Key Contemporary Thinkers

Georgia Warnke, Gadamer: Hermeneutics, Tradition and Reason

Peter Burke, The Annales School

Jocelyn Dunphy, Paul Ricoeur: Hermeneutics, Philosophy and Social Theory

David Frisby, Walter Benjamin: An Introduction to his Social Theory

John Hall, Raymond Aron: A Study in French Intellectual Culture

Philip Hansen, Hannah Arendt: History, Politics and Citizenship

Adrian Hayes, Talcott Parsons and the Theory of Action

Douglas Kellner, Jean Baudrillard: From Marxism to Post-modernism and Beyond

Eileen Manion, Mary Daly: Philosophy, Theology and Radical Feminism

Michael Moriarty, Roland Barthes

William Outhwaite, Habermas

Philip Pettit, Rawls

Simon Schaffer, Kuhn

Geoff Stokes, Popper


Language, Experience and Reality

Christopher Hookway

Polity Press

Copyright © Christopher Hookway 1988

First published 1988 by Polity Press
in association with Basil Blackwell.

Editorial Office:
Polity Press, Dales Brewery, Gwydir Street,
Cambridge CB1 2LJ, UK

Basil Blackwell Ltd
108 Cowley Road, Oxford OX4 1JF, UK

All rights reserved. Except for the quotation of short passages for the purposes of criticism and review, no part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the publisher.

Except in the United States of America, this book is sold subject to the condition that it shall not, by way of trade or otherwise, be lent, resold, hired out, or otherwise circulated without the publisher’s prior consent in any form of binding or cover other than that in which it is published and without a similar condition including this condition being imposed on the subsequent purchaser.

British Library Cataloguing in Publication Data
Hookway, Christopher
Quine: language, experience and reality.
— (Key contemporary thinkers).
1. Quine, W. V.
I. Title    II. Series
191    B945.Q54

ISBN: 978-0-7456-6724-9 (Single-user ebook)

Material from Word and Object by W. V. O. Quine is reprinted by permission of MIT Press. Copyright © 1960 by the Massachusetts Institute of Technology.
Material from From a Logical Point of View and Theories and Things is reprinted by permission of the Harvard University Press.

Typeset in 11 on 12½ pt Garamond
by Columns of Reading

Printed and bound in Great Britain by Marston Lindsay Ross International Ltd, Oxfordshire



