even formulated in thought). It would

clearly be much better to have a finitely axiomatized theory that entails all

the T-sentences. Then we could at least have a look at it!It will turn out that such a theory must analyze each sentence into

its constituent parts, and that is why it is of interest to those engaged in

semantic theory (see the next chapter). The way Tarski’s theory actually




works is that we do not take each sentence as primitive; instead we provide

a structural analysis of each sentence, and based on that analysis we gener-

ate a T-sentence for every sentence. So we don’t have to form an infiniteconjunction of all the T-sentences, even though that would have satisfied

Tarski’s condition of material adequacy. Strictly speaking, we should amend

Tarski’s condition to read: the theory must entail all the T-sentences from a

finite number of axioms.

How do we produce something that generates all the infinitely many

T-sentences without conjoining them into an infinite conjunction? Tarski

proposes that what we want is something that is in effect a logical conjunc-

tion of all the T-sentences. He makes this point in the following paragraph:

Now at last we are able to put into a precise form the conditions under which we will

consider the usage and the definition of the term “true” as adequate from the mate-

rial point of view: we wish to use the term “true” in such a way that all equivalences

of the form (T) can be asserted, and we shall call a definition of truth “adequate” if all

these equivalences follow from it. … The general definition has to be, in a certain sense,

a logical conjunction of all these partial definitions.4

“In a certain sense” it has to be the logical conjunction of the partial defi-

nitions, but not in the straightforward sense of simply conjoining them as

they stand. What Tarski has is a technical way of constructing something

that is in effect a logical conjunction without being an actual logical con-

junction. Soon we will see what that way is.

8.6 Satisfaction

Tarski next makes a few points about semantic notions and formal lan-

guages. Semantic notions he defines as relational, with the two most impor-

tant semantic notions being designation and satisfaction. I doubt the Rolling

Stones had anything Tarskian in mind when they wrote their song “(I Can’t

Get No) Satisfaction,” but the lyrics fit quite nicely. It is indeed not easy to

get no satisfaction. As Tarski shows, you need to be quite ingenious to get

satisfaction—obstacles must be overcome. These two semantic notions are

relational because they connect language to things in the world (I suspect

the Stones were singing of relational connections too). An example wouldbe the name “Mick Jagger” designating the writhing entity that is Sir Mick

Jagger. Satisfaction is very similar, but satisfaction is a semantic relation

that applies to predicates and not singular terms. Satisfaction is a relation







162 Chapter 8

left we mention an expression and on the right we use that same expres-

sion (if the metalanguage is the same as the object language). We could

call this an S-sentence, by way of analogy with a T-sentence. An S-sentencetells us under what conditions a given predicate is satisfied by an object.

We could also say that each S-sentence is a partial definition of satisfaction

for the language in question. All the S-sentences together give a complete

definition of satisfaction for the language. There are only finitely many

basic S-sentences, because there are only finitely many primitive predicates

in the language (three, to be precise). These are sometimes called satisfac-

tion axioms. (We can also write out “designation axioms” for the individual

constants, which will have the form: “‘a’ designates a.”)We have taken something that is a part of a sentence, the predicate, and

then defined the semantic relation of satisfaction for that part, which is

analogous to the way we would define truth for a whole sentence. We are

left with the following formula for “white”: “x satisfies the predicate ‘white’

if and only if x is white.” We define satisfaction for each of the predicates

by using in the metalanguage the expression we refer to in the object lan-

guage. But from this finite formulation, we can generate infinitely many

S-sentences. This is because we can use devices like “not” and “and” to pro-

duce arbitrarily complex predicates, such as “x is white and x is cold and x

is not ice cream.” This operation is called a recursive procedure, which Tarski

explains thus:

In defining the notion of a sentential function in formalized languages, we usually

apply what is called a “recursive procedure”; i.e., we first describe sentential func-

tions of the simplest structure (which ordinarily presents no difficulty), and then we

indicate the operations by means of which compound functions can be constructed

from simpler ones. Such an operation may consist, for instance, in forming the logi-

cal disjunction or conjunction of two given functions, i.e., by combining them by

the word “or” or “and.” A sentence can now be defined simply as a sentential func-

tion which contains no free variables.6

He is making the point here that we must remember that in addition to

primitive predicates there are also complex predicates built up by using

connectives. Consider the complex predicate “is white or red.” An object

satisfies “is white or red” if and only if that object satisfies “white” or it

satisfies “red.” We can then generalize this over all predicates to get a gen-

eral rule for “or”: for any predicates “ F ” and “G,” x satisfies “ F or G” if and

only if x satisfies “ F ” or x satisfies “G.” Now we have covered every possible




disjunction of predicates with this one axiom. Here is how Tarski explains

the idea:

To obtain a definition of satisfaction we have rather to apply again a recursive proce-dure. We indicate which objects satisfy the simplest sentential functions; and then

we state the conditions under which given objects satisfy a compound function—as-

suming that we know which objects satisfy the simpler functions from which the

compound one has been constructed. Thus, for instance, we say that given numbers

satisfy “x is greater than y or x is equal to y ” if they satisfy at least one of the func-

tions “x is greater than y ” or “x is equal to y .”7

Once we have a recursive definition of satisfaction we can generate S-sen-

tences for any complex predicate in the language. This means that we getan infinite number of these S-sentences from a finite number of axioms—

axioms for each primitive predicate and axioms for each of the connectives

used to form complex predicates. In other words, we get the effect of an

infinite disjunction of S-sentences from a finite basis. We have analyzed the

complex predicates into their parts and then said something general about

the parts. This solves the problem raised by the infinity of complex expres-

sions in the language. The theory has been finitely axiomatized.

The final stage of the truth definition consists in linking satisfactionto truth. Tarski writes: “Hence we arrive at a definition of truth and false-

hood simply by saying that a sentence is true if it is satisfied by all objects, and

false otherwise.” In effect, Tarski has recursively defined “true of” by using

disquotational S-sentences and then linked “true of” to “true” by invok-

ing the idea of a sentence being true of all objects. This is all really just a

technical way to implement the underlying idea of the T-sentences, which

themselves already contain partial definitions of truth. Thus Tarski achieves

his stated conditions of adequacy.

In the next chapter we will look in more depth at the scope and limits of

Tarski’s construction, as we examine Davidson’s claim that Tarski-style truth

theories provide a framework for doing the semantics of natural languages.

Here we will ask what general significance Tarski’s theory has, beyond that

of recursively defining “true” for particular formal languages. From a purely

logical point of view, it appears that Tarski has accomplished what he set

out to accomplish. The more difficult question is the philosophical upshot

of his work, if any.





9 Davidson’s Semantics for Natural Language

9.1 Background

Tarski’s intention was to define the concept of truth for formalized lan-

guages. Donald Davidson’s aim is to use Tarski’s theory of truth for formal-

ized languages to give a theory of meaning for natural languages. Davidson

is therefore using Tarski’s theory for a purpose that he did not originally

intend—as a form of semantic theory for a natural language. Tarski was

restricting his definition of truth to a limited formal language, taking the

concept of translation (sameness of meaning) for granted, while Davidson

is redeploying his theory to give a theory of meaning for a full natural lan-

guage. Tarski was explaining the nature of truth; Davidson is using truth

to explain the nature of meaning. If Davidson is right, Tarski’s theory has a

much greater significance than its originator realized. It is both a theory of

truth in a limited setting and a theory of meaning in an unlimited setting.

Before we discuss Davidson’s article “Semantics for Natural Language,”

let us make a few background comments. Two thoughts about meaning arevery much in the air during the whole of twentieth-century philosophy of

language, beginning with Frege. One thought is that meaning and truth

are somehow intimately connected. A second thought is that meaning is

essentially compositional: that is, the meaning of a sentence results from the

meanings of its parts. Meaning is something that works constructively, pro-

ceeding from simpler elements to determine the meaning of more complex

elements by means of rules. Putting the two thoughts together, meaning

is something that operates compositionally to generate sentences that are

true or false.

These ideas are present in Frege’s writings, because when Frege is dis-

cussing sense and reference one of his concerns is the reference of parts of



166 Chapter 9

sentences, and the reference is what determines the truth-value of a sen-

tence. Moreover, sense is the “route to reference,” so that sense is under-

stood in terms of the concept of reference. In his view the reference of asentence is its truth-value. So sense is what contributes to truth-value via

reference. Obviously, whether a sentence is true depends on what it means.

The connection between meaning and truth is thus quite overt in Frege and

was made more explicit by later philosophers. A simple formulation of the

connection is that the meaning of a sentence is its truth condition. Let us

talk about that for a moment, so that we get a grip on the underlying ideas

before we discuss what Davidson has to say.

Take a sentence like our old friend “Snow is white”: it means something.If we wanted to say what it means, the most straightforward way would be

to declare: “‘Snow is white’ means that snow is white.” Again, do not sup-

pose that this is trivial just because I have written the same sentence twice.

The proposition expressed is no tautology but a contingent, informative

proposition. If you know that “Snow is white” means that snow is white,

you know something substantial about that sentence. A person who does

not know English could know this proposition. I could say of a monolin-

gual Frenchman “Pierre knows that ‘Snow is white’ means that snow is

white,” thereby attributing to him knowledge of the meaning of a single

English sentence (he need not know what “means” means in English to

know that proposition). You need not know the metalanguage in order that

this language can be used to describe what you know. I can use English to

ascribe knowledge to animals, but I don’t suppose that they know English.

Notice that the sentence “‘Snow is white’ means that snow is white” has the

characteristic structure that we talked about in connection with Tarski. It

both mentions and uses the same sentence. It does not have the same form

as “‘Snow is white’ means ‘La neige est blanche,’” in which two sentences

are mentioned. This sentence reports the correct translation of an English

sentence into a French sentence. So there are two different ways to “give

the meaning” of a sentence: one is by mentioning a sentence that has the

same meaning as the given sentence (giving a translation), the other is by

using a sentence to state the meaning of a mentioned sentence. In the latter

case you can know the proposition expressed without knowing the lan-guage that is used to express it. Thus we can say of a completely monolin-

gual Frenchman “Pierre knows that ‘La neige est blanche’ means that snow

is white” without imputing any knowledge of English to him. But you can’t



Davidson’s Semantics for Natural Language 167

do that if you quote “Snow is white” after “means,” because then you are

ascribing knowledge of an English expression to him.

So in our sample meaning-ascription (“‘Snow is white’ means that snowis white”), a sentence is mentioned on the left-hand side and then used on

the right-hand side, just like a T-sentence (see the previous chapter). The idea

that meaning and truth are connected comes from the very simple observa-

tion that we can substitute for the words “means that” the words “is true if

and only if.” We thereby obtain something that is well formed grammati-

cally and that duplicates the pattern of use and mention that we have noted.

This suggests that to know what a sentence means is to know the condi-

tions under which it is true. To know the meaning of a sentence is to knowits truth condition. At the least, to know a sentence’s truth condition is to

know something about its meaning. Acquiring such knowledge is removing

semantic ignorance to some degree. You may be wondering what a particular

sentence in a foreign language means, and then someone tells you that the

sentence in question is true if and only if the sky is blue. Haven’t you learned

that the sentence means that the sky is blue? Learning the truth condition of

a sentence is learning what the sentence means, apparently. At any rate, it is

learning something important about its meaning.

Let us then entertain the hypothesis that when a person understands

a sentence he or she knows what its truth conditions are. Knowledge of

meaning is knowledge of truth conditions. Many philosophers embraced

this idea about meaning throughout the twentieth century (Wittgenstein

in the Tractatus most prominently). Davidson is in the same tradition. He

is supposing that meaning and truth conditions are at the very least inti-

mately connected. An issue we will discuss later is whether truth conditions

are sufficient for meaning, but they do seem to be necessary, because it

would not be possible to know the meaning of a sentence without know-

ing its truth conditions. How could I know what “Snow is white” means if

I was completely ignorant that “Snow is white” is true if and only if snow

is white? Still, we may wonder whether knowing the truth conditions of a

sentence is sufficient for knowing the meaning of the sentence.

To give you an intuitive sense of things, it seems very natural to suppose

that “Hesperus is a planet” has the same truth conditions as “Phosphorus isa planet,” because truth conditions are determined by reference. The truth

condition that makes both these statements true is that a certain object,

namely Venus, is a planet. As we know from Frege, those two names do not



168 Chapter 9

have the same meaning, however; so identity of truth conditions is not suf-

ficient for synonymy. Referential truth conditions do not add up to sense.

We will come back to this. In any case, it seems as though truth conditionsare very intimately connected to meaning, because they involve reference,

which is fixed by sense. If we did not grasp the truth conditions of a sen-

tence, we would not know its meaning. Thus Davidson’s first idea is that

meaning and truth conditions are connected. A theory of truth conditions

would therefore be a theory of meaning, or something close to it.

Davidson’s second idea is the compositional idea. It is hardly deniable

that language forms a compositional structure. There are a finite number of

primitive elements (“words”) that crop up in various combinations. Theseelements join together according to syntactic rules that generate phrases,

and these phrases in turn combine to form sentences. A sentence is a

complex entity, made up of parts, and these parts can recur in other sen-

tences. It seems just obvious that the meaning of a sentence in a language

is derived from the meanings of the elements that compose it—as obvious

as the fact that buildings are put together from simpler parts. The units of

language are, moreover, exceptionally mobile, because they can jump from

one sentence to another—as when I say “John is quick” and “Jill is quick.”

We human speakers spend our lives recombining old words into new pat-

terns—we seem to get a kick out of it.

Now put those two ideas together and you get the following: the truth

conditions of a sentence depend compositionally on the words that com-

pose the sentence. The compositionality of meaning is the compositional-

ity of truth conditions. The meaning of a sentence is its truth condition,

and the compositionality of meaning is the compositionality of truth con-

ditions. Thus, if we had a compositional theory of truth conditions, we

would have a compositional theory of meaning. The question then is what

a compositional theory of truth conditions would look like.

9.2 The Merits of Tarski’s Theory as Applied to Meaning

Davidson’s proposal emerges from the background outlined above. He

presupposes that background when arguing for the relevance of Tarski tothe theory of meaning. Let us trace how he comes to his conclusion. He

begins by declaring that a theory of meaning should give the meaning of

every meaningful expression. He says this as if it were obvious, but it is not




so obvious. Many philosophers have given theories of meaning without

supposing that a theory of meaning must actually specify the meaning

of every meaningful expression. They proceed at a more abstract theo-retical level, saying variously that meaning is an image in the mind or a

behavioral disposition or a social practice or a certain kind of intention.

Davidson is influenced here by linguistics and Chomsky’s conception of

what a theory of syntax should look like. A syntactic theory is a theory

that specifies (finitely, recursively) which strings of words are grammatical.

It supplies a set of rules that determine which strings are grammatical or

well formed. Such a theory is deemed adequate if and only if the rules cor-

rectly decide which strings of words are grammatical. It gets pretty detailedand specific. Davidson thinks that a theory of semantics should likewise

encompass the whole language, giving meaning rules for every expression.

A syntactic theory tells us, for any string of words, whether it is meaning-

ful; a semantic theory (for Davidson) tells us what each meaningful string

of words means.

However, there is the question what form such a specification of mean-

ings might take. In other words, how should we specify the meaning of

every meaningful expression? Davidson does not in this article give any

examples of alternatives to the theory that he himself favors, but we can

give a few illustrations of what he has in mind. One thing we could do is

to specify meanings by presenting what is called a translation manual. We

could give a specification of meaning for English by providing a translation

of every word and sentence of English into some other language. Thus we

say things like “white” means “blanche.” We could also provide synonyms

from within the same language, as in “bachelor” means “unmarried male.”

We could even give the trivial identity translation: “white” means “white.”

The form of these translation manuals is always the same: there is a pair

of quoted expressions linked by the relation-word “means” or “means the

same as.” If we wanted to do it seriously, we would devise a translation

manual that is compositional. We would not want to provide translations

for every sentence outright, because they are infinitely many sentences.

There would be a finite set of rules by which sentences are translated from

one language to another. Davidson does not think that a good theory ofmeaning should take the form of a translation manual, but that is one obvi-

ous way in which we might set about giving the meaning of every mean-

ingful expression. One might wonder whether there is any other feasible



170 Chapter 9

way. How else can we give the meaning of an expression other than by

providing a synonym for that expression?

A Fregean might suggest that we assign a sense to every expression in thelanguage. Thus we say things of the form: “The word ‘w ’ has sense S.” We saw

when discussing Frege that there are problems with that, because there are

questions as to how we assign a sense to a word. We need somehow to refer to

a sense, but how exactly do we refer to senses? The only way to refer to senses

appears to be to link them with expressions, as in “the sense of ‘white.’”

Then we end up saying, “The word ‘w ’ has the sense of the word ‘w* ,’” where

“w* ” is a synonym of “w .” But this is the translation manual again. So it is

difficult to see how we might implement a systematic assignment of Fregeansense to all meaningful expressions of the language that gets beyond a trans-

lation manual. Still, this would be one conceivable framework for assigning

meanings to expressions. Another approach might be to wheel in psychol-

ogy. Locke and others thought that the meaning of a word is an image in the

speaker’s mind when she utters the word. A specification of meaning might

then involve a specification of which images are associated with the word.

Thus: “The meaning of ‘w ’ is image I .” For example, the meaning of “red”

is an image of red. Here the problem is not so much the form of the speci-

fication but the plausibility of the underlying theory—because the image

theory has been thoroughly discredited (how does it work for the meaning

of “not” and “number” and “believes”?). At any rate, those are some pos-

sibilities about how we might specify meanings, to be set beside Davidson’s

positive proposal. His proposal is quite different and avoids altogether the

locution “Word ‘w ’ means X ,” whatever X may be. His is a theory of mean-

ing in which we don’t talk about the things that words and sentences mean.

Davidson’s first point about the proper form of a meaning specification

is that it must be one that is structurally based, finitely stated, and capable

of generating an infinite output. A natural language like English has infi-

nitely many sentences, so whatever your theory of meaning is it has to

specify meanings for all of those infinitely many sentences. It should not

perform this function one sentence at a time, because then it would be an

infinite specification. What is desirable is a finite number of axioms with

an infinite array of consequences, so that the theory of meaning will besomething that works recursively. Davidson believes that a theory of mean-

ing must have this recursive shape, and this is one main reason he thinks

that Tarski’s theory is right for the job.




A point that Davidson brings up in this connection has a Chomskian

flavor: the theory must be finite because human languages are learnable.

An ordinary child with a finite brain can learn a language that containsinfinitely many sentences. So the child’s potentially infinite mastery of the

language must be finitely based, that is, grounded in a finite number of

semantic principles. The child, being finite, can only learn something that

is finitely specifiable. If it were only infinitely specifiable, then no finite

being could learn it. Learnable languages must be finitely based. Thus the

language must be based on repeatable rules that govern infinitely many

potential cases. At any moment you might hear a sentence you have never

heard before and understand it in an instant. You never learn the meaningof a sentence by learning the meaning as a whole. The way you understand

novel sentences is by analyzing them into their constituent words. Once

you understand the rules that combine those words, you can generate from

that basis what the sentence means. Our understanding of language is a

compositional operation. For a language to be learnable and represented in

a finite brain, the language itself must have finitely many basic semantic

structures with generative potential. Any theory of meaning must reveal

what that generative semantic structure is. If it did not perform that func-

tion, it would treat every sentence as a semantic primitive. Such a theory

would be inadequate because it would not represent an essential feature

of the semantics of natural language, and hence our understanding of

language.

Meaning must be compositional and languages learnable, so we need

a finite semantics. Meaning is also closely linked with truth conditions.

So we need a finite statement of truth conditions if we are to capture the

essence of what meaning is. This is what we know about meaning before

we settle on any specific theory. Davidson is saying that these are general

facts about meaning that any theory of meaning should respect. His bold

proposal, then, is that Tarski’s truth theory fulfills these conditions and

accommodates the general features of meaning we have articulated. Thus

Davidson claims that Tarski’s theory has the right form to qualify as a the-

ory of meaning. It is a finite, structural, recursive assignment of meanings

(i.e., truth conditions) to sentences, capable of generating a potential infin-ity of semantic assignments.

Let us examine in a particular case how the theory recursively gener-

ates truth conditions by analyzing the structure of sentences. We take an



172 Chapter 9

ordinary English sentence like “Snow is white” and we analyze it into a

singular term “snow” and a one-place predicate “is white.” Then we give a

designation axiom for “snow”: “‘Snow’ designates (in English) snow.” Nowwe give a satisfaction axiom for “is white”: “An object x satisfies ‘is white’

(in English) if and only if x is white.” We have broken down the sentence

into its constituent parts and assigned semantic properties to the parts.

Now we need to derive truth conditions for “Snow is white” based on our

axioms. Since this is a subject-predicate sentence, we have a rule that says

that such a sentence is true if and only if the designation of the subject term

satisfies the predicate term. We then consult our axioms to ascertain what

the designation of the subject term “snow” is and what are the satisfactionconditions of the attached predicate “is white.” We find that these are as

specified just now. We can then deduce that “Snow is white” is true if and

only if snow is white. We simply substitute “snow” for “the designation of

‘snow’” and “is white” for “satisfies ‘is white.’” We broke the sentence into

its syntactic parts and then derived the truth conditions from our axioms

dealing with those primitive parts. Thus we derived the truth conditions of

the whole sentence from the semantic properties of its parts. Since meaning

coincides with truth conditions, we have derived the meaning of the whole

from the meanings of its parts.

If we now add axioms for connectives like “and” and “not,” as outlined

at the end of the last chapter, we can derive truth conditions for complex

sentences compounded from these connectives, such as “Snow is white

and grass is not blue.” And now we have a language with potentially infi-

nitely many sentences in it. The primitive expressions recur in different

sentences, so we just need to have axioms that cover those expressions;

the full range of sentences results simply from repetition. Thus Davidson

thinks that Tarski’s theory performs one of the key functions of a semantic

theory: it shows how the meaning of a sentence depends on the words that

make up the sentence, because it shows how truth conditions result from

the structure of a sentence.

Here is a quotation from Davidson summing up much of the above:

What properties do we want [of a theory of meaning]? An acceptable theory should,

as we have said, account for the meaning (or conditions of truth) of every sentenceby analyzing it as composed, in truth-relevant ways, of elements drawn from a finite

stock. A second natural demand is that the theory provides a method for deciding,

given an arbitrary sentence, what its meaning is. (By satisfying these two condi-




tions the theory may be said to show that the language it describes is learnable and

scrutable.) A third condition is that the statement of truth conditions for individual

sentences entailed by the theory should, in some way yet to be made precise, draw

upon the same concepts as the sentences whose truth conditions they state.1

One of the things that Davidson aims to do is articulate the conditions for

what a theory of meaning should accomplish. Other philosophers have

often neglected this point. He wants us to be clear about what a theory of

meaning should aim for, and he gives a set of criteria to determine whether

some proposed theory is a good theory or not. We have talked about the

first two of those conditions, but we have not yet talked about the third

one.A salient feature of Tarski’s theory is that it often evokes a sense of trivial-

ity. It is always saying things like “‘Snow is white’ is true if and only if snow

is white.” If the same sentence is repeated on the right of the biconditional

as appears on the left, this does not appear to be saying much that is inter-

esting about the original sentence. It is not trivial if the object language sen-

tence comes from another language, but in our own language it does seem

quite trivial. Shouldn’t we do a bit more to say what the meaning of the

sentence “Snow is white” is? Shouldn’t we try to be more ambitious, more

informative, and more analytical? We already knew perfectly well that

“Snow is white” means that snow is white. Tell me something I don’t know!

Davidson thinks that this alleged drawback is actually a virtue of the

theory. He thinks it is good that the theory does not draw on any concep-

tual resources not contained in the sentence with which we started. He

thinks the theory should not draw on any fresh or innovative conceptual

resources. He does not give any particular arguments or reasons for that

position, but his basic thought is that the one thing every speaker indisput-

ably knows is that the sentence “Snow is white” is true if and only if snow

is white—and also that it means that snow is white. If our aim is to provide

a meaning specification that captures what the speaker meant by uttering

a sentence, there are no questions or doubts about that specification when

using Tarskian T-sentences. By being conservative in our meaning-ascrip-

tions, we do not go beyond what the speaker ordinarily knows in knowing

the meaning of a sentence. We are not attributing to the speaker dubiousitems of knowledge that he may simply not possess. There is a word for this

conservative approach that Davidson does not use in the article we are dis-

cussing: “homophonic.” That word means that what is on the right-hand



174 Chapter 9

side should be the very same sentence as what is mentioned on the left, or

a straightforward translation of it. What it should not be is an analysis or

reduction or paraphrase or spinning out of the object language sentence(that is, heterophonic). If the T-sentence is homophonic, then we can be

sure that it does not attribute to the speaker more knowledge than he actu-

ally possesses regarding the truth conditions of the sentence whose mean-

ing he grasps. The only concepts he needs to understand “Snow is white”

are the concepts snow and white, so our description of his knowledge should

be restricted to those concepts.

We may wonder what the homophony requirement rules out. David-

son gives the example of modal expressions. Suppose we are interested insentences like “Necessarily 2 + 2 = 4” and we want to provide a T-sentence

for that. A homophonic T-sentence will simply repeat that sentence neat

on the right-hand side. We just peel the quotes off it. But many philoso-

phers have supposed that the semantics of modals should be more adven-

turous. They suppose for various reasons that it is illuminating to invoke

the apparatus of possible worlds. Thus we can analyze the modal operator

“necessarily” as a quantifier over possible worlds—as in “for all worlds w .”

Adopting this analysis, we can write our T-sentence accordingly: “‘Necessar-

ily 2 + 2 = 4’ is true if and only if, for all worlds w , 2 + 2 = 4 in w .” Davidson

would protest that bringing in the ontology of possible worlds introduces

conceptual resources not contained in the original sentence. The original

sentence said nothing about possible words, nor did it have a quantifier in

it. The sentence we began with has been enriched or expanded by bringing

in these alien concepts. The speaker might even complain when presented

with such a T-sentence: “But I don’t believe in the ontology of possible

worlds, and that is not what I meant by ‘necessarily’!”

But this issue is controversial because it is not clear at what point we

have introduced alien concepts into our T-sentence. A possible world the-

orist might insist that he has not introduced alien concepts because the

ontology of possible worlds is implicitly contained in our ordinary talk

about necessity. It is not just a philosopher’s invention—it is the under-

lying meaning of modal sentences. Do we add alien concepts if we write

a T-sentence for “John is a bachelor” by using on the right the sentence“John is an unmarried male”? It becomes rather difficult to argue the issue

because it is not always clear what people ordinarily mean by the sentences




they use. This is no doubt why Davidson hedges his homophony require-

ment with the phrase “in some way yet to be made precise.”

9.3 Applying Tarski’s Theory to Natural Languages

When dealing with a language based on ordinary predicate logic, David-

son’s use of Tarski’s theory of truth to provide a theory of meaning is quite

straightforward. It can be essentially the same theory as Tarski constructed.

The Davidsonian theory of meaning will consist of the Tarskian apparatus

of base axioms, recursive axioms, and rules of combination. But Tarski rec-

ognized that his theory applied only to precise formalized languages, notmessy natural languages. Of course, that limited type of language is not

the whole of language, so there is a question about the rest of language.

Hasn’t the theory dealt with only a fragment of language as we have it? This

is already a problem for Tarski’s aim of defining truth, because the word

“true” applies to many sentences of English that go beyond the resources of

predicate logic languages. He is therefore unable to say what “true” means

when it applies to sentences not translatable into the formal language. But

the problem has special force for Davidson because he is claiming to apply

Tarski’s theory to natural languages in their entirety. If Tarski’s methods do

not apply to certain sentences in natural language, then Davidson cannot

rely on Tarski to give a complete theory of meaning for natural languages.

So Davidson has some obligation to explain to us how we can extend Tar-

ski’s methods to different areas of language. How do we give the meaning

of the parts of language that don’t fit the forms of classical predicate logic?

Davidson is fully aware of this potential problem. He writes of his style

of semantic theory:

What would emerge as the deep problems are the difficulties of reference, of giving a

satisfactory semantics for modal sentences, sentences about propositional attitudes,

mass terms, adverbial modification, attributive adjectives, imperatives, and inter-

rogatives, and so on through a long list familiar, for the most part, to philosophers.2

As he sees it, we need to find ways to assimilate these idioms to semantic

forms that are already susceptible to Tarskian treatment. Let’s now con-

sider some of these idioms. Adverbs provide a particularly instructive case.A theory of truth for sentences containing adverbs needs to specify how

adverbs contribute to the truth conditions of sentences. We need suitable

semantic axioms for adverbs. There is no obvious way to apply Tarski’s



176 Chapter 9

apparatus to sentences such as “John ran quickly,” simply because there

are no adverbs in the formal languages he is concerned with. We certainly

can’t say that objects like John satisfy “quickly”—that makes no sense. Itis necessary to give a different sort of theory of how such adverbial sen-

tences might work. Davidson accomplishes this by paraphrasing adverbial

sentences into sentences that quantify over events and then recasts the

adverbs as predicates of events. For example, he would paraphrase the sen-

tence “John ran quickly” as “There was an event e such that e was a running

by John and e was quick.” We have thereby replaced the adverb “quickly”

with the adjective “quick” and applied it to an event (not to John himself).

Then we can give a satisfaction axiom for the predicate “quick” in the usualway: an event e satisfies “quick” if and only if e is quick. What Davidson

has done here is translate the grammatically adverbial sentence into a sen-

tence without adverbs and replaced them with adjectives (predicates) that

apply to events. Thus the forms familiar from predicate logic are shown

capable of including the adverbial constructions of English and other natu-

ral languages.

Another example involves the so-called intensional operators. These

go all the way back to Frege. John believes that Hesperus is a planet, but

John does not believe that Phosphorus is a planet, despite the identity of

Hesperus and Phosphorus. Since “Hesperus” denotes the same planet as

“Phosphorus,” we find that we cannot substitute co-denoting names inside

belief contexts. Contexts of this kind are called opaque. As Frege pointed

out, the truth conditions of sentences containing intensional operators like

“believes that” depend on the sense of the embedded name, not the refer-

ence. We therefore cannot have a comprehensive axiom for a name that

simply gives its reference, because that cannot capture the contribution

that the name makes to sentences containing intensional operators. Some-

times the name affects the truth-value of a sentence in a way that goes

beyond its reference and brings into play what Frege calls sense. For this

reason, our account of the semantics of names is incomplete if it only gives

their reference. Something else must be added to that, and it is not clear

how we can accommodate these cases within the framework Tarski laid

down. Tarski’s theory just specifies references for singular terms by meansof designation axioms. Sense is ignored. For Tarski’s own purposes, that is

fine, because he is only interested in defining truth for languages without

intensional operators. But Davidson wants the Tarskian framework to apply




to all the linguistic constructions of natural language—and that is a much

taller order. How can a semantics devised for purely extensional languages

deal with intensional languages?As it happens, Davidson does have a theory of intensional contexts that

is ingenious and purports to solve the problem (see his paper “On Saying

That”3). Consider the sentence “John says that the sky is blue.” Davidson

proposes that we analyze that sentence in the following way: “The sky is

blue. John said that.” We divide the original sentence into two, separated

by a period, but linked by the demonstrative “that” that refers back to the

first sentence. It is just like you saying something and I reply, “I just said

that!” The point of this analysis (sometimes called the paratactic theory ) isthat by removing the embedded sentence we abolish the intensional opera-

tor. There is no opaque context left anymore. In the sentence “The sky is

blue” by itself we can substitute any co-denoting term and not alter the

truth-value of the sentence. It does not occur within an intensional context

as part of a complex sentence. It is a separate sentence, so everything here

is extensional. We can therefore apply Tarski’s extensional theory and not

run into any problems. Likewise, the sentence “John said that” is com-

pletely extensional—in particular, we can substitute any term that refers to

the same thing as “that” and not alter truth-value. This demonstrative can

be taken to refer to the proposition expressed by the first sentence, so any

term that refers to the same proposition will not change truth-value. So, by

ingenious paraphrase, we can bring these apparently intensional contexts

into the Tarskian fold: they turn out to be extensional after all. (There is a

lot more that can be said about this proposal of Davidson’s, as well as about

his theory of adverbs, but we are just trying briefly to provide a flavor of

how he would go about extending Tarski’s framework to natural languages.)