Note on References


Part I   The Evolution of Empiricism


1.1 Introduction

1.2 Meaning and Naming

1.3 Words and Sentences

1.4 Objects and Ontology

1.5 Russell’s Theory of Descriptions and Quine’s Criterion of Ontological Commitment


2.1 Meaning, Rationality and the A Priori

2.2 Classical A nalysis

2.3 Carnap’s Views

2.4 Holism

2.5 ‘Two Dogmas of Empiricism’: the Third and Fourth Milestones

2.6 Empiricism without the Dogmas: Moderate Holism

2.7 Understanding without Analyticity


3.1 Quine, Carnap and Philosophy

3.2 Quine and Pragmatism

3.3 Pragmatism and Realism

3.4 Realism and Naturalism

Part II   Logic and Reality


4.1 Science and Philosophy

4.2 Quine’s Physicalism: Some Questions

4.3 Expression and Description

4.4 Truth and Method

4.5 Physicalism

4.6 Reduction and Reality


5.1 Logic: Calculus and Categorial Framework

5.2 The Canonical Notation for Science

5.3 Extensionality

5.4 Formal Logic, Canonical Notation and Ordinary English


6.1 I ntensionality: Examples

6.2 Time and Tense

6.3 Objects and Events

6.4 Physics and Extensionality


7.1 Necessity and Extensionality

7.2 De dicto and de re

7.3 Tense Modality

7.4 Essentialism

7.5 Conclusion

Part III Mind and Meaning


8.1 The Thesis

8.2 The Arguments

8.3 Epistemological Underdetermination, Ontological Indeterminacy

8.4 Consequences: Mind and Meaning

8.5 Consequences: Reference and Truth


9.1 Introduction

9.2 Rabbits

9.3 Translation and Meaning

9.4 Psychological Theory and the Theory of Meaning

9.5 Physicalism and Psychological Explanation

9.6 Some Questions


10.1 Holism and the Theory of Meaning

10.2 Some Holistic Theses

10.3 Radical Interpretation

10.4 Normative Principles and the Holistic Character of Interpretation

10.5 Interpretation and Cognitive Science

10.6 Indeterminacy: Quine and Davidson

Part IV Knowledge and Reality


11.1 Epistemology: the Cartesian Paradigm

11.2 Positivism and Epistemology

11.3 Epistemology Naturalized

11.4 Sceptical Arguments

11.5 A Scientific Refutation of Science?

11.6 Conclusion


12.1 Introduction

12.2 Truth

12.3 Taking Our Own Physics Seriously

12.4 Physicalism

12.5 Empiricism

12.6 Conclusion







For my mother and father


This volume examines the views of the most influential American philosopher of the post-war period, Willard van Orman Quine. His views are interesting and important in their own right, but they are of value too in providing the background to much recent analytical philosophy. Many philosophers who do not agree with Quine consciously develop their views in response to his, and I hope that the book will help with understanding these developments. In line with the aims of the series, I have tried to make the book accessible to non-philosophers and to students. To this end, there is little discussion of technical issues in logic and the philosophy of mathematics, and I have tried to explain all of the logical notation that I have used.

Writing about a contemporary philosopher calls for a balance between exposition and critical evaluation. A compromise is needed between a careful exposition which risks suggesting that the subject cannot provide a clear statement of his own views, and an extended critical engagement which may leave the reader uncertain why the author thinks his subject’s views are important. Quine’s writings are not easily understood: he is a systematic philosopher, and the systematic underpinnings of his positions are not always apparent. Hence, the early chapters are weighted more heavily towards exegesis, and towards placing Quine’s views within this wider context, but the amount of critical evaluation grows as the volume proceeds.

I am grateful to Jonathan Dancy for comments upon an early draft, which led to many improvements of style and substance, to Harold Noonan, whose comments on part of the text saved me from several mistakes, and to Michael Bryon for helpful discussions of Quine and Carnap. My greatest debt is to my wife, Jo, who provided indispensable personal support during the writing, and technical support in coping with a recalcitrant word processor, and even found time to improve my English after reading the final draft.

C. H.

Note on References

Works cited in the text are listed in the References (pp. 221–4). This list is in two main parts, the first covering works by Quine and subdivided into ‘Books’, and ‘Articles which are not reprinted in any of the books’. In each case the list is given in date order of first publication, and an abbreviation is shown by which the work is referred to in the text as source for quotations. For ease of reference book initials are given in parentheses after first mention of articles that appear in the books.

The second part of the References, ‘Other References’, lists works by other writers. It is arranged alphabetically and, for authors with more than one entry, in date order of first publication; where applicable, however, page references in the text are to the subsequent edition cited. Works in this part are referred to in the text by author/date.


Quine was born in 1908. He studied as a graduate student at Harvard, and apart from short visits to Oxford, Paris and other centres of learning, he stayed there as a philosophy teacher until his retirement in the mid-1970s. No contemporary thinker can equal the influence he has upon recent analytical philosophy, through both his teaching and his extensive publications.

These publications include at least fifteen books together with numerous articles. Many of these, including most of the early ones, are concerned with formal logic. The works with the greatest philosophical impact are From a Logical Point of View, published in 1953, and Word and Object, from 1960. A number of monographs and collections have appeared since then, and his philosophical views have been clarified and developed in many ways. However, the core of his position is present in these relatively early works. The former is a collection of papers, including two classics, ‘On what there is’ and ‘Two dogmas of empiricism’. These contain trenchant criticisms of many of the assumptions of twentieth-century empiricism, and advocate what is described as a kind of pragmatism.

Word and Object is an extended treatise on philosophy of language. Many themes from the earlier book remain, but they are systematically developed and related to a naturalistic perspective which had not been evident in 1953. This naturalism involves stressing continuities between philosophy and empirical science, and approaching philosophical issues from the point of view of an austere, somewhat behaviourist standpoint. It leads to one of Quine’s most famous, and most controversial, doctrines, the indeterminacy of translation.

Particularly because he avowed ‘pragmatism’ in 1953, and because he is linked to John Dewey by his defence of ‘naturalism’, some people are tempted to view Quine as a distinctively American philosopher. He is seen as continuing the pragmatist tradition of Peirce, William James, Dewey, Mead and others. While he was influenced by his Harvard teacher C. I. Lewis, who belonged to that tradition, such an interpretation would be highly misleading. It encourages a distorted reading of the work of the earlier pragmatists, who would have found many of Quine’s views uncongenial, and, more important, it can prevent our appreciating Quine’s relations to the logical empiricist or logical positivist movement. While in his early twenties Quine had visited the Vienna Circle, the home of logical positivism, and this seems to have shaped his philosophical outlook. His views cannot be understood other than as a response to the positions defended by Rudolf Carnap and the other positivists. The force of his critical arguments is missed when it is not appreciated that he is arguing against the form of empiricism which he encountered in the work of Carnap. And the reader will fail to understand his positive views if it is not seen that he remained faithful to the underlying spirit of positivism.