There is also the matter of nonindicative sentences, which apparently

lack truth conditions altogether. The imperative “Shut the door!” does

not appear to be true or false. A straightforward method here would be to

translate these sentences into indicative sentences. We paraphrase “Shut

the door!” into “I order you to shut the door.” The latter sentence can be

true or false, depending on whether I did indeed order you to shut the

door. And it would generally be true because in saying “I order you” I did order you (these kinds of speech acts are called performatives). Again, we

find a suitable paraphrase of the original sentence that can be subjected to

the Tarskian treatment—since the paraphrase does have truth conditions.



178 Chapter 9

These examples illustrate the kind of massaging that needs to be done on

natural language sentences in order to make the Tarskian semantic frame-

work applicable to natural languages. At the least, Davidson would say, itis not obvious that we cannot extend Tarski’s truth theory further than

initially appears. Trying to do so constitutes a research program (this means

it will keep eager graduate students busy for a few years).

The case of indexicals poses an awkward problem for the homophony

requirement. Suppose we give a homophonic T-sentence for “I am hot”—

that is, “‘I am hot’ is true (in English) if and only if I am hot.” The problem

is obvious: no one can say truly “I am hot” unless I (Colin McGinn) am hot.

But people other than me can be hot and can truly say they are with thesentence “I am hot”—without me being hot. Davidson’s homophony con-

dition clearly fails here. We obviously need to write our T-sentence along

the following lines: “‘I am hot’ is true for speaker S at time t if and only if S is

hot at t .” That is the correct truth condition for the English sentence “I am

hot.” Fine, but that T-sentence is not homophonic, because the right-hand

side does not repeat the sentence mentioned on the left. We must eliminate

the word “I” altogether and add in “S” and “t .” We thus use conceptual

resources not present in “I am hot”—the right-side is just not a synonym

of the sentence mentioned on the left. Homophony violated! That seems

the right way to go, but then we wonder how Davidson can stipulate his

homophony requirement in the first place. How can he formulate it so as

to rule anything out, while making an exception for indexicals? Add to this

the point that his treatment of adverbs also seems to violate homophony,

with the added quantifiers and ontology of events, and the requirement

begins to lose any bite. How can it rule out possible world paraphrases of

modal idioms, say, if it lets in nonhomophonic T-sentences for indexicals

and adverbs?

Davidson’s theory of meaning makes no attempt to define the semantic

primitives. There is only an assignment of logical form. He makes a sharp

distinction between defining the primitive expressions and giving the log-

ical form of sentences. In his way of looking at things, base axioms for

primitive terms would be something like the following: “‘Snow’ designates

snow,” and “An object satisfies ‘white’ if and only if it is white.” His theoryanalyzes the logical structure of sentences but it does not analyze individual

words. The theory will tell us that a sentence is made up of a singular term

and a one-place predicate or that a complex sentence is a conjunction, but




it won’t tell us that “bachelor” means “unmarried male,” say. This kind of

theory has been described as modest , because it refrains from venturing into

the analysis of word meaning. The label is not very apt, because giving logi-cal form is not something that is trivial or obvious or uncontroversial. But

the idea is that giving logical form is something quite different from analyz-

ing individual expressions. The former is thought necessary and admirable,

while the latter is taken to be optional and vaguely taboo.

Describing logical form involves determining the semantic categories of

words. This can be quite nontrivial. Consider again the word “snow” and

the sentence “Snow is white.” If we treat that sentence as having the logical

form of a straightforward subject-predicate sentence, as we have so far, thenwe are treating “snow” as a singular term, a name of snow, whatever snow

may be (is it the totality of all lumps of snow or is it more like a Platonic

universal—the Form of Snow?). Then, we would write an axiom for “snow”

that would be just like the axiom for a name like “Hesperus” (“‘Snow’ desig-

nates snow,” “‘Hesperus’ designates Hesperus”). But if we thought that the

word “snow” is not a singular term but really a predicate, then we would

formulate its axiom in the following style: “x satisfies ‘snow’ if and only if

x is (a piece of) snow”—it would get a satisfaction axiom, not a designation

axiom. Such a semantic classification would award a different logical form

to the sentence “Snow is white.” Instead of having the logical form “ Fa,” a

singular term plus a predicate, it would have the logical form of a universal

quantification, as in “For all x, if x is (a piece of) snow, then x is white.” The

word “snow” would then be placed in a different semantic category—that

of predicate, not singular term. (In fact, “snow” is what is called a “mass

term,” and we have just described two ways of handling mass terms—as

names or predicates.) Similarly, in Davidson’s treatment of adverbs, a word

like “quickly” is transformed into a predicate in the assignment of logi-

cal form. In his treatment of indirect discourse, the word “that” in “John

said that the sky is blue” gets classified as a demonstrative and hence as a

context-dependent singular term. None of this semantic classifying is espe-

cially modest—it’s pretty bold.

Formal languages are supposed to be unambiguous, so what about ambi-

guity in natural languages? For instance, the word “bank” is ambiguous,meaning either the bank of a river or a bank for money. This is called lexical

ambiguity . But there is also syntactic ambiguity , as in the example Davidson

cites: “They came by slow boat and plane”—was it just the boat that was



180 Chapter 9

slow or also the plane? Clearly truth conditions will vary when there are

ambiguities, so we will need to resolve ambiguities before constructing our

T-sentences. We don’t want to end up with monstrosities like “‘Samanthalay down on the [river] bank” is true if and only if Samantha lay down on

the [money] bank.” Here we might simply index the word “bank” so that

the ambiguity is removed—“Rbank” and “Mbank.” For the syntactic ambi-

guity, a bracketing device would suffice, as in “They came by (slow boat and

plane)” and “They came by (slow boat) and plane.” (This bracketing device

is used in standard logic to indicate scope.)

It is important to note that the T-sentences themselves are not the whole

story. They assign truth conditions, and hence meanings, but the meat ofthe theory does not reside in the T-sentences alone. There is also the proof

of the T-sentences. Davidson makes the point that we have to derive the

T-sentences from a finite set of axioms that reflect a recursive structure—

the repeatable occurrence of semantic primitives. The illumination comes

not just from the final results—the theorems—but also from the process

of deriving the theorems from an analysis of the semantic structure of the

sentences. We see how the constituent words generate the sentence’s truth

condition. The point for Davidson is that the theory must be structural

and hence explain how an infinite language can derive from a finite foun-

dation. There is much more to Tarski’s theory than the output of T-sen-

tences, lovely as these are; there is the whole complex apparatus of axioms

and derivations that generates that output. It’s the journey as well as the

destination.

Another merit Davidson sees in this theory is that it allows us to give

a theory of meaning without postulating meanings as entities. W. V. O.

Quine lurks in the background here. Quine is notorious for his antipathy

to meanings as entities (“creatures of darkness,” he calls them, threats to

clean living, etc.). Quine wonders how we might set about counting these

elusive entities, distinguishing them one from another. How many mean-

ings are there in this book, for example? Davidson thinks it’s a big advan-

tage of Tarskian semantics that there is no need to assign any “meanings”

to words in the theory of meaning. This is a theory of meaning that does

without any special entities called meanings (senses, intensions). Instead,it assigns references to words—and references are honest, civilized citizens,

not murky shadows hovering in the vicinity of words. We say, “‘Hespe-

rus’ refers to Hesperus” with confidence in our theory, but we say nothing




about any alleged semantic wraiths describing themselves as “senses.” Yet

we succeed in saying what sentences mean (supposedly: see below). In the

case of predicates, the theory assigns no entity at all, not even a reference.We simply reuse the predicate in our satisfaction axiom. Consider again an

axiom like the following: “x satisfies ‘white’ if and only if x is white.” Notice

that there is no reference here to anything that might be denoted by the

predicate “white.” We could have said, “‘White’ designates Whiteness,” but

we said no such thing. Instead, we say that an object satisfies “white” if and

only if it is white, with no reference to any supposed abstract entity named

“Whiteness.” There is no singular term in this statement for anything that

is being assigned to the predicate—no properties, universals, senses, or thelike. The axiom gives a condition under which the predicate is satisfied but

without committing us to any strange entities of the kind that so horrified

the fastidious Quine. The only entities referred to in the satisfaction axiom

are the ordinary objects we need anyway—the spatiotemporal objects that

are white. Similarly, Tarski does not interpret the connectives by specify-

ing a reference for them—he doesn’t say “‘And’ designates conjunction.”

He just says a sentence of the form “ p and q” is true if and only if “ p” is

true and “q” is true. Using the word “and” on the right-hand side does

not mean we must assign any reference to the word. It is a theory of mean-

ing without things called “meanings”—without peculiar semantic entities.

Words and sentences mean something, and we can say what they mean,

but there are no meaning entities such that words and sentences mean

them. So Quine need not worry that talk of “theories of meaning” threatens

to unleash a disreputable ontology of meanings that besmirch his clean and

tidy universe.

9.4 Empirical Truth Theory

With meanings safely out of the way, Davidson broaches the question of

the empirical status of Tarski-type truth theories. That is, how would you

verify that a particular theory is correct? There are two cases to consider:

the case where the object language and the metalanguage converge, and

the case where the object language is different from the metalanguage. Letus take the simple case where we are giving a truth theory for our own

language. How could we verify that its theorems are correct? Davidson sug-

gests that it is quite simple to do so, because we can look at the theorems



182 Chapter 9

and see from their very form whether they are correct. If the theory says

“‘Snow is white’ is true if and only if snow is white,” we can see immedi-

ately that it has to be correct. But if it said “‘Snow is white’ is true if andonly if the stock market is about to collapse,” we would know it had gone

badly wrong somewhere, since that is a far cry from what “Snow is white”

means. Our own semantic competence enables us to judge whether the the-

ory has got the truth conditions right for a given sentence. The T-sentence

is empirically correct just when the sentence used on the right is the same

as the sentence mentioned on the left. It is thus easy to tell in our own case

whether the T-sentences are correct. (Actually, he seems to be forgetting

here that not all T-sentences are homophonic. Is it so easy to judge that aT-sentence incorporating his own theory of adverbs is correct? Certainly we

cannot just check to see if we have the same sentence twice, since we don’t.

It might be quite controversial whether the following T-sentence is true:

“‘John ran quickly’ is true if and only if there is an event e such that e was

a running and e was by John and e was quick.” But it is at least true that we

can decide such questions by consulting our own competence, since we do

understand “John ran quickly.”)

Davidson makes the rather interesting observation that it can be easier

to judge the truth of a T-sentence than to judge the grammaticality of the

sentence at issue. He writes:

It may in fact be easier in many cases for the speaker to say what the truth condi-

tions of a sentence are than to say whether the sentence is grammatical. It may not

be clear whether “The child seems sleeping” is grammatical; but surely “The child

seems sleeping” is true if and only if the child seems sleeping.4

This implies that it is easier to know what a sentence means than to knowwhether it is meaningful. One might have thought that we first decide if a

sentence is meaningful and then inquire into its meaning, but if Davidson

is right it can be the other way around. But how far can this go? Do I know

that “The ocean swims nightly to itself” is true if and only if the ocean

swims nightly to itself, even if I doubt that this sentence makes sense?

What about “‘Dawn and not sun upward grim’ is true if and only if dawn

and not sun upward grim”? Or “‘The’ is true if and only if the”? Repetition

is surely not enough if the sentence is totally whacky.These are Davidson’s reflections on the domestic case, but what about

verifying a truth theory for a foreign language? How do we know when we

have got someone else’s truth conditions right? We can’t consult our own




competence in the language, because we don’t have any. We have to find

out what the alien speakers mean by their words. At this point in the paper

Davidson alludes to Quine’s discussion of radical translation. Quine hasa famous thought experiment where a traveler goes to a foreign land and

comes across a tribe of speakers whose language has never been translated.

The traveler fancies himself a field linguist. He engages in radical transla-

tion, translation from scratch, with no dictionary. Quine’s question is how

he would begin the process of radical translation and whether it is pos-

sible to arrive at a definitively correct translation scheme for the language.

Davidson takes over that question because he is interested in how we might

verify a truth theory for a radically foreign language. In other words, hewants to determine how we could empirically assign truth conditions to

sentences.

Quine gives the example of the word “gavagai” in the radical translation

thought experiment. The traveler immerses himself in the tribe’s culture,

observing their linguistic behavior, and sets out to discover what they mean

when they utter the word “gavagai.” The traveler has no dictionary because

this is radical translation. How could our traveler figure out the meaning of

the word? There is no use asking the natives because he won’t understand

what they say, and they don’t speak his language either. The first thing he

would do is to discover when and where they utter “gavagai”—in response

to what immediate sensory presentations. What, say, are they looking at

when they say “gavagai”? Suppose our traveler notices that the natives say

“gavagai” just as a rabbit runs by them. Our traveler might conclude that he

knows what “gavagai” means—it means “rabbit.” The general idea is that

he looks around the natives when they utter the word and makes a hypoth-

esis about its meaning. We might agree with the traveler that the correct

translation of the word “gavagai” is indeed “rabbit,” because the natives are

observed to utter that word if and only if there is a rabbit running by. Since

our traveler is a keen student of Tarski, he records his hypothesis in the

form of a satisfaction axiom: “x satisfies ‘gavagai’ if and only if x is a rabbit.”

Quine would put the situation by saying that a rabbit is part of the stimu-

lus meaning of the word. The natives are stimulated to utter “gavagai” just

when there is a rabbit in their sensory vicinity. If you trace the stimulusback from their sense organs to the environment, you will find a rabbit at

the other end. Now Quine makes his killer point: even though the natives

might intone “gavagai” when and only when they see a rabbit, that does



184 Chapter 9

not necessarily imply that “gavagai” means “rabbit.” As logicians say, it does

not imply that the set of rabbits constitutes the extension of “gavagai.”

Even though it is quite true that rabbits are included in the stimulus mean-ing, there are other things that are included in the stimulus meaning too.

One of the things that is included in the stimulus meaning of “gavagai”—in

addition to rabbits—is parts of rabbits, say, the ears. So “gavagai” might

mean “rabbit ears.” Whenever there is a rabbit present, rabbit ears are also

present. Of course, there could be a case where our traveler holds in his

hand a pair of severed rabbit ears, and he finds that the natives don’t utter

“gavagai” in the presence of the ears alone. He can then rule out the rabbit

ear hypothesis. But then it occurs to our wily translator that “gavagai” couldstill mean, “rabbit ears still attached to a living rabbit.” And then he realizes

that it could also mean “temporal stage of a rabbit” or “retinal cause of my

sensations of rabbit” or even “visual percept of a rabbit” (the native never

utters “gavagai” unless he is perceiving a rabbit). It might indeed mean “rab-

bit flea” so long as rabbits and their fleas keep close enough company. The

point is that there may be many things in the causally operative environ-

ment (or even in the native’s own heads) that the word might mean. We

cannot easily determine just what specifically the word means (what its

extension is). As a result of these reflections, Quine comes to the startling

conclusion that what the native means is radically indeterminate (and he

even extends the indeterminacy thesis to what we mean by our words).

There is no “fact of the matter” as to what “gavagai” means (or what our

word “rabbit” means come that that).

Davidson is not in this paper concerned with indeterminacy, though

elsewhere he expresses general agreement with Quine’s thesis. His concern

here is with Quine’s general picture of how we set about devising and test-

ing interpretations of the language of others. This brings us to Davidson’s

theory of what he calls radical interpretation. Davidson has a whole article

(“Radical Interpretation”5) that delves into this question. Here we must be

brief. Roughly, he thinks that we need to assign truth conditions accord-

ing to the external environmental causes of utterances. If a native “holds

true” a sentence just when a certain state of affairs objectively obtains in

the environment, we must suppose that the sentence is true just whensuch a state of affairs obtains. If this leaves gaping indeterminacies, then

so be it. As a way to constrain our interpretations further, Davidson advo-

cates what is called the principle of charity —that is, the interpreter must




interpret his subjects in such a way that their beliefs and assertions come

out mainly true. We cannot suppose that our native is massively in error,

totally deluded, riddled with false beliefs. Of course, the native could bemistaken about there being a rabbit in front of him when he utters the

word “gavagai”—he might be prone to rabbit hallucinations (that strange

plant he insists on smoking all the time). But Davidson maintains that we

have to attribute mainly true beliefs to our native if we are to understand

him at all. Speakers are uninterpretable (indeed, for Davidson, impossible)

unless the principle of charity applies to them. Since we ourselves are inter-

pretable (and apparently possible) that means that we cannot be massively

in error either. And that implies that skepticism about our beliefs has to bewrong: we must have mainly true beliefs, no matter what the skeptic says.

There is a whole discussion about this set of issues, ranging into philoso-

phy of mind and epistemology, but we will not cover it here. We have said

enough to indicate how Davidson views the project of verifying theories of

meaning for alien speakers.

9.5 Criticisms of Davidson’s Theory

Let us assemble some criticisms of Davidson’s theory of meaning. First we

can ask whether Davidson says enough about what meaning is—and about

what our grasp of meaning consists in. His guiding idea is that a theory of

meaning assigns truth conditions to sentences, and a speaker’s understand-

ing of a sentence consists in his knowing what the truth conditions are.

Thus, to understand “Snow is white,” the speaker merely needs to know

that this sentence is true if and only if snow is white. This account of mean-

ing raises the question: is it enough to say that knowledge of meaning is

knowledge of truth conditions—especially when we restrict ourselves to

homophonic statements of truth conditions? Isn’t this just way too mini-

mal? Can’t we ask what this knowledge of truth conditions itself involves?

There are different options to take in response to this line of criticism.

Davidson’s response would be that we do not need to dig any deeper into

linguistic understanding to have an acceptable theory of meaning. Maybe

a psychologist could say more about linguistic understanding, but fromthe point of view of philosophical semantics we have achieved our aim,

namely, to specify meanings systematically and show how an infinite mas-

tery can proceed from a finite basis. To venture further is to stray into vague



186 Chapter 9

and swampy territory. If we stick with Tarski, pure and simple, we have

all the security of rigorous formal logic, without speculations about what

might be secretly on the speaker’s mind when she understands sentences.Alternatively, we might borrow an idea from Wittgenstein’s Tractatus. He

there takes the view that when a speaker understands a sentence he grasps

what possible state of affairs would make the sentence true. To understand

“Snow is black” you must grasp the state of affairs that would make that

sentence true. That is a merely possible state of affairs, not an actual state of

affairs. We apprehend possibilities by the faculty of imagination. We imag-

ine such a state of affairs when we grasp the meaning of “Snow is black.”

So when I understand the sentence “Snow is black” what I do is imagina-tively conjure up a possible state of affairs where snow is black. I might

form a mental image of black snow. The fact that I imagine that state of

affairs and not some other state of affairs is what my grasp of the sentence’s

meaning consists in. If I were to imagine the state of affairs of snow being

blue, I would not have imagined the state of affairs that corresponds to the

sentence “Snow is black”—I would have misunderstood the sentence. This

Wittgensteinian account of knowledge of truth conditions goes beyond

Davidson’s austere minimalist account. It is Tarski plus the modal imagi-

nation. The speaker has to employ her modal imagination in order to get

her mind around meaning. This is a richer psychological story than David-

son’s proudly modest account. It attempts to illuminate nontrivially what

knowledge of truth conditions involves psychologically.

Another approach that many philosophers have favored is to bring in

the notion of verification. The ability to verify the sentence “Snow is white”

is what knowing its truth condition amounts to. To verify this sentence we

need to seek out some snow, check it over, and decide what color it is. We

need to see with our own eyes that it is white. To do that, we have to know

where to look—we have to know that snow falls from the sky and covers

the hills and dales in winter. If someone tried to verify the sentence “Snow

is white” by examining the lava spewing from volcanoes, he would show

himself not to understand “Snow is white.” The ability to verify the sen-

tence in the right way is clearly connected to knowing its truth condition.

If you know the truth condition of a sentence, you generally have a prettyshrewd idea of how to verify it. If you don’t, you are clueless. Some phi-

losophers (often describing themselves as positivists) try to elevate these tru-

isms into a theory of what knowing truth conditions is—it comes down to




knowing what kind of evidence counts in favor of the truth of a sentence.

This turns knowledge of truth conditions into knowledge of verification

conditions. The view may be horribly misguided, but at least it attempts tospell out what knowing truth conditions might be. (The right view is that

we generally have two pieces of knowledge regarding a sentence: knowledge

of what state of affairs would render it true, and knowledge of the kind of

evidence that would warrant assent to it.)

A second line of criticism brings us back to Frege. Tarski’s axioms for

names are designation axioms, and so they only assign reference to names.

For Tarski that is fine, because sentences containing names are true only

depending what the names refer to. If we are only interested in definingtruth, it does not matter what name we use, so long as denotation is pre-

served. If “Hesperus is a planet” is true, then “Phosphorus is a planet” is also

true. But those sentences do not mean the same. This is why Frege brought

in sense to beef things up. We need to assign more than reference to a name

if we are to capture its full meaning. We need something like sense. But Tar-

ski’s semantic apparatus does not specify sense. How then can it function as

a theory of meaning? At best, it is a theory of reference.

A third criticism is that Davidson’s theory provides no explanation of

how words come to have semantic properties. The axioms say things like

“‘Hesperus’ denotes Hesperus,” but nothing in the theory tells us how it

is that a word like “Hesperus” acquires reference. Similarly for predicates

and satisfaction. The axioms don’t explain what gives marks and sounds

the semantic features they have. What constitutes reference? Many philoso-

phers of language have felt that we need an explanation of relations like

denotation. We can’t just accept them as primitive. In other words, a satis-

factory theory of meaning must propose an account of denotation. Some

tough-minded philosophers have even undertaken to explain reference and

satisfaction in physical terms. But in Davidson’s Tarski-based theory, deno-

tation is taken for granted. At the least, we need to supplement the Tarskian

semantics with some sort of explanatory theory of denotation; it is not by

itself a complete account of meaning in natural languages.

Fourth, Davidson distinguishes sharply between giving logical forms for

sentences and giving analyses of individual words. But how robust is thatdistinction? The intuitive idea Davidson is working with is that in attrib-

uting logical forms we do not break words down into parts, but in lexical

analyses we do. He is skeptical about the whole idea of analyzing primitive



188 Chapter 9

predicates, but he is an enthusiast of logical form attributions. But con-

sider Russell’s theory of descriptions (see chapter 3): here we break the word

“the” down into a complex quantified conjunction. Why isn’t this lexicalanalysis? It certainly involves taking a unitary word and analyzing its mean-

ing into separate and more primitive parts. How does this differ from ana-

lyzing “bachelor” as “unmarried male”? Similarly, Davidson’s own theory

of adverbs construes sentences containing adverbs as quantifications over

events with predicates of events. The logical form here is quite different

from the superficial syntax of the sentence. The paraphrase finds hidden

semantic complexity in adverbs. Why isn’t this a case of lexical analysis?

What about modal words like “possibly”? A standard account has it that“possibly” means “There exists a possible world.” The modal adverb goes

over into an existential quantifier over worlds. This looks like an exercise

in conceptual analysis, but it is also an attribution of logical form. If we

want to know what the logical form of “possibly p” is, we are told that this

sentence means the same as “There exists a world w such that p in w .” But

this is at the same time a conceptual analysis of “possibly.” Again, there

is no clear distinction between accounts of logical form and lexical analy-

ses. The alleged distinction evaporates on closer inspection. Yet Davidson

seems wedded to ruling out lexical analysis while championing logical form

assignment. One suspects that he has bought into Quine’s rejection of the

analytic–synthetic distinction, while seeing the merits of certain theories of

the meaning of syntactically simple terms. In fact, these two stances are in

tension. However, this is a subject beyond our current purposes, so we will

not pursue it.

We must finally scrutinize a rather fraught passage from Davidson:

A theory of truth entails, for each sentence s, a statement of the form “s is true if and

only if p” where in the simplest case p is replaced by s. Since the words “is true if and

only if” are invariant, we may interpret them if we please to mean “means that.” So

construed, a sample might then read “‘Socrates is wise’ means that Socrates is wise.”6

Davidson seems to believe that we can substitute “means that” for “is true

if and only if” in a Tarskian T-sentence (“if we please”) and say essentially

the same thing (what this has to do with the “invariance” of “if and only if”

remains a mystery). Thus a truth theory can do duty as a meaning theory.The gap from truth to meaning has been crossed by this simple substitution.

If Davidson does believe this, he is wrong. The biconditional “is true if and

only if” does not mean “means that”—not by a long chalk. In elementary




logic “if and only if” is called a “material biconditional” and a sentence

containing it is true when the two sentences on either side of it are true.

Thus “Snow is white if and only if grass is green” is a true sentence. By thesame token, “‘Snow is white’ is true if and only if grass is green” is true, if “if

and only if” is the material biconditional (i.e., is a truth function). Now per-

form Davidson’s substitution, replacing “if and only if” with “means that.”

We obtain the sentence “‘Snow is white’ means that grass is green.” That is

false, egregiously so. The English sentence “Snow is white” most certainly

does not mean that grass is green! If Davidson were right, any sentence of

English would mean the same as any other sentence sharing its truth-value.

That would be the total collapse of meaning and disqualifies any theory ofmeaning entailing it from serious consideration.

But it may be replied that this is only if we adopt the material bicondi-

tional interpretation of “if and only if.” Davidson does appear to be doing

just that, but maybe this is a slip. Can’t we offer him a stronger bicon-

ditional? Not the material biconditional but the strict biconditional. The

strict biconditional requires not just actual identity of truth-value for the

joined sentences but identity of truth-value in all possible worlds, that is,

necessary coincidence of truth-value. The sentences “Snow is white” and

“Grass is green” have the same truth-value in the actual world but not in all

worlds, because in some worlds grass may be blue and snow is still white.

But we can quickly see that this will not solve the basic problem. Suppose

we have a sentence like “2 + 2 = 4 if and only if 3 + 3 = 6.” The joined sen-

tences are both true in all possible worlds, so this biconditional sentence

is true under a strict modal interpretation of “if and only if.” Now we can

run through the same argument again. If we replace “if and only if” with

“means that” in a T-sentence we get: “‘2 + 2 = 4’ means that 3 + 3 = 6.” That

is no better than before. This ascription of meaning is quite wrong.

The simple fact is that “means that” is far more restrictive about substi-

tutions within its scope than “is true if and only if,” no matter how strict

you are about the biconditional. The only way to get something that adds

up to “means that” out of “is true if and only if” is to stipulate that you are

going to mean the former by the latter. But that would be a futile verbal

ruse, gaining us nothing. It would also completely wreck the idea of usingTarski’s theory of truth as a theory of meaning, since the words “is true if

and only if” would no longer mean what they now do. In sum, the passage

from Davidson is an error.





10 Grice’s Theory of Speaker Meaning

10.1 Background: Speakers and Sentences

We now move to a discussion of a short but groundbreaking paper by H.

P. Grice, called simply “Meaning.”1 The paper is very dense and there isn’t

much reiteration of points, so it calls for careful reading. Let’s begin by

explaining the larger project that Grice was trying to make progress on in

this paper. He is really interested in how words and sentences come to mean

what they do mean—how word meaning and sentence meaning arise. What

makes bits of language express meaning? Grice has a very natural, intuitive

answer to that question: it has something to do with the way speakers mean

things. It is not that words mean what they do because there is some fact of

nature that makes them do so. It is not as though meaningful words came

preformed in nature and humans decided to exploit this naturally given

fact. Words are not like apples on a tree, patiently waiting for us to pick

them. Meaningful language is not an independent phenomenon that we

tap into. Language does not predate the existence of speakers. English wasnot just lying around and we discovered it. Words in themselves are just

sounds or marks that we produce with our voices or our hands—there is

nothing about them intrinsically that determines what they mean, or that

they mean anything. The meaning of words is arbitrary and conventional,

the result of a kind of decision. Meaning is conferred on words. But it is not

conferred by nature or by God—it is conferred by us. It is what we humans

do with words that makes them mean what they do. Presumably this brings

in the human mind somehow, because it can hardly be the human body

that gives meaning to words (the kidneys, the toes).

Grice focuses on the notion of an agent meaning something by his or

her actions. More specifically, he introduces the notion of speaker meaning .



192 Chapter 10

Not only do words and sentences mean things; speakers mean things by

their words. We use the word “means” in both ways. We can say that the

sentence “Snow is white” means that snow is white, and we can also saythat a speaker meant that snow is white by uttering the sentence. We must

distinguish sentence meaning from speaker meaning—words do the for-

mer, human agents do the latter. But we must also inquire into how the two

types of meaning are connected.

Grice proposes that sentence meaning derives from speaker meaning.

It is because people mean things by their words that those words come to

mean what they do. We have not yet explained or analyzed the notion of

speaker meaning, but the notion is already familiar enough for us to under-stand the thesis that speaker meaning is the basis and origin of sentence

meaning. Words come to mean what they do in virtue of the fact that we

mean various things by them. We confer meaning on them by meaning

something by them. Linguistic meaning thus comes from us—we create it

by acts of speaker meaning. Following that initial thought, Grice is propos-

ing that we analyze word meaning in terms of speaker meaning. If we can

do that, we will have explained how words mean what they do. This will be

a major philosophical achievement. But first we need to know exactly what

speaker meaning is, as well as how it connects with sentence meaning.

Sentence meaning is properly described as semantic meaning : it relates

to words considered independently of speakers. We make no reference to

a speaker when we say “‘Snow is white’ means that snow is white.” But

speaker meaning is properly described as pragmatic meaning because it

explicitly refers to speakers—people mean things in this sense. The word

“pragmatic” here has nothing to do with the doctrine called “pragmatism,”

still less with the notion of the merely practical. It merely connotes the fact

that speaker meaning is about the relation between agents and language.

Semantics is about words themselves and what they mean, but pragmatics

is about speakers and what they do with language. (Syntax is about words

considered independently of their meaning.) Put in these terms, Grice is

suggesting that pragmatic meaning has primacy over semantic meaning.