The outlook of the Viennese positivists involved several related components. Most important was a commitment to scientism: scientific knowledge serves as a paradigm for all knowledge; and philosophy can be a respectable activity only if it can itself be pursued as a science. In most cases, this was coupled with the claim that all of the sciences could be unified into a single body of knowledge grounded in physics, and with the view that physics tells the whole story about the fundamental character of reality. This led to a repudiation of areas of discourse that did not meet scientific standards. Metaphysics, religious claims, ethical and aesthetic propositions were rejected as meaningless. These views were grounded semantically, employing the empiricist view that the meaning of a word or sentence somehow involved a connection with experience. If we do not know what experience would show that a proposition was true, we do not understand the proposition. Philosophers could employ logical analysis to clarify the meanings of troublesome words and sentences – or to reveal that they had no meaning.

In Two dogmas of empiricism’ (FLPV), Quine attacked the semantic doctrines of the positivists. He denied that we can talk sensibly about the links between particular propositions and experience, rejected the positivists’ ideal of philosophical analysis, challenged their reductionist assumptions, and insisted upon the holistic character of the relations between our beliefs and our experience. He concluded that philosophers’ use of the notion of meaning was indefensible, and thus challenged the whole idea of philosophical analysis.

However, his commitment to other positivist dogmas remained. He never rejected empiricism, and he continued to believe that philosophy must be scientific. This is the source of his naturalism; and his austere behaviourist approach reflects his continuing loyalty to the view that reality is a physical system. His greatest philosophical contribution has probably been to develop, in a consistent and rigorous fashion, the consequences of a set of assumptions whose appeal cannot be denied even by those philosophers who reject them. All our knowledge of external reality comes through the senses; the only real knowledge is scientific knowledge; and the universe is, fundamentally, a physical system. Above all, Quine is a systematic philosopher who has articulated this empiricist, physicalist vision of knowledge and reality with great clarity.

This volume is divided into four parts. The first of these examines the views defended in From a Logical Point of View, and introduces the sources of Quine’s naturalism. The second part explains the metaphysical and logical doctrines which determine the character of many of his views, and which come to the fore in Word and Object. We here consider his physicalism and his view that an adequate language for science is ‘extensional’. In the third part, we examine the indeterminacy of translation, and compare Quine’s views with those of a philosopher much influenced by him, Donald Davidson. Davidson exploits Quine’s insights about language while rejecting some of the underlying commitments which link him to the positivist movement. This enables us, in the final part, to begin to evaluate Quine’s physicalist naturalism and his empiricism. By the end of this part, we shall unravel some of the complexities of Quine’s position and see how it is possible to dissent from it.

Part I

The Evolution of Empiricism


Language and the


In ‘Five milestones of empiricism’, reprinted in Theories and Things, Quine describes ‘five points where empiricism has taken a turn for the better’ (TT, p. 67) since the seventeenth century. He sees his own work as the culmination of this process of improvement. Examining these five ‘milestones’ will enable us to introduce some of the doctrines for which he is best known, and will also help us to see how Quine himself views the historical context of his philosophical position. This will occupy us for the first three chapters.

We do not require a precise definition of ‘empiricism’: it is enough that empiricists take seriously the claims of the sciences to provide our best knowledge of reality, and hold that this knowledge is grounded in sensory experience. When we raise the philosophical question of how such knowledge of reality is possible at all, we tend to focus first upon questions of evidential support: how does experience enable us to sort our beliefs into those that are true and those that are false? But there is a prior question about how thoughts and utterances can be about the world at all: what is it for a sound, an inscription on paper or a blackboard, or a state of someone’s mind, to represent some external state of affairs? What is involved in understanding a thought or utterance, in knowing what it means? These questions raise a host of issues about representation, meaning and reference which have been fundamental for twentieth-century analytical philosophy.