We could state his position in another way: semantic meaning is ulti-

mately psychological, because for a sentence to mean something is forspeakers to use the sentence while in a certain psychological state—that

of meaning something by the sentence. We shall see later just what this

psychological state is. What Grice is proposing, in effect, is that we can



Grice’s Theory of Speaker Meaning 193

explain semantics in terms of psychology. Sentence meaning comes down

to psychological facts about speakers. This seems to conflict with Frege’s

doctrine (expounded in chapter 1) that senses are not psychological. Sensesare abstract entities, according to Frege, objective things that do not depend

on minds at all. The Gricean approach to meaning appears to be rejecting

that Fregean position. In effect, Grice takes the meanings of words to be

reducible to psychological facts, contra Frege.

That is the program that hovers in the background of Grice’s paper

“Meaning.” In subsequent papers, he labors to develop the program of try-

ing to reduce semantics to psychology, and others joined him. In the pres-

ent paper he focuses on understanding just what speaker meaning is. Let usturn to this seminal paper.

10.2 Two Types of Meaning

Grice begins by distinguishing between two types of meaning, which he

calls “natural meaning” and “non-natural meaning.” He then devotes the

rest of the paper to explaining non-natural meaning. It is easy to grasp this

distinction at an intuitive level. As an example of natural meaning, Grice

gives “Those spots mean (meant) measles,” which might be paraphrased

“Those spots are a symptom of measles.” We can infer measles from the

spots, so the spots mean measles. The spots are a natural sign of measles.

Another example: “The recent budget means that we will have a hard year

ahead.” Given the tightness of the budget, money will be scarce in the com-

ing year. We can infer hard times ahead from the budget. A third example

not given by Grice would be: “Those clouds mean rain.” This says some-

thing like, “There is a natural association between clouds and rain, so we

can infer the latter from the former.”

We can contrast these examples of natural meaning with the following

examples of non-natural meaning: “Those three rings on the bell (of the

bus) mean that ‘the bus is full,’” and “That remark, ‘Smith couldn’t get on

without his trouble and strife,’ meant that Smith found his wife indispens-

able.” These are very British examples, perhaps not familiar to all readers. In

Grice’s day (circa 1957) conductors on buses would ring the bell three timesto indicate that the bus is full (they had different numbers of rings for start-

ing and stopping). The second example involves what is called “Cockney

rhyming slang,” a dialect of east London that substitutes charming phrases



194 Chapter 10

for common words: “trouble and strife” for “wife,” “apples and pears” for

“stairs,” and so on. The speaker says, “I can’t get on without my trouble and

strife” and he means that he finds his wife indispensable.We can see intuitively that “means” is being used differently in these

two types of case, and Grice suggests some marks that distinguish the cases.

Spots “mean” measles in a different sense from the sense in which three

rings “mean” that the bus is full. In the measles case we cannot say, “Those

spots mean measles, but he hasn’t got measles,” but in the three rings case

we can say, “Those three rings meant the bus is full, but the bus is not full.”

The bus conductor may have made a mistake when he thought the bus was

full, but the spots can’t make mistakes. The fact that the conductor meant itdoes not entail that it is true. The Cockney made his colorful statement and

meant to compliment his wife, but this does not entail that he really does

find his wife indispensable—he might be quite able to survive without her.

Just because someone makes an assertion and means something by it does

not imply that the assertion is true.

Another difference is that in the case of non-natural meaning we can

substitute an expression in quotation marks after “means” but we can’t do

this for natural meaning. We can say that the conductor meant, “The bus is

full” by his three rings, but we can’t say that the spots meant, “The patient

has measles.” What this really comes to is that three rings are synonymous

with the sentence “The bus is full,” but spots are not synonymous with the

sentence “The patient has measles”—they are not synonymous with any-

thing, even though they naturally mean something. Spots are not words.

A third difference is that in the case of reports of natural meaning, there

is no indication that an agent or person is involved in the fact of meaning.

When spots mean measles, there is no agent or person who is meaning any-

thing. But in the case of non-natural meaning, there is always the implica-

tion of an agent or a person involved. When there is non-natural meaning,

there is always an agent of that meaning—a bus conductor or an uxorious

Cockney, say. People mean things in the non-natural sense, but objects or

events mean things in the natural sense. This is connected to the further

point that in non-natural cases we can speak of “what was meant” (by an

agent), but we cannot speak this way about natural meaning—we cannotrefer to “what was meant” by the spots.

Grice’s terminology here is not altogether perfect, though it has become

entrenched. He talks of “non-natural meaning,” but there is really nothing




non-natural about it. We typically use the word “non-natural” to indicate

things that are supernatural or lie outside of nature. But Grice does not

mean to suggest that idea when he speaks of non-natural meaning. He isnot using the word “non-natural” as G. E. Moore used it when describing

the property of being good as non-natural, that is, not part of the natural

causal order. It is not a very descriptive label and it has some misleading

connotations. We could instead call it “semantic meaning” or “speaker

meaning” or “agent meaning.” At any rate, it is good to keep these alterna-

tive labels in mind when using the phrase “non-natural meaning.” It is, in

fact, not easy to come up with the perfect nomenclature for the distinction

Grice is making, despite the clarity of the distinction.

10.3 What Is Speaker Meaning?

The question is what constitutes this so-called non-natural meaning. Here

Grice is looking for the necessary and sufficient conditions for instances

of non-natural meaning—that is, he seeks an analysis of the notion. His

procedure then is to try out various analyses and see if there are any coun-

terexamples. He begins by examining a suggestion of C. L. Stevenson’s that

he calls the causal theory of meaning. This theory is very tempting because

it reflects some obvious facts about language. Let us take an ordinary asser-

tion, such as my asserting to you, “Nadal won the French Open in 2012.”

By making this assertion I meant precisely that Nadal won the French Open

in 2012. Why does the speech act mean that? Well, here are two obvious

facts: my utterance of that sentence tends to produce in my audience the

belief that Nadal won the French Open in 2012, and the utterance itself

was produced by my having that same belief. The utterance expresses my

belief and it induces the same belief in you. I have a tendency to say it

given my beliefs, and you have the tendency to believe it because you hear

me say it. The assertion has these causes and effects, and they seem bound

up with what I meant. We might then propose the following definition of

non-natural speaker meaning: X means that p by uttering s if and only if X ’s

uttering s is caused by his belief that p and his uttering s causes in his audi-

ence the belief that p. Less formally, you mean that p by an action if andonly if that action causes observers of the action to believe that p.

Grice gives a counterexample to this analysis, questioning its sufficiency.

He describes a man putting on a tailcoat when he is about to go to a dance.



196 Chapter 10

This causes an observer to believe that the man is about to go to a dance.

The observer believes this because putting on a tailcoat is very good evi-

dence that someone is going to a dance. The tailcoat wearer also believes heis going to a dance. Given the causal theory of meaning, we should be able

to conclude that putting on the tailcoat meant that he was going to a dance.

Indeed, we should be able to conclude that in putting on the tailcoat, the

wearer meant that he was going to a dance. We should be able to report that

“what was meant” in performing the action was that the agent was going

to a dance. Grice’s point is that no such thing was meant. The agent did

not mean anything by his action—he was just preparing for the dance. His

action was not any kind of assertion, not a case of speaker meaning. He wasnot trying to communicate anything. So inducing beliefs in others by one’s

actions is not sufficient for those actions to be cases of non-natural mean-

ing. This is really quite obvious, because most of your actions are not cases

of meaning anything to anybody, even though observers will form beliefs

about you given your actions. I may comb my hair to keep it neat, and you

may be caused to believe that I am trying to keep my hair neat by observ-

ing me comb it, but my action of combing was not a case of my meaning

anything to anybody—I wasn’t trying to tell you anything. These kinds of

examples put paid to the causal theory of speaker meaning.

Another type of case Grice gives that is equally devastating to the causal

theory involves the sentence “Jones is an athlete.” What I meant by that

utterance is that Jones is an athlete. Hearing this, my audience might form

the belief that Jones must be tall, because athletes usually are—and maybe

it is true that Jones is tall, and that I believe it. But did I mean that Jones is

tall when I said, “Jones is an athlete”? No, I did not. The utterance of “Jones

is an athlete” has a tendency to induce the belief that Jones is tall, but it

does not mean that. Again, this point is obvious and generalizes. Whenever

I utter a sentence of English, my utterance has a tendency to induce the

belief that I am speaking English, but I certainly do not mean that I am

speaking English whenever I open my mouth to speak in that language.

The utterance may also cause the belief in my audience that I am alive, but

again this is not something I can be said to mean. If this were sufficient for

speaker meaning, every time I speak I would mean hugely many things—allthe things that people might believe who hear me speak. The conditions

proposed by the causal theory are hopelessly weak.




Grice now switches to a different kind of theory. Instead of using the

idea of a causal tendency to induce a belief in an audience, this new theory

brings in the idea of intention—specifically, the intention to produce abelief in the audience. So the speaker means something by an action if he

intends to produce a certain psychological effect. This intention was lack-

ing in the tailcoat and “athlete” examples. To mean something you have to

intend to get a belief across to an audience, not merely get a belief across in

any old way. If I assert that p, I intend to get you to believe that p by making

that utterance. This sounds on the right track, because in meaning some-

thing I surely do intend to have a certain effect on my audience.

But Grice produces a counterexample to this analysis: the handkerchiefcase. I leave B’s handkerchief at the scene of a murder in order to induce the

detective to believe that B is the murderer. Thus I intend to produce in the

detective the belief that B committed the murder by leaving the incrimi-

nating handkerchief around the murder scene. I might thereby fulfill my

intention of producing in the detective the belief that B is the murderer.

But did I, by this action, mean that B is the murderer? Clearly not: all I did

was intentionally fabricate evidence from which the detective inferred that

B is the murderer.

What is intuitively missing in this example is that the detective does

not know that I intended him to form the belief by leaving the handker-

chief there. I concealed my intention completely by secretly depositing the

handkerchief. If he knew that I had left it there, he would not have formed

the belief that B is the murderer, because he would know that I was trying

to frame B. So let us add the condition that the agent must not only intend

to produce the belief but must also intend that the audience should recog-

nize this intention. Now we have an extra intention—the intention that

the first intention should be out in the open. The agent intends to produce

the belief in the audience and he intends the audience to realize that he has

the former intention. Thus there is a double intention, where the second

intention refers back to the first. We might call this the transparency condi-

tion. The agent’s belief-inducing intentions must be intentionally transpar-

ent to the audience, if the agent is to mean something by the action.

This is beginning to sound pretty good, but Grice is not through yetwith his counterexamples. He gives the grisly example of Herod presenting

Salome with the head of John the Baptist on a charger. He thereby intends

her to form the belief that John the Baptist is dead, but he also intends that



198 Chapter 10

Salome should recognize this intention. He does not attempt to conceal his

intention—no doubt because he is not afraid of her knowing that he has

that intention. The severed head is ample evidence that John the Baptist isdead and Herod transparently proffers this evidence to Salome, so that all

his intentions are out in the open. But, Grice insists, this act by Herod is not

a case of meaning that John the Baptist is dead. It is not a way of telling her

he is dead. So we have still not captured what is distinctive of non-natural

speaker meaning. It is not like saying , “John the Baptist is dead.”

Now we reach the crux of Grice’s argument, which is contained in the

following passage:

The way out is perhaps as follows. Compare the following two cases: (1) I show Mr.

X a photograph of Mr. Y displaying undue familiarity to Mrs. X. (2) I draw a picture

of Mr. Y behaving in this manner and show it to Mr. X . I find that I want to deny

that in (1) the photograph (or my showing it to Mr. X ) meantNN anything at all;

while I want to assert that in (2) the picture (or my drawing and showing it) meantNN

something (that Mr. Y had been unduly familiar), or at least that I had meantNN by it

that Mr. Y had been unduly familiar. What is the difference between the two cases?

Surely that in case (1) Mr. X ’s recognition of my intention to make him believe that

there is something between Mr. Y and Mrs. X is (more or less) irrelevant to the pro-

duction of this effect by the photograph. Mr. X would be led by the photograph at

least to suspect Mrs. X even if instead of showing it to him I had left it in his room

by accident; and I (the photograph shower) would not be unaware of this. But it will

make a difference to the effect of my picture on Mr. X whether or not he takes me to

be intending to inform him (make him believe something) about Mrs. X , and not to

be just doodling or trying to produce a work of art.2

The distinction to which Grice is drawing attention is clear enough (despite

his rather convoluted grammar). In the photograph case, the audience’s

reason for forming the infidelity belief is the evidence contained in the

photograph itself—it makes no difference how Mr. X views my intention

in showing it to him. He might find the photograph in his wife’s closet, so

there is no act of showing at all. But in the drawing case, Mr. X ’s reason for

forming the infidelity belief is not the drawing itself—the drawing by itself

is not good evidence for that belief—but rather the reason is that Mr. X

infers that I intend him to form the infidelity belief. In this case, if we ask

Mr. X why he has that belief he will say that it is because he knows that I

intended him to form the belief—and he goes by my intention because he

knows me to be trustworthy about such matters. Nothing like this holds in

the photograph case: here his knowledge of my communicative intentions




plays no role in his belief formation. What I intend in the drawing case

is that Mr. X should form the belief because of my intention to get him to

believe it, and not because my drawing is somehow solid gold evidencefor the belief. The drawing is relevant only because it is evidence of my

communicative intention, but this is not so for the photograph. It is the

audience’s recognition of my belief-inducing intention that supplies him

with a reason for forming that belief, not any independently plausible evi-

dence. In short, his sole reason for forming the belief is that he sees that I

am intending him to form it. Thus for the agent to mean something, it is

necessary that he should intend the audience to form a belief by means of

the recognition that the agent has such an intention. The agent intendsthe audience to engage in a piece of reasoning of the following form: the

speaker intends me to form the belief that p, therefore I will form the belief

that p. This is quite unlike the photograph and severed head cases, because

in these cases the audience reasons as follows: I have solid evidence that p

based on a photograph/severed head; therefore I will believe that p.

10.4 Consequences and Criticisms

So now we know what speaker meaning is. It is intending people to form

beliefs based on a recognition that that is what you intend. What can we

do with this information? We can use it to define sentence meaning. A

sentence s means that p if and only if people regularly use s to mean that

p, where a speaker’s meaning that p is equated with the intention to induce

a belief in an audience by means of the audience’s recognition of that

intention. No doubt we will have to say more about the notion of “regular

use,” but the thrust is clear: a sentence means what it does because people

utter sentences with the kinds of intentions specified by Grice. Meaning

something non-naturally is a matter of performing actions with Gricean

intentions, and semantic meaning has its origin in speaker meaning. So

semantics reduces in the end to intentions—that is, to a certain kind of psy-

chological state. Languages like English exist because humans are capable

of Gricean communicative intentions. It is in virtue of these intentions that

words have meaning.It is worth spelling out a bit further Grice’s picture of language and its rai-

son d’être. We all have a great many beliefs about the world, often acquired

by observation. Imagine a time before language evolved, but when people



200 Chapter 10

still had a stock of beliefs. As a social species, we might wish to induce

some of our beliefs in others—that is, we want to share our knowledge (this

can be useful in child rearing, among other things). How do we set aboutdoing this? The obvious way would be to present other people with the

evidence that led you to form your beliefs and let them come to their own

conclusions. If you want others to know where the succulent fruit is, you

might take them there so that they can see it for themselves. Alternatively,

you might preserve the evidence you had and bring it with you to show

others—so you bring them a piece of fruit as evidence that you know the

location of fruit and then have them follow you. But this is not always

feasible, because evidence is often perishable and nonportable. You hadthe evidence but you can’t present it to others to induce the same belief in

them. So you have a problem of belief transmission: how do you get them

to share your belief? The only apparent solution is that you have to present

them with evidence that you have the belief in question, and then rely on

their ability to reason that if you believe it, there must be a reason to believe

it. In other words, their reason to believe that p is that you believe that p. Of

course, that wasn’t your reason—you had solid evidence, but the evidence

has long since perished. So you need to intend to produce a belief in others

by getting them to recognize that you have the belief yourself, so that they

can reason that if you believe it, that is a reason for them to believe it.