It is an assumption of much twentieth-century philosophy that we naturally fall victim to certain deeply mistaken pictures of how thought and language relate to the world. They tend to be uncritically accepted, but seriously distort our philosophical thinking; indeed, these false pictures often give rise to apparent philosophical problems, which can be dismissed once the pictures that produce them are rejected. For many analytical philosophers, all of the traditional ‘problems of philosophy’ result from this kind of distortion. The ‘milestones’ to which Quine refers all involve developments in our philosophical understanding of representation: they promise philosophical enlightenment by overthrowing entrenched, but mistaken, conceptions of how thought and language work.

We can pass over the first milestone rapidly. It is ‘the shift of attention from ideas to words’; focusing the analysis of representation upon linguistic expressions or utterances rather than upon thoughts or ideas. The merit of this shift was that attention could turn from shadowy objects of introspection to more easily examined public representations. My concern in this chapter is with the second milestone, ‘the shift of semantic focus from terms to sentences’. This introduces some of the most important foundational doctrines for contemporary philosophy of language. Examining these will help us to explain Quine’s approach to issues of what he calls ‘ontology’, in his classic paper ‘On what there is’ (FLPV).


A natural starting point for an explanation of how language works is that words stand for things; we understand a word when we know what thing it stands for. Thus, I understand the word ‘London’ when I know which city it refers to or denotes, and I understand ‘Quine’ when I know which person it names. A sentence can then be looked on as a sequence or arrangement of words, and our understanding of the sentence is built out of our knowledge of what the words stand for. Finally, we can say that a sentence is true when the arrangement of words in the sentence corresponds, in some fashion, to the arrangement in reality of the things that those words stand for. This is only a vague sketch of a possible theory – the notion of arrangement conceals a host of problems – but it will do as a stalking horse for our present discussion; it cannot be denied that it has considerable initial plausibility. In this section, I shall introduce some problems faced by any theory of this general shape. We can then investigate how Quine’s second milestone enables us to move beyond this theory and respond to these problems.

By way of preparation, we must labour the obvious point that a language such as English contains expressions of different kinds. Consider the sentence:

Quine is American.

The name ‘Quine’ functions as a subject expression which purports to pick out a unique individual: we shall call it a singular term. ‘London’ is also a singular term, and it is clear too that a more complex phrase, a ‘definite description’ such as ‘The author of Word and Object’, can also be used to pick out a single individual. The expression ‘is American’ does not purport to pick out a single individual, but rather expresses a general characteristic which can be applied to many things: such expressions can be called predicates. The sentence inset above is formally analogous to

London is populous.

Each employs a singular term together with a predicate which is used to apply some characteristic to the individual that the singular term refers to. Using upper case letters ‘F’, ‘G’ etc. to mark the places occupied by predicates, and lower cases letters ‘a’, ‘b’, etc. to mark the places of singular terms, we can express this common form:

Fa (It is a logician’s convention that the predicate is written first.)

A sentence such as:

Brutus killed Caesar

contains two singular terms (two subject expressions), ‘Brutus’ and ‘Caesar’, together with a predicate expression that expresses a relation between two persons, that of killing. Using ‘R’ etc. to mark the places of relational predicates, we can express the form of this sentence:


The sentence

London is south of Birmingham

is also of this form: it concerns a relation between two things.

There is one other kind of expression to which I want to draw attention here. In a sentence like

It is not the case that Quine is German,

the expression ‘it is not the case that’ is attached to the complete sentence ‘Quine is German’. Similarly, two complete sentences are conjoined by ‘and’ in:

Quine is American and Frege is German.

Following logician’s practice, I shall use ‘~’ to express ‘it is not the case that’ (negation) and ‘&’ to express ‘and’ (conjunction). The forms of our last two sentences can be expressed:


Fa & Gb.

These expressions which attach to, or connect, complete sentences will be called connectives or operators. ‘Or’ (disjunction), formally expressed by ‘v’, functions analogously to ‘and’.

Thus, we have three kinds of expressions: singular terms, including names; predicates including relational expressions; and various operators. There is no suggestion that this exhausts the resources of a natural language, nor that it accounts for all occurrences of the expressions that we have mentioned. But it provides us with a useful account of a fragment of most natural languages, and contains just enough complexity to enable us to understand some important philosophical doctrines: we can formulate difficulties for our plausible account of language.

Let us begin by looking at how predicates work. It seems easy enough to find the things that singular terms stand for: ‘Quine’ stands for a man, ‘London’ for a city, and these are comfortably concrete observable objects. But what of an expression like ‘is red’, ‘is American’ or ‘killed’? These do not stand for concrete observable objects. ‘Red’ cannot stand for any particular red object for it could not then be used to say truly of any other object that it is red. The only candidate for the referent of ‘is red’ is that it stands for the attribute or general character of redness or of being-red. This does not seem to be a concrete or observable thing: I can see particular red things, but I cannot see the general character of redness. We seem to be committed by our account of representation to the view that there are such general characters, that we are aware of them, and that we only understand predicates by somehow associating them with such general objects.

A parallel problem arises from the use of operators or connectives: what do ‘not’ and ‘and’ stand for? There does not seem to be anything in our experience to serve as the meanings of these expressions, yet the theory of meaning under discussion requires that there be such objects and that understanding the expressions involves associating them with these objects. Once we extend the fragment of language with which we are dealing, it looks as if we shall be led into such absurdities as the claim that there is something, viz. nothing, which the expression ‘nothing’ stands for.

Finally, let us consider the expressions for which the referential theory of meaning seems best suited, names and other singular terms. We all of us, including Quine, presumably understand the two sentences below:

Hamlet killed Laertes

Pegasus was a winged horse

It does not matter for the present whether we think those sentences are true or false. It is enough that we can understand them, for they contain names – ‘Hamlet’, ‘Laertes’, and ‘Pegasus’ – for creatures from fiction or mythology. Although we can observe actors portraying Hamlet, we cannot see the Prince of Denmark himself. He stands in no causal relations to other concrete objects, and he has no location in space and time. Like Pegasus, he seems to be a non-existent object. If the views about names described above are correct, then our understanding of the names employed in these sentences shows that there are – and that we can talk about – real things which do not exist.

Even if this is accepted, there is scope for considerable disagreement about just what these names refer to. Some hold – implausibly – that Hamlet is an idea in someone’s mind; others claim that he is a merely possible object; others that he is a sui generis fictional entity. We do not need to get involved in these debates, since Quine does not consider that these examples raise a serious problem for the referential theory. Adopting a view that was anticipated by Frege, and has subsequently been developed in much more detail by John Searle and Gareth Evans, Quine suggests that talk of Pegasus, Hamlet and their properties is not serious factual discourse. We ‘frivolously’ pretend to make assertions about winged horses, and to talk about the goings-on in Elsinore, whenever our idiom deviates from talk of such concrete objects as texts and the inscriptions that they contain. Hence, I only pretend to use ‘Hamlet’ as an ordinary proper name, so my usage does not show that I recognize the reality of Hamlet (FLPV, p. 103).

However, as my last sentence indicates, I can use names for characters of fiction in what are plainly serious assertions. For example, I may truly say:

Hamlet did not really exist

Pegasus did not really exist

I understand those sentences and think that they are true. If I only understand the names they contain by knowing who or what the expressions name, then I must know that both ‘Hamlet’ and ‘Pegasus’ refer to something non-existent. So, Quine – and all of us – must agree that there are non-existent objects. Unless the names referred to things, the two sentences could not be true; and since the sentences are true, those referents must be non-existent. Hence, the theory of meaning we are considering suggests that there are non-existent objects.

These problems do not refute the theory of language from which we started, but they point towards philosophical problems that become very pressing once the theory is accepted. What are these objects? How do we know about them? What are their properties? It is easy to find the resulting view of the world very embarrassing. The world seems to contain far more objects than are explicitly discussed in the sciences. And our grasp of our familiar language appears to involve an acquaintance with objects which are not evident to our senses. For a philosopher who is sympathetic to empiricism, the burgeoning population of abstract objects is something to be avoided. Empiricists often prize their down-to-earth common sense, but the common-sense view of the world does not seem to find room for these curious abstractions.

A prejudice in favour of the concrete is commonly found among philosophers who take science seriously, or who believe that our knowledge derives primarily from the senses. They show no reluctance to admit that the world contains objects which are visible; and invisible objects like protons are accepted because they stand in causal relations to other objects and are causally implicated in our ways of coming to knowledge about them. But abstract objects – things like numbers, classes, attributes, non-existent or fictional entities, which do not enter into the causal structures studied by the physical sciences, and have no location in space and time – are treated with suspicion. That Quine shares this prejudice is evident from the first paragraph of a paper jointly written with Nelson Goodman in 1947, ‘Steps towards a constructive nominalism’. This reads:

We do not believe in abstract entities. No one supposes that abstract entities – classes, relations, properties, etc. – exist in space–time; but we mean more than this. We renounce them altogether.