That is, you need Gricean intentions if you are to solve the problem

of perishable, nonportable evidence in belief transmission. Since Gricean

intentions constitute meaningful language, you need to invent language

to fill the evidential gap. So language exists because evidence vanishes or is

unobtainable for other reasons. Your beliefs can persist through time and

space, even though the evidence on which they are based is confined to a

particular time and place. So you can use the existence of your beliefs to

persuade others to believe as you do. When you do that, the stage is set

for speaker meaning and hence language. Thus language exists to let peo-

ple know what we believe so that they can form the same beliefs. Gricean

intentions are substitutes for actual hard evidence. They allow us to trans-

mit our beliefs by testimony , instead of by lugging evidential stuff around.

On occasion, our audience may refuse to form the belief we want themto, perhaps distrusting our own belief-forming powers; then we might say,

“You don’t believe me, but let me show you this”—and then we whip out

the clinching piece of hard evidence. Sentences, on this conception, are




really evidence surrogates—what we resort to when we can’t just point to

the facts or produce the smoking gun. Sentences take up the evidential

slack. This is the lesson buried in Grice’s account of speaker meaning: youdon’t have the photograph, so you produce the drawing, intending the

audience to infer the belief from the fact that you have the belief.

What objections might be raised against Grice’s account of meaning?

The actual analysis of speaker meaning looks pretty strong, so it is hard

to object to that. But there are questions about the precise philosophical

significance of his analysis. If we are to provide an explanation of sentence

meaning in terms of speaker meaning, it must be that speaker meaning

does not presuppose sentence meaning. Since speaker meaning consists ina complex array of intentions and beliefs, these intentions and beliefs must

not presuppose sentence meaning. That is, the intentions and beliefs must

not be linguistic in character. But two sorts of argument suggest that they

do build in sentence meaning. One line of thought is that it is not possible

to possess Gricean intentions without already being a language user: the

intentions must be formulated in the very language the speaker is using.

Thus, when I utter, “Snow is white” with Gricean intentions, I must be

thinking something like, “I intend to produce the belief that snow is white

by means of my audience’s recognition of my intention.” But this is itself

an English sentence, so my intention presupposes the notion of sentence

meaning. In other words, if thoughts are inherently expressed in language,

they cannot be used to explain language.

The natural reply to this is that thoughts are not inherently expressed

in language. There can be thought without language. Animals have inten-

tions and beliefs, but they do not speak a language. Human infants have

thoughts before they acquire their native tongue. So thought itself does not

presuppose mastery of a language. Moreover, people who speak different

languages can have the same thoughts, even though their sentences are

not the same—so there must be a psychological level that is independent

of particular spoken languages. Perceptual states are surely not inseparable

from spoken language, so why should thoughts be? I surely don’t see in

English, so why should my thoughts have English written into their iden-

tity? I can express my thoughts to others in English, but they are not them-selves English sentences running through my mind. I could have had the

same thoughts and yet never have learned English—I might have been a

French speaker, say.



202 Chapter 10

Fair enough, it may be said—English is not essentially the medium of

my thoughts, even though I am an English speaker. But might my thoughts

not have a more subtle connection to language—what about the idea of alanguage of thought ? True, I do not think in English, but my thoughts must

exist in some sort of symbolic medium; and this medium must have the

characteristics of a language—combinatorial, finitely based, recursive, ref-

erential. Aren’t my concepts essentially symbolic entities that join together

to make thoughts? Thus the brain has a language of its own in which beliefs

and intentions are encoded. This is not a conventional natural language

but a universal, species-wide language that the brain employs to carry out

cognitive operations. When I think that snow is white, my brain activatesits special words for snow and whiteness, maybe in the form of a binary

code that neurons can embody. These brain symbols will have reference,

and maybe even sense, and they can combine to produce strings that have

truth conditions. So having a mind depends on having a brain language.

But then, sentence meaning is basic after all, because Gricean intentions are

grounded in brain sentence meaning. Natural language sentence meaning

may be explicable in terms of psychological states, but psychological states

themselves must be explained in terms of a universal language of thought—

so at the bottom of the whole thing, we find sentence meaning staring up

at us. There is then still the question of what gives these brain sentences

their meaning—because it can’t be that they are uttered with certain kinds

of intentions. How do brain symbols come to mean what they do? That

question remains unanswered.

At this point we stray into the area of philosophy of mind. We are now

inquiring into the semantics of thought. That is a subject for another kind

of book. What we can say here is that these questions are not going to

be easy. But no matter how the deep questions are resolved, Grice does at

least provide an illuminating and plausible account of speaker meaning. Its

precise significance for the general nature of meaning, however, remains

debatable.



Appendix: Kripke’s Puzzle about Belief

Finally, let us consider Kripke’s paper “A Puzzle about Belief,”1 mainly

because of its intrinsic interest, impact, and connection to issues already

discussed. It is also fun to think about. I confine this topic to an appendix

because the issue is more about the nature of belief than the nature of

language, and because Kripke is not offering any theory but presenting a

puzzle. I am going to describe my own version of the puzzle, which I think

reveals its essential structure, without irrelevant distractions. Kripke’s ver-

sion involves a bilingual speaker, Pierre, who speaks only French at one

time, and on the basis of his verbal behavior we ascribe to him the belief

that London is pretty. He assents to “Londres est jolie” on the strength of

what he has read about London in rosy travel books. Then he comes to

London and learns English, living in a seedy part of the city. He now assents

to “London is not pretty.” Of course, he doesn’t realize that the place he

is living in is actually the reference of the French word “Londres.” On the

basis of his assent behavior, we now ascribe to him the belief that London is

not pretty. So we have ascribed contradictory beliefs to him. Yet he is guiltyof no logical blunder. He has manifested no irrationality. His situation is

perfectly intelligible.

I will now describe a case that exhibits much the same structure but

without the reliance on two different languages (Kripke himself recognizes

that his puzzling cases do not require two different languages). A psycholo-

gist is conducting experiments on the interpretation of faces. She asks the

subjects to respond to photographs of faces according to whether the sub-

ject finds the person photographed trustworthy or untrustworthy, going

by their facial expression. Then the psychologist tells the subjects that

although the pictures may appear to be of the same person they are all in

fact pictures of different people. In reality, however, the pictures are all of



204 Appendix

the same person. And so the subject’s beliefs are all about a single person

throughout the experiment, though he thinks they are about different peo-

ple. Let us suppose that the subject’s response takes the form of checking thesentence “That person is trustworthy” or “That person is untrustworthy.”

The experiment is run and the data show that subjects systematically vary

in their responses according to facial expression. Logically, this example is

just like Kripke’s Pierre example: “Londres” and “London” refer to the same

city, but Pierre does not realize this. He may in fact believe explicitly that

they are not identical. In the experiment, the subject keeps seeing a picture

of the same person but doesn’t realize it and even disbelieves it.

To begin the experiment, the psychologist shows the subject the firstpicture and asks him if that person is trustworthy. Based on the expression

of the individual’s face in the picture, the subject says yes. The psycholo-

gist then shows the subject another picture, and based on that person’s

expression, the subject responds that that person is not trustworthy. Keep

in mind that the subject thinks there is a different person in each picture.

The experimenter goes on to show the subject ten different pictures and

based on the subject’s judgments ascribes beliefs to the subject. Using the

normal method of belief ascription, the experimenter would ascribe con-

tradictory beliefs to the subject in exactly the way we would in Kripke’s

Pierre example. The subject believes that that person is trustworthy and

that that person is untrustworthy—yet these are same person. Suppose the

experimenter says to the subject, “Just for convenience I’m going to call

all these different people photographed ‘Albert,’ so I want you to respond

to the sentence ‘Albert is trustworthy.’” In fact, the single person in all the

pictures is named Albert. Then after the first photograph is presented, the

experimenter asks, “Do you think that Albert is trustworthy?” The subject

responds affirmatively, thus showing that he believes that Albert is trust-

worthy. Now, on the second trial, the subject responds negatively, thus

showing that he believes that Albert is untrustworthy. Already in two pre-

sentations of the pictures the subject has exhibited contradictory beliefs:

he believes that Albert is trustworthy and he believes that Albert is untrust-

worthy. The subject could continue to form contradictory beliefs about the

same person throughout the experiment. What is going on, intuitively, isthat the subject does not realize it is the same person in each photograph,

so he feels free to form different beliefs from trial to trial. However, the

experimenter knows that the subject is forming beliefs about the same



Appendix 205

person. This is a perfectly intelligible situation, just as Kripke’s Pierre case

is, and what makes it intelligible is that people can fail to realize they are

forming contrary beliefs about the same thing. It is not a given that per-ceived objects are the same—one might have false beliefs about this. Even if

the objects are presented in a qualitatively identical way, and in fact really

are the same object, a person can intelligibly suppose that two numerically

distinct objects are being presented. It is intelligible that someone might

take a single individual to be really twins, and hence freely form contradic-

tory beliefs about this single person.

We could imagine an alternative experiment where the experimenter

tells the subject that all the photographs are of the same person. Now con-sider what happens. The experimenter presents the first photograph and

asks if the person depicted (“Albert”) is trustworthy. The subject assents to

this proposition, thus showing that he believes that Albert is trustworthy.

Then she presents the second photograph and asks the same question. The

subject replies, “But I already told you I think Albert is trustworthy.” The

experimenter persists with her question, pointing out the extremely shifty

expression on the person’s face. She asks: “Are you so sure now that Albert

is trustworthy?” The subject may hesitate and then comment, “Perhaps

I should revise my belief about Albert—this expression could only come

from an untrustworthy man.” The subject, as we say, changes his mind,

forming a new belief and rejecting the old belief. He is rationally committed

to changing his earlier belief when he acquires good contrary evidence. It

would be highly irrational of him to persist with the first belief in the light

of acquiring the second. Why? Because he believes (truly) that the same

person is being presented, and it is irrational to attribute contrary predi-

cates to the same individual, given that you know it is the same individual.

This thought experiment is structurally the same as Kripke’s, but it is

more streamlined because it requires that we use only one language. We

also make explicit the subject’s beliefs about the identity of the things he

has beliefs about. In both cases, however, we wind up ascribing contradic-

tory beliefs to the subject.

We are now starting to see what these kinds of examples depend on.

Here is another example. Consider someone who has some eccentric meta-physical views about the world. He doesn’t think objects persist for more

than two seconds. He subscribes to the doctrine of repeat creationism, main-

taining that God actually re-creates the world every two seconds. But God



206 Appendix

makes it look as if there is seamless continuity to the human senses. In

fact, he thinks, God annihilates the particles that compose objects and cre-

ates new particlesab initio

every two seconds. He is omnipotent, after all,and he likes to keep himself busy. (Note: we are supposing that this meta-

physical system is false.) In addition to this belief, our eccentric metaphy-

sician believes that objects change their natures in important ways every

two seconds—specifically, they become made up of different kinds of par-

ticles every two seconds. So suppose that at time t he assents to “This table

is made of electrons” but at time t plus two seconds he assents to “This

table is not made of electrons”—despite the fact that he refers to the same

table both times (contrary to his metaphysical beliefs). Does he not havecontradictory beliefs? Of course, he doesn’t think so, because he doesn’t

think the same table is referred to by the two demonstratives; but from

our point of view, we can see that he both believes that this table is made

of electrons and believes that this table is not made of electrons. We arrive

at these ascriptions of belief simply by taking seriously his assent to “This

table is made of electrons” at t and his assent to “This table is not made

of electrons” at time t plus two seconds. If we gave the table a name, say

“Bill,” then we could convict our metaphysician of believing both that Bill

is made of electrons and that Bill is not made of electrons. He sees no clash

in his beliefs, because he thinks they are about different objects; but we

know better, so we detect a clash—and we are right, because objects do per-

sist over time. It is rather as if Pierre were to assert outright that “Londres”

and “London” do not refer to the same city, were we to suggest to him that

perhaps they are the same. He has a false nonidentity belief, just like our

metaphysician.

Suppose you have used the name “Larry” to refer to someone of your

acquaintance, assuming (truly) that there is only one Larry you have been

calling by that name. Perhaps you have noted that Larry seems rather a

changeable kind of man. Then you come to the conclusion that there is

no single Larry—you have been calling two men by the same name. This

conclusion, however, is false. You might now start to feel a new libera-

tion in your assent to sentences containing “Larry,” because now you can

attribute his varying characteristics to two different men. But using ourusual practice of belief ascription, we find ourselves ascribing contradic-

tory beliefs to you, because you are in fact referring to the same man with

“Larry” though you think you are not. You might believe that both men are



Appendix 207

really called “Larry” because you have heard others refer to “them” by that

name, but there is nothing impossible there—different people often have

the same name. The problem is that you have a false identity belief con-cerning Larry—you believe that Larry1 and Larry2 (as you put it to yourself)

are not identical, when they are.

Here is one final example. Consider Peter, a born-and-bred Londoner.

Peter was raised in Hackney, a less than salubrious part of London. As a

result of his experiences in Hackney, he concludes (a little hastily) that Lon-

don is not genteel—he assents readily to “London is not genteel.” But then,

at age eighteen, he is kidnapped and taken to Hampstead, also a part of

London. Hampstead is so different from Hackney that Peter does not thinkhe is in the same city. He notices that people refer to the city of which

Hampstead is a part as “London,” but he assumes that this is just a case of

different places having the same name—a common occurrence, as he has

learned from geography classes. If you ask him what he thinks of the sen-

tence “London is genteel” after the move to Hampstead, he enthusiastically

assents to it. Of course, he thinks that this “London” refers to a different

city from the other “London.” Given our usual practices of belief ascription,

we would have to conclude that he believes both that London is not gen-

teel and that London is genteel. Certainly, his assent behavior in the two

places would warrant such ascriptions of belief considered separately—it is

only the fact that we can make both ascriptions that might make us hesi-

tate. The word “London” in his idiolect refers only to one city, which is why

we can make the contradictory ascriptions, but Peter does not realize this,

which is why he can so smoothly believe both things.

It should be clear that in none of these cases are we talking merely about

contradictions between de re beliefs. There is nothing puzzling or paradoxi-

cal about attributing to someone the belief of Harvey that he is shady and

also the belief of Harvey that he is not shady. You just need to observe Har-

vey acting suspiciously in one situation and observe him acting irreproach-

ably in another, without realizing you have observed the same individual

twice. In this kind of case, there is no de dicto attribution of the form “ X

believes both that Harvey is shady and that Harvey is not shady.” All we

have is the de re attribution “ X believes of Harvey that he is shady and alsoof Harvey that he is not shady.” Kripke’s examples involve contradictory

de dicto beliefs, not just contradictory de re beliefs. The latter are not puz-

zling at all. There is no suggestion in these cases that the subject believes



208 Appendix

contradictory propositions. But in the Kripke cases that is precisely the situa-

tion. The same is true of the further examples I have described.

Although we cannot resolve the paradox, we can at least examine howit arises—its inner logic. There are two kinds of case in which a person has

contradictory beliefs: there is the case where a person has contradictory

beliefs because he is irrational, and there is the case where a person has con-

tradictory beliefs without being irrational. What is the difference between

those two cases?

Suppose you ask someone, “Do you think that a is F ?” and he replies,

“Yes.” Now you ask, “Do you think that a is identical to b?” and he again

replies, “Yes.” Then you ask, “Do you think that b is F ?” He replies, “No.”This case is an example of outright irrationality, because it logically follows

from “a is F ” and “a is identical to b” that “b is F ” is true. This is a simple

consequence of Leibniz’s self-evident law of the indiscernibility of identicals,

that is, that if a is identical to b then anything true of a must be true of b.

If someone were to reply as just described, you would be within your rights

to protest, “Then you don’t really believe that a and b are identical!” But, of

course, it is not irrational to refuse to make the inference to “b is F ” from “a

is F ” if you do not believe “a is identical to b.” You simply lack the identity

premise that would make the inference go through. Indeed, it would be

irrational to make the inference without the identity premise. You are not

guilty of any irrationality if you refuse to infer that Phosphorus is a planet

just from the premise that Hesperus is a planet, but you are if you refuse

to make that inference given that premise and the premise that Hesperus is

identical to Phosphorus. These are just totally different kinds of psychologi-

cal situations, not to be confused or assimilated.

In Kripke’s example, Pierre does not believe the identity “Londres is

identical to London”—he will not assent to this sentence. The same is true

in all the cases I described. The subject lacks belief in a crucial identity

premise. So he is not being irrational; in fact, he is being perfectly rational.

There are cases of rationally contradictory belief —those in which the subject

believes that p and believes that not- p without violating any principle of

logical inference. These arise when the subject does not believe an identity

proposition that connects two names or demonstratives or descriptions.It is not irrational to have the beliefs that Pierre has, because he formed

them perfectly rationally. What would be irrational is to believe that Lon-

don is pretty and that London is not pretty while assenting to “Londres is



Appendix 209

identical to London.” That is, if we confronted Pierre’s assent to “London

is not pretty” and his assent to “Londres est jolie” with the information

that “Londres” refers to the same city as “London,” and he accepted thatidentity but refused to give up either assent, then he would be irrational. He

couldn’t rationally suppose that the place he called “Londres” is the same

as the place he calls “London,” and yet the former place is pretty and the

latter is not. Everything depends on his view of a certain identity question.

Pierre and his puzzling kindred spirits are really no more irrational than

someone with contradictory de re beliefs, that is, not irrational at all. It is

not irrational to believe of a that it is F and of a that it is not F , because

in such a case you don’t subscribe to an identity judgment regarding theobjects of your beliefs. You fail to realize that your beliefs are about the

same thing. You only lapse into irrationality if you accept the identity “a

is identical to b” but persist in assenting to “a is F ” and “b is not F .” In all

the puzzling Pierre-type cases we have described there is a crucial nonac-

ceptance of an identity statement—a true identity statement.

This is not meant to solve or remove Kripke’s puzzle, which does reveal

something strange in our normal practice of belief ascription, only to diag-

nose how it arises. We need to see clearly the difference between irratio-

nal contradictory beliefs and rational contradictory beliefs. That difference

turns on the role of identity judgments in the subject’s reasoning. What is

surprising is that a nonparadoxical rejection of a true identity statement can

lead so quickly to a puzzling assignment of contradictory beliefs, given that

we insist on sticking to our usual practice of belief ascription. Being logical

can lead to an appearance of illogicality. This appearance is the same as in

genuine irrationality. But the underlying state of mind is quite different.





Notes

1 Frege on Sense and Reference

1. Gottlob Frege, “On Sense and Reference,” in Philosophy of Language: The Central

Topics, ed. Susana Nuccetelli and Gary Seay (New York: Rowman & Littlefield, 2008),

113.

2. Ibid.

3. Ibid.

4. Ibid.

5. Ibid.

6. Ibid., 113–114.

7. Ibid., 114.

8. Ibid.

9. Ibid.

10. Ibid.

11. Ibid.

12. Ibid., 114–115.

13. Ibid., 115.

14. Ibid., 115–116.

15. Ibid., 116.

16. Ibid., 117.

17. Ibid.



212 Notes

2 Kripke on Names

1. Discussion in this chapter follows the excerpt from Saul Kripke’s Naming and

Necessity (Lecture II) in Philosophy of Language: The Central Topics, 128–146.


Topics, 126.

3 Russell on Definite Descriptions

1. Bertrand Russell, “Descriptions,” in Philosophy of Language: The Central Topics,

147.

2. Ibid., 148.

3. Ibid., 150–151.

4. Ibid., 153–154.

4 Donnellan’s Distinction

1. Keith Donnellan, “Reference and Definite Descriptions,” in Philosophy of Lan-

guage: The Central Topics, 157.

2. Ibid., 164.

3. Stephen Neale, Descriptions, excerpted in Philosophy of Language: The Central

Topics, 170.

5 Kaplan on Demonstratives

1. David Kaplan, “Demonstratives,” in Philosophy of Language: The Central Topics,181.

2. Ibid., 187.

3. Ibid.

6 Evans on Understanding Demonstratives

1. Gareth Evans, “Understanding Demonstratives,” in Philosophy of Language: The

Central Topics, 201.



Notes 213

7 Putnam on Semantic Externalism

1. Hilary Putnam, “Meaning and Reference,” in Philosophy of Language: The Central

Topics, 275.

8 Tarski’s Theory of Truth

1. Alfred Tarski, “The Semantic Conception of Truth,” in Philosophy of Language: The

Central Topics, 30.


Topics, 117.

3. Alfred Tarski, “The Semantic Conception of Truth,” 30–31.

4. Ibid., 32.

5. Ibid., 38.

6. Ibid.

7. Ibid.

9 Davidson’s Semantics for Natural Language

1. Donald Davidson, “Semantics for Natural Languages,” in Philosophy of Language:

The Central Topics, 58.

2. Ibid., 62.

3. Donald Davidson, “On Saying That,” in his Inquiries into Truth and Interpretation

(Oxford: Oxford University Press, 2001). .

4. Davidson, “Semantics for Natural Languages,” 61.

5. Davidson, “Radical Interpretation,” in Inquiries into Truth and Interpretation..

6. Davidson, “Semantics for Natural Languages,” 60.

10 Grice’s Theory of Speaker Meaning

1. Discussion in this chapter follows H. P. Grice’s paper “Meaning” in Philosophy of

Language: The Central Topics, 69–76.

2. Ibid., 72–73.



214 Notes

Appendix: Kripke’s Puzzle about Belief

1. The discussion in this appendix concerns an excerpt from Saul Kripke’s “A Puzzle

about Belief,” in Philosophy of Language: The Central Issues, 257–263.



Index

Agent meaning, 195. See also Meaning

Analytic–synthetic distinction, 4, 6–7,

40–41, 45, 46, 129, 188

A posteriori propositions, 6, 9, 40

Appearances, 6

A priori propositions, 4, 6, 7, 40–41, 45, 46

Aristotle and redundancy theory,

151–155

Aspect of object and sense, 14–16, 145

Atheism, 59

Attributive view of descriptions, 78–84

Axiomatized arithmetic, 51

Base axioms, 175, 178

Belief, Kripke’s puzzle about, 203, 209.

See also Kripke, Saul

Belief ascription, 203–207, 209

Biconditionals, 152–156, 188–189. See

also T-sentences

material, 189

strict, 189

Brain and language, 202

Causal chain theory, 48–49

Causal theory of meaning, 195, 196

Causal theory of perception, 47

Character, 107–108, 110–111, 117,

130–131, 140

content and, 107–108, 110, 111, 130,

131

context and, 107–108, 110

defined, 107

as function from context to content,

110

nature of, 107, 110

Character compositionality, 112

Character- vs. content-type meaning,

111. See also Character: content and

Charity, principle of, 184–185

Circumstances of evaluation. See Condi-

tions of evaluation

Cockney rhyming slang, 193–194

Cogito ergo sum (Descartes), 106

Coherence theory of truth, 148

Complete expressions, 31

Compositional idea, 168

Compositionality, semantic, 112

Compositional theory of meaning, 168

Compositional theory of truth condi-

tions, 168

Conditions of evaluation, 104–110

context of use and, 105–109

Conjunctions. See Sentence connectives

Content, 117

defined, 108

Content compositionality, 112

Content- vs. character-type meaning,

111. See also Character: content and

Context dependence, 102

definite descriptions and, 102, 108–

109, 111

of indexicals, 102, 106–109, 111, 142



216 Index

Context-dependent singular term, 179

Context independence, 106–107, 109–

111, 117, 133

Context of use, 105–109

Contingency, 40–45, 52–53, 98

Contingent facts, 41–42

Contingent properties, 43

Contingent sentences, 40, 97

Contradictions, 56–57

Contradictory beliefs, 203–209

Conversational implicature, 90, 91. See

also Implications and implicature

Coreference, 29, 30, 101, 123, 127

Correspondence theory of truth, 148,

149, 153–154

Creationism, repeat, 205–206

Davidson, Donald

applying Tarski’s theory to natural lan-

guages, 175–181

empirical truth theory and, 181–185

and the merits of Tarski’s theory as ap-

plied to meaning, 168–175

semantics for natural language,

165–168

theory of meaning, 175, 178–181

criticisms of, 185–190

Definite descriptions, 40, 41, 46, 95–96,

99

context dependence and, 102, 108–

109, 111

defined, 13

impure descriptions and, 53–54

vs. indefinite descriptions, 55–60, 65,

67

indexical expression and, 102, 109, 111

meaning and, 17, 61, 70, 77, 95, 98,

126–127, 143

mode of presentation and, 38

names and, 36, 43, 60, 62, 65 (see also

under Description theory of names)

proper names and, 13, 53–55, 60, 62,

69–71, 102, 119–120

as quantifiers, 55, 59, 60, 78, 96, 104

reference and, 13, 14, 16, 17, 20, 48,

51–52, 61, 67, 68, 79, 84–85, 91,

109, 117, 126–127, 134, 143

referential and attributive uses of,

78–84

rigid designation and, 42–44, 99

Russell on, 55–75, 104

semantic ambiguity and, 92

sense and, 13, 14, 17, 38, 117, 125

as singular terms, 55, 61–62, 68

theories of, 60–63, 77

“Definite Descriptions” (Russell), 60, 70.

See also Definite descriptions; Russell,

Bertrand

Demonstrative reference, 54

Demonstratives, 54, 88, 96, 142

defined, 101–102

Donnellan on, 80, 88

Evans on understanding, 127 (see also

Evans, Gareth)

Frege and, 127

identity propositions that connect,

208

indexicals and, 101–102

Kaplan on, 97 (see also Kaplan, David)

nature of, 71, 96

proper names and, 54, 71, 88

Putnam on, 142

Russell and, 70–71, 88, 96

Strawson on, 88

Denoting, 84–85

vs. referring, 84

Descriptions. See also Definite descrip-

tions; Indefinite descriptions

essential, 52–53

impure, 53–54

as reference dependent, 127 (see also

Reference dependency)

Russell’s theory of, 67–72, 116

semantics of, 90

theory of (see Description theory of

names)



Index 217

Description theory of (indexical) sense.

See Description theory of names: in-

dexicals and

Description theory of names, 53–54, 69

definite descriptions and, 36–39

demonstratives and, 54

Evans and, 117, 119, 127

Frege and, 35, 36, 39, 50, 116, 117,

119

historical perspective on, 35, 39

indexicals and, 102, 116–119, 127

John Stuart Mill and, 50, 125

Kaplan and, 102

Kripke’s critique of, 35, 36, 39–45, 52,

54, 102 (see also Kripke, Saul: critique

of Frege)

objections to, 49–52, 54

names and, 46, 48

objects and, 47

overview, 36, 38

Perry and, 117–119

reference and, 49, 50, 54, 102

Russell and, 69, 95, 116

social/socialized, 51

Designation, 159

Designation, mode of, 8, 10–11

Designation axioms, 162, 172, 176, 187

Direct designation, 70

Direct reference, 101–105

vs. rigid designation, 103–106

semantics of, 100, 101

Direct reference model, 101

Disappearance theory. See Redundancy

theory of truth

Disquotational theory of truth. See Re-

dundancy theory of truth

Donnellan, Keith

critique of Russell, 78, 91–94

distinction between denoting and re-

ferring, 84–85

evaluating, 87–90

distinction between referential and at-

tributive uses, 78–84

on implication and implicature, 90–91

Neale’s criticism of, 90–92

P. F. Strawson and, 82, 85–86, 90

Russell and, 78, 83–87

truth-value and, 78

on truth-value gaps, 85–87

Dual-aspect semantics, 111

Empirical truth theory, 181–185

Empty names, 125–126

Equality (mathematics), 3–4

Essential indexical, 119. See also “I”

Evans, Gareth

description theory of names and, 117,

119, 127

Frege and, 115, 117, 120–125, 128–

129, 131

on indexicals, 115–118, 131

John Perry and, 117–119

Kaplan and, 115, 131

on reference dependence, 123–127

on Russellian terms, 123, 127

on saying vs. showing, 122–123

on senses, 119–128

terminology, 116

theory of sense and reference for in-

dexicals, 119–122

thesis that names are Russellian, 126

on “today” and “yesterday,” 128–130

on understanding demonstratives, 127

view of names, 123–124, 126–128

Exaggeration, 93

Existential quantifiers, 64, 67, 188

Existential statements, 73–75

Extension, 135

character and, 110

context and, 110, 111

defined, 97

of Frege’s theory beyond singular

terms, 25–29

two types of dependence of, 106

Extensional contexts, 177

Extensional languages, 177



218 Index

Extensional sentences, 177

Extensional theory, Tarski’s, 177

Externalism, 133. See also Semantic

externalism

psychological, 145

Famous deeds theory, 45

Fictional characters, 61, 66

Fictional names, 124

First-level concepts, 59, 120

Formal correctness, 149, 150, 155, 157

Formal languages, 160–161, 179

Frege, Gottlob, 1–2. See also Kripke,

Saul: critique of Frege


36, 39, 50, 116, 117, 119

on ideas (in people’s minds) vs. sense

and reference (of words), 20–23

on identity, 3–9

on indexicals, 108–109, 115–118

leveled system, 22–23, 112

mock sense and, 124–125

on mode of presentation, 10–11

“On Sense and Reference,” 2, 3, 10,

24, 33, 35, 111

on opaque contexts, 32–33

on ordinary and extraordinary use,

18–20

on proper names, 13–14, 16, 22, 27,

31, 35, 55, 56, 60, 77, 119–120

on reference, 3–5, 8–23, 121

Russell and, 55, 127

on sense(s), 3–5, 8–16, 19–23, 120,

121, 123

theory of definite descriptions, 61

theory of meaning, 112

theory of sense and reference

compared with Russell’s theory, 77

extension beyond singular terms,

25–29

problems with, 23–24

theory of truth, 153, 154

on truth-value, 25–32, 78, 97, 120

Genuine names, 65, 69, 73, 86, 100

God, question of the existence of, 59

Grammar as logically misleading, 69

Grice, H. Paul, 90, 92

theory of speaker meaning, 191–199

consequences and criticisms,

199–202

Homophony, 173–175, 178

Hyperbole, 93

“I,” 105–107, 110, 111, 116–119, 133,

139, 143, 178

Ideal language, 24, 63, 65

Ideas, 20–23

compositional, 168

as objects of reference, 22

vs. senses, 22

Identicals, indiscernibility of, 208

Identity, 3–9

law of, 6–7

as a relation between names, 8

theories about personal, 44–45, 53

types of, 4

Identity statements, 3–5, 7–10, 123,

129, 209. See also Frege, Gottlob:

theory of sense and reference

Evans and, 123

false, 5, 71

Frege on, 11, 123

informative, 6, 13, 16, 38, 48, 53, 71

linguistic theory of the content of, 9

logically proper names and, 71, 77

as modes of presentation, 24

Russell and, 71

Wittgenstein on, 24

Implications and implicature, 90–94

Incomplete expressions, 31, 120

Indefinite descriptions, 67–68

vs. definite descriptions, 55–60, 65, 67

identity and, 63–65

as quantifiers, 58

Indeterminacy, radical, 184



Index 219

Indexical expression and definite de-

scriptions, 102, 109, 111

Indexical “I.” See also “I”