(Goodman 1972, p. 173)

Although Quine subsequently came to acknowledge some abstract entities, and described this passage as expressing merely the hypothetical basis for the investigations occupying the rest of the paper, it is plain that he sympathizes with the mistrust of abstract entities here expressed.

In their 1947 paper, Quine and Goodman attribute some of this suspicion of the abstract to the belief that it is not possible to make sense of at least one sort of abstract object – the classes studied in set theory – without either running into contradiction or relying upon ad hoc and unnatural analyses. But, fundamentally, their refusal to admit the abstract objects with which mathematicians and others seem to deal ‘is based on a philosophical intuition that cannot be justified by appeal to anything more ultimate’ (p. 174).

In ‘On what there is’, Quine raises a question about the reality of attributes and considers it in relation to some of the assumptions about language that we are examining:

There are red houses, red roses, red sunsets; this much is prephilosophical common sense in which we all must agree. These houses, roses and sunsets, then, must have something in common; and this which they have in common is … the attribute of redness.

(FLPV, p. 10)

When he asks why we should admit that there is this abstract entity – the attribute, property, or ‘universal’, redness – the arguments he considers have the following basic strategy. When I make an assertion such as

The rose is red,

I employ a predicate expression – ‘red’ or ‘is red’. This expression is plainly meaningful; I understand it. If an expression is meaningful, then there is something it stands for, it has a meaning. The meaning of our predicate is this abstract entity, the attribute of redness. By granting that general terms are meaningful, we admit the reality of some abstract entities – their meanings. Universals are the meanings of general terms. If the referential theory of meaning is adopted, this seems hard to resist.

Thus, each of the three classes of expressions we have considered presents philosophical problems if our natural theory of meaning is adopted. At different times, some philosophers have been prepared to bite the bullet, accepting that there are far more things around than the sciences would have us believe. In different ways, around the turn of the century, Moore, Russell and Meinong were prepared to take a strongly realist view of many of these strange objects. The problems they raise are primarily of two kinds. The first, already alluded to, is epistemological: in order to account for our understanding of language, we must show that we are in cognitive contact with these things. It seems easiest to do that by positing a faculty for intellectual acquaintance with them which will grate with the prejudices of an empiricist.

The second problem is metaphysical: it is often claimed that we understand the nature of objects of a particular kind, and we can talk about them intelligibly, only when we can understand identity statements involving terms referring to the object. We may say such things as

There is a number that is prime

There is number which is the sum of two and seven

But we do not know what numbers are until, as well as using numerals to refer to them, we understand what is involved in saying

Nine is the same number as the sum of two and seven.

We know when two singular terms stand for the same number. It is a condition of our treating persons as objects that we know how to answer questions about whether two terms refer to the same person, and so on. Hence, one basis for questioning the reality of abstract objects concerns whether a clear sense has been attached to the appropriate statements of identity or sameness. (See FLPV, p. 4, where Quine uses just such an argument against the claim that there are possible objects – recall the claim in the joint paper with Goodman that reference to abstract entities is likely to bring paradox and unclarity with it.)


The second milestone of empiricism was a doctrine introduced, and used to great effect, by Bentham; it was also embraced by Frege and Russell (TT, pp. 68–70). It involves a ‘shift of attention from words to sentences’: the primary vehicle of meaning is seen no longer as the word but as the sentence’ (TT, pp. 68–9); ‘the meanings of words are abstractions from the truth conditions of sentences that contain them’ (TT, p. 69).

Verbal behaviour primarily involves uttering sentences. These are the shortest units of language which can actually be used to perform linguistic actions. Unless a specific conversational context is provided, we can make no sense of the utterance of a single word. Our grasp of the meaning of a word enables us to use and understand the sentences in which it occurs. It is a natural corollary of this that an account of how language works is adequate if it makes sense of what we do with sentences: words are best viewed simply as recurrent features of sentences. Focusing on individual words in abstraction from a wider sentential context can then appear as a source of philosophical illusion. Thus, Hume asked what idea or impression we associate with any given word, assuming that if we cannot provide one, then the word has no meaning. We may now suspect that this challenge has no force: if I can use and understand all of the sentences in which a word appears, that shows that I understand it; it is beside the point that I associate no image or impression with the word alone. An empiricist who thinks that our grasp of our language involves linking it to experience need claim only that we understand an expression when we know how to test the sentences in which it occurs against experience. He need not match experiences to individual words.