indispensability of, 119

Indexicality, the point of, 118–119

Indexicals

context dependence, 102, 106–109,

111, 142


116–119, 127

Evans on, 115–122, 131

Fregean theory of, 108–109, 115–118

Kaplan on, 100–105, 115

nature of, 101–102

possible worlds, meaning, and,

109–112

two principles of, 102–105

Indirect perspective, 33

Indirect reference, 19

Indirect sense, 19, 32–33

Indirect speech, 19–20

Indiscernibility of identicals, law of, 208

Information, 130–131

as an epistemic notion, 131

Intension, 109–111, 133, 134, 140

defined, 97

extension and, 97–100, 106, 110, 111,

133, 134, 140

Intensional contexts, 177

Intensional languages, 177

Intensional operators, 176, 177

Intension-based semantic theory, 111

Intention, 80–82, 197–202

Intentional operators, 160

Internalism, 133

Interpretation

radical, 184

substitutional, 64

Irony, 93

Kaplan, David, 97

on character and content, 107–108,

131, 140

on context of use and conditions of

evaluation, 104–109

and Fregean model, 101

on indexicals, 100–105, 115, 131, 133

and possible worlds and meaning,

109–112

on intension and extension, 98–100,

133

return to Russellian semantics, 100

on rigid designators, 103

on rigidity and direct reference,

104–105

on “today” and “yesterday,” 113

Kripke, Saul, 35–36, 99, 138

causal chain theory, 48–49

critique of description theory of

names, 35, 36, 39, 52, 54, 102

objections to, 49–52, 54

critique of Frege, 39–42

epistemic objections, 45–48

objections to, 49–51

essential descriptions, 52–53

impure descriptions, 53–54

modal arguments, 42–45, 48, 50, 53

Naming and Necessity , 104

“Naming and Necessity,” 35, 45

necessity of origin example, 103

puzzle/puzzling cases, 203, 209

on rigid designation, 42–45

social character of names, 51–52

Language, philosophy of. See also spe-

cific topics

questions in, 1

Language acquisition. See Learning

language

Language of thought, 202

Learning, 9–10

Learning language, 171, 173

Levels, Frege’s system of, 22–23, 112

Lexical ambiguity, 179. See also Seman-

tic ambiguity

Lexical meaning, types of, 111



220 Index

Linguistic deference, 51

Linguistic labor, universality of the divi-

sion of, 140–141

Logical form, 73, 154, 155, 178–179,

187–188

Logically proper names, 69–71, 77, 100

Material adequacy, 149–150, 157, 159

Material biconditional, 189

Meaning, 108, 181, 191. See also Speaker

meaning

as compositional, 165, 168, 171

definite descriptions and, 17, 61, 70,

77, 95, 98, 126–127, 143

definitions, 140

“meanings are not in the head,” 135–142

merits of Tarski’s theory as applied to,

168–175

as a social phenomenon, 141

theories of, 112, 181, 185, 188–189

(see also Causal theory of meaning)

Frege’s theory, 33, 112, 115, 147

Kaplan’s theory, 112

for natural languages, 147, 165, 175

(see also Davidson, Donald: theory

of meaning)

possible world semantics and, 111

properties/conditions for, 119, 168–

173, 175, 185, 187, 189, 190

referential, 61

Russell’s theory, 61, 69, 112

theories of truth and, 147, 165, 168–

173, 175, 185, 189

types of, 111, 193–199

“Meaning and Reference” (Putnam),

133

Meaning-ascription, 167

“Meaning” (Grice), 191, 193. See also

Grice, H. Paul

Meinong, Alexius, 59–63, 125

theory of definite descriptions, 61–63

Meinongian ontology, Russell’s rejec-

tion of, 65–67

Metaphors, 93

Mill, John Stuart, 50, 125

Mock sense, 124–125

Mode of designation, 8, 10–11, 80

Mode of presentation, 8, 226

of aspect, 15–16

as an aspect of an object, 14

definite descriptions and, 38

Frege and, 10–13, 15–16, 23–24, 33,

38, 120

identity statements and, 11, 12, 23–24

vs. mode of designation, 10

nature of, 10, 13, 14, 120

perception and, 11, 13

of reference, 12, 15

second- and third-order, 33

sense and, 12, 13, 15–16, 23, 33, 120

Names. See also Description theory of

names; Kripke, Saul

Evans’s view of, 126–128

genuine, 65, 69, 73, 86, 100

Russell on, 104

as Russellian, 126

social character of, 51–52

Naming and Necessity (Kripke), 35, 36,

39, 45, 104. See also Kripke, Saul

Natural kinds, 141

Natural kind terms, 133–134, 142

Natural language(s), 63, 160, 165, 202.

See also Davidson, Donald

applying Tarski’s theory to, 175–181

Natural vs. non-natural meaning,

193–195

Neale, Stephen

criticism of Donnellan, 90–92

Descriptions, 90

on Donnellan and Grice, 92

Necessity, 75, 99, 103, 174. See also

Naming and Necessity

Necessity of origin, 44, 103

Neurolinguistics. See Brain and language

Nonindicative sentences, 95, 177



Index 221

Non-natural meaning, 193–195

Numerical identity, 4

Object language and metalanguage,

155–15

Object(s). See also Aspect of object and

sense

Frege on, 26, 31

terms that introduce, 100

truth-value as, 25, 26, 28, 30, 31, 120

“On Sense and Reference” (Frege), 2, 3,

6, 10, 19, 24, 33, 35, 111, 153. See

also Frege, Gottlob

purpose, 10

Opaque contexts, 32–33, 176–177

Ordinary and extraordinary use, 18–20

Paratactic theory, 177

Partial definitions, 157–159, 162, 163

Peano’s axioms, 51–52

Perception

causal theory of, 47

and mode of presentation, 11, 13

Performatives, 177

Perry, John, 117–119

Personal identity theories, 44–45, 53

Perspective, 130–131. See also Character

on a perspective, 33

Positivists, 186–187

Possible worlds

analytic–synthetic distinction and,

40, 41 (see also Analytic–synthetic

distinction)

content and, 108

idea of, 42

indexicals, meaning, and, 109–112

intension, extension, and, 97–100,

106, 109–111, 134

ontology of, 174

“possibly” and, 188

rigid designation and, 42–45, 99,

104

truth-value and, 97, 98, 108, 189

Possible world semantics, 97–100,

104–106

Pragmatic meaning, 192

Pragmatics, 78, 84, 85, 88–92. See also

Donnellan, Keith

Pragmatic theory of truth, 148, 149

Predication, 65, 68

Presentation. See Mode of presentation

Pretend-sense theory, 125–126

Primary and secondary occurrences (of

descriptions), 74–75

“Proper knowledge,” 8–9

Proper name(s), 47, 119–120

ambiguity, 17, 35, 36, 71

aspects and, 13–14, 16

as complete expressions, 31

context dependence and, 102

definite descriptions and, 13, 53–55,

60, 62, 69–71, 102, 119–120

demonstratives and, 54, 71, 88

as descriptions, 54

as designating objects, 31

Frege on, 13–14, 16, 22, 27, 31, 35, 55,

56, 60, 77, 119–120

identity statements and, 71

Kripke and, 35, 36, 42

logically, 69–71, 77, 100

Meinong’s theory and, 60

multiple entities with the same, 17

ordinary names and, 13, 69–70,

119–120

reference and, 13–14, 16–18, 22, 27,

55, 56, 61, 69, 77, 119, 135, 142

as rigid designators, 42

Russell on, 13, 55, 56, 58, 60–62, 69–

73, 77, 100

Russell’s criteria for, 69, 71

sense and, 13, 16–18, 27, 35, 55

as singular terms, 55

Propositional function, 57–60

Propositions

contradictory, 207–208 (see also Con-

tradictions; Contradictory beliefs)



222 Index

Propositions (cont.)

the proposition expressed vs. the

proposition meant, 91–93 (see also

Implications and implicature)

singularity of (see Singular

propositions)

Putnam, Hilary

“meanings are not in the head,”

135–142

on semantic externalism, 133–146

Twin Earth thought experiment,

134–145

“Puzzle about Belief, A” (Kripke), 203.

See also Kripke, Saul

Qualitative identity, 4

Quantified proposition vs. particular

proposition, 82

Quantifier expressions, 59, 62, 92, 120

Quantifier phrases, 58–59, 80, 96

Quantifiers, 64

definite descriptions as, 55, 59, 60, 78,

96, 104

descriptions as, 73

existential, 64, 67, 188

indefinite descriptions as, 58

vs. names, 58

names as, 73

Quantifier view of descriptions, 78. See

also under Quantifiers

Quantifier words, 59, 89

Radical indeterminacy, 184

Radical interpretation, 184

Radical translation (thought experi-

ment), 183

Rationally contradictory beliefs, 208, 209

Real-world correlate, 130, 131

Recursive procedure, 162, 163

Redundancy theory of truth, 152–155

Aristotle and, 151–155

Reference, 12, 16–18, 181, 187. See also

specific topics


17, 20, 48, 51–52, 61, 67, 68, 79, 84–

85, 91, 109, 117, 126–127, 134, 143

levels of, 112

meaning as determining, 139–140

proper names and, 13–14, 16–18, 22,

27, 55, 56, 61, 69, 77, 119, 135, 142

sense and, 16–18, 20–23, 129, 166

of a sentence, Frege on, 25

theories of, 15, 61, 119–120, 187

truth-value and, 25–32, 97, 106, 120,

166

“Reference and Definite Descriptions”

(Donnellan), 84–85. See also Donnel-

lan, Keith

Reference dependency, 123–127

Referential theory of meaning, 61

Referential view of descriptions, 78–84

Referring vs. denoting, 84

Regular use, 199

Relational connections and semantic

notions, 159–160

Repeat creationism, doctrine of,

205–206

Representational entities, 15

Rigid descriptions, 45, 103, 104

Rigid designators

de facto, 104

de jure, 104

indexicals as, 102–103

vs. non-rigid designators, 42–44, 99,

103, 109–110

Rigidity/rigid designation, 42–45, 104

definite descriptions and, 42–44, 99

vs. direct reference, 103–106

possible worlds and, 42–45, 99, 104

Russell, Bertrand

on definite descriptions, 55–75, 104

demonstratives and, 70–71, 88, 96


95, 116

Donnellan and, 78, 83–87

Donnellan’s critique of, 78, 91–94



Index 223

Frege and, 61, 116

on indefinite descriptions and iden-

tity, 63–65

on indefinite descriptions as quanti-

fiers, 58–59

John Stuart Mill and, 50

Meinong and, 59–63

Russell’s rejection of Meinong’s on-

tology, 65–67

on names, 100, 104–105 (see also under

Proper name(s))

on primary and secondary occur-

rences, 74–75

Principia Mathematica, 63


on propositional function and in-

stance, 57–59

reference dependency and, 123

referential theory of meaning, 61, 69

theories of definite descriptions, 60–63

(see also Definite descriptions)

theory of descriptions, 67–72, 116

compared with Frege’s theory, 77

objections to, 94–96


theory of meaning, 112

truth-value and, 71–72, 77, 78, 94, 95

Russellian terms, 123

as Fregean, 127

that differ in their sense, 127

Satisfaction, 159–163

Satisfaction axioms, 162–163, 172, 176,

179–181, 183

Saying vs. showing, 122–123

Schematic letter, 155

Scientific sense of words, 124

Secondary occurrences. See Primary and

secondary occurrences

Second-level concepts, 59, 120

Self. See “I”

Semantic ambiguity, 16–17, 35–36, 65,

74, 75, 85, 89, 92, 93, 117

Semantic compositionality, types of,

112

Semantic conception of truth, 160

“Semantic Conception of Truth, The”

(Tarski), 147. See also Tarski, Alfred

Semantic denoting, 84–85

Semantic externalism, 142

Putnam on, 133–146

Semantic meaning, 192, 195. See also

Meaning

Semantics, 78. See also specific topics

of natural languages, 163 (see also Nat-

ural language)

Semantic theory, 119

Semantic value, 120

Sense, 108, 181. See also specific topics

aspect and, 14–16, 73

authentic non-nonsense sense and

specious phony sense, 124–125

conception of, 12–16


38, 117, 125

definition and meaning of the term,

14

Evans on, 119–128

Frege’s conception of, 12–16 (see also

under Frege, Gottlob)

Frege’s introduction of the term/con-

cept, 20

as a label, 12

proper names and, 13, 16–18, 27, 35,

55

as the route to reference, 166 (see also

Reference: sense and)

Sense-specifying reference assignment,

121

Sentence connectives, 89, 155, 161

Sentence meaning. See Semantic

meaning

Sentences and propositions, 2–3

Sentential functions, 161

Showing vs. saying, 122–123



224 Index

Simple object theory (identity state-

ments), 9

Singular propositions, 100, 102–105,

115

defined, 100

Singular terms, 30, 31, 33, 120, 154–

155, 176, 178–179

context-dependent, 179

coreferential, 29 (see also Coreference)

definite descriptions as, 55, 61–62

extension of Frege’s theory beyond,

25–29

Kaplan and, 100, 104

nature of, 25, 55, 126

Russell and, 55, 61–62, 68, 90

Skepticism, 94

Speaker meaning, 93, 191–193, 195. See

also Meaning

nature of, 192, 195–199

Speakers and sentences, 191–193

S-sentences, 162, 163

Statements, 3. See also Sentences and

propositions

Stimulus meaning, 183–184

Strawson, P. F.

critique of Russell, 94

on demonstratives, 88

Donnellan and, 82, 85–86, 90

“On Referring,” 77

referential view of descriptions,

78

truth-value and, 72, 77, 85– 86, 94

on truth-value gaps, 85–86

Subjective ideas, 22

Subjective View: Secondary Qualities and

Indexical Thoughts, The (McGinn),

116, 118–119

Subsentential expressions, 25, 32

Substitutional interpretation, 64

Synonymy, 16, 121

Syntactic ambiguity, 85, 179–180. See

also Semantic ambiguity

Syntactic theory, 169

Tarski, Alfred

criteria of acceptability, 149–151 (see

also Truth theory(ies): criteria of

acceptability)

object language, metalanguage, and,

155–158

theory of truth, 147–163, 165

Tarskian biconditionals, 152–156. See

also Biconditionals; T-sentences

Tautology, 6, 9, 71, 73. See also Ana-

lytic–synthetic distinction

Temporality. See “Today” and

“yesterday”

Testimony, 200

“The,” 55, 68. See also Definite

descriptions

Thought, language of, 202

“Thought, The” (Frege), 111

“Today” and “yesterday,” Evans on,

128–130

Translation, 165, 166

radical, 183

Translation manual, 169, 170

Transparency condition, 197

Truth, definitions of, 147, 149–151, 153,

155–161, 163, 165. See also Partial

definitions; Truth theory(ies)

Truth biconditionals. See Tarskian

biconditionals

Truth conditions, theory of, 168. See

also Truth theory(ies)

Truth function, 120

Truth theory(ies), 148–149, 158. See

also Redundancy theory of truth;

Semantic conception of truth; Tarski,

Alfred: theory of truth; Truth, defini-

tions of

criteria of acceptability, 149–151, 175,

188

empirical, 181–185

Truth-value, 75, 120, 166, 177

character, context, content, and, 105–

106, 108, 110–111



Index 225

concepts and, 120

contingent sentences and, 97

Davidson and, 189

Donnellan and, 78 (see also Truth-

value gaps)

Frege on, 25–32, 78, 97, 120

intensions, extensions, and, 97–99

Kaplan and, 98

Kripke and, 99

Documents

Colin McGinn-Philosophy of Language_ The Classics Explained-The MIT Press (2015).pdf