If this is correct, we should be wary of assuming that every word corresponds to some thing or idea – its meaning. So long as we can explain how the presence of a word is reflected in what can be done with the sentences that contain it, we have all the account we need of its meaning. Hence, Frege insisted in his early writings that we should not ask after the meaning of a name in isolation, but only in the context of a sentence: once we know how the name functions in sentences, we know its meaning.

This insight can be applied in a number of different ways. For example, it can liberate us from the urge to look for some object which functions as the ‘meaning’ of a predicate. We can focus, instead, upon the roles of predicates in sentences, and upon the kinds of claims about meanings that are actually made in ordinary talk about talk. Thus, in ‘On what there is’, Quine writes:

The useful ways in which people ordinarily talk or seem to talk about meanings boil down to two: the having of meanings, which is significance, and sameness of meaning, or synonymy. What is called giving the meaning of an utterance is simply the uttering of a synonym, couched, ordinarily, in clearer language than the original. If we are allergic to meanings as such, we can speak directly of utterances as significant or insignificant, and as synonymous or heteronymous one with another. The problem of explaining these adjectives … is as difficult as it is important. But the explanatory value of special and irreducible intermediary entities called meanings is surely illusory.

(FLPV, pp. 11–12)

The point being made is that we should be able to explain what is involved in expressions being meaningful, to account for what we do when we learn or explain meanings, and to explain what is involved when expressions mean the same, without introducing meanings as abstract objects. Ordinary talk of meaning does not force this way of thinking upon us; and we should not assume that an expression is meaningful only if there is some meaning that it has. Positing objects like meanings need not be part of the best explanatory theory of how we are able to speak a language.

In the previous section, we also raised a problem about the meaning of connectives: what do ‘and’ and ‘not’ stand for? It was one of the fundamental thoughts of Wittgenstein’s Tractatus Logico-Philosophicus that these expressions (called the ‘logical constants’ because of their fundamental role in inference) ‘do not represent’. Exploring what this might involve will offer a further illustration of the primacy of sentence meaning. The idea is that these expressions are truth functional. Consider the following three assertions:

Quine is American

Frege is German

Quine is American and Frege is German

The third does not report some additional fact about or feature of the world which is not captured in the two previous claims. Rather, I understand the word ‘and’, in uses of these kinds, when I grasp that the conjunction of two sentences is true when both are true, and that otherwise it is false. I do not need acquaintance with some abstract entity conjunction, and there are not features of the world which can only be described in a language which contains ‘and’. Once I know this rule for determining the truth values of conjunctive sentences on the basis of the truth values of the parts, then I understand ‘and’, and can sensibly use conjunctive sentences. Similarly, I understand ‘it is not the case that’ when I see that the complex sentence formed by prefixing that operator to another sentence is true when the prefixed sentence is false, and false when it is true. The connectives are ‘truth functional’ because the truth values of the sentences we construct with their aid are functions of (depend upon) the truth values of their parts. The primacy of sentence meaning enables us to avoid the problem of accounting for our acquaintance with the abstract objects conjunction and negation.

There is another way to apply the insight, which appears to involve a wholly different strategy. Quine often links the primacy of sentence meaning to the use, by Bentham and Russell, of the method of paraphrase or contextual definition (TT, pp. 68–70, 75). As we have seen, one argument for the existence of attributes exploits the fact that we appear to refer to them with singular terms. Thus

Redness is a colour.

We might try to remove the illusion that ‘redness’ was a singular term standing for some mysterious abstract entity by paraphrasing this sentence as

Anything red is coloured.

This does not yield an analysis of the word ‘redness’. There is no part of the second sentence which means the same as ‘redness’. Rather, it suggests a method for paraphrasing every sentence that contains the word ‘redness’ into a clearer one that lacks this term. We offer, not an account of the word ‘redness’, but an account of all of the sentences in which that word occurs. In Quine’s view, only the primacy of sentence meaning can explain why such analyses seem satisfactory. A familiar example of the same strategy involves sentences like:

The average car owner drives 200 miles each week.

This appears to ascribe a property to a curious abstract object, the ‘average car owner’. Anyone possessed of ontological scruples may complain that there is no such thing. The quoted sentence provides a misleading expression of a thought whose structure is more perspicuously revealed by the paraphrase:

If you divide the number of miles driven by cars each week by the number of drivers, the answer is 200.

The apparent reference to ‘the average car owner’ has vanished.

Our two applications of this second ‘milestone’ are importantly different. The first forces us to refocus our sense of what has to be explained when we account for our ability to use language. By turning our attention away from words to sentences, it enables us to see that there need not be an object (its meaning) for which each word stands. However, it does not direct our attention away from the task of explaining our understanding of an ordinary natural language. The paraphrase technique, on the other hand, simply provides another sentence which, supposedly, ‘reveals’ the content that it shares with the first. Rather than offering insight into how a natural language works, we see a suspicion of natural languages as vehicles for the perspicuous representation of thoughts: we must always be on the lookout for deceptive formulations.

It is important to see how this second application is related to the primacy of sentence meaning. If word meaning is primary, we can only clarify the meaning of a sentence by clarifying the meanings of the words that it contains. A revealing paraphrase would have to contain paraphrases or analyses of the more troublesome terms in the sentence. Quine holds that the primacy of sentence meaning frees us of this requirement. A perspicuous paraphrase of a puzzling sentence may contain no constituent corresponding to some word or phrase in the sentence it paraphrases. No portion of our paraphrases captured the meaning of ‘the average driver’ or ‘redness’. However, we can feel confident that we have understood the original sentences because we have a recipe for constructing paraphrases of any sentences in which these expressions occur. Only a prejudice that the meaning of each word must be separately captured in our analysis could lead us to be dissatisfied with this, according to Quine.

Problems emerge when we ask for a fuller account of the relation between the original sentences and their paraphrases. One position is that the paraphrases provide ‘analyses’ of the original sentences: they characterize the structure of the ‘thoughts’ which were clothed in potentially misleading linguistic garb for communicative purposes; or they elucidate some underlying logical or grammatical structure ‘present in’ the original sentences. As we shall subsequently see, Quine, unlike Russell, is unhappy with this sort of account of what is going on. He would favour a weaker claim, that the two can ‘serve the same purpose’, and that the paraphrases are preferable, probably because they are clearer. He is likely to insist that our expressive capacities would not be limited if we revised our language, using only the long-winded paraphrases and banning the more familiar idioms. While such a change might interfere with rhetorical or conversational aspects of language use, it would not leave us unable to describe states of affairs which we could previously discuss. Moreover, in using this clumsy, awkward language we should escape the misleading surface features of English which deceive us into being puzzled by unreal philosophical problems.

The claim is not that we should actually speak this revised form of English. Rather, we should restrict our attention to it when our interests are philosophical. Looking at the paraphrases in this way enables us to obtain their philosophical benefits without believing in hidden structures which guide our thought but are disguised by the surface forms of English. (For a fuller discussion, see chapter 5 below.)


When we return to the problems about names for fictional entities and other ‘non-existent’ objects, we may think that the primacy of sentence meaning promises a straightforward solution. A name may be meaningful even if it fails to refer, so long as we understand the sentences in which it occurs: it is only prejudice that makes us suppose that a name cannot be meaningful unless it refers. However, Quine does not rest with that observation, and it is important that we should understand why.

First, note that names typically do refer to objects. It is a natural objection to the referential theory of meaning that it takes a model of meaning which applies to names and generalizes it to cover expressions of all kinds. Thus, we should not be overhasty in ignoring the referents of names and other singular terms. Secondly, there are philosophical reasons for being wary. We naturally view the world as composed of objects which have different characteristics. Theories differ in what objects they take the world to contain. Quine speaks of the set of objects which a speaker takes to exist as his ‘ontology’: an assertion ‘ontologically commits’ its author to objects of a certain kind if the assertion would only be true if objects of that kind existed. If we are to retain this view of the world, then two problems require attention. First, we need a philosophical elucidation of the concept of an object: just what is it for something to be an object? Secondly, we need an explanation of how we establish which objects a given theory or theorist is ‘committed’ to. How do we establish whether, for example, elementary arithmetic is true only if there are such abstract objects as numbers?

Perhaps the fact that names typically refer to objects provides the key to the solution of these problems. Some philosophers have claimed that we can only elucidate the concept of an object through the concept of a name or singular term: all we can say about objects is that they are the the things that singular terms refer to. If that were so, then we might undermine our concept of an object by blithely rejecting a referential account of the functioning of names, brandishing the slogan that the sentence is the unit of meaning. Moreover, it is desirable that there be a way of telling from a theory or assertion just what its ontological commitments are: we would like a criterion of ontological commitment. It is natural to suppose that a theory is committed to there being objects of a certain kind if it makes claims which involve names or singular terms for objects of that kind. Unless singular terms stand for objects, how are we to establish what objects a given theory says there are? Names appear to be intimately related to objects, and it would be wrong to lose sight of this.

In ‘On what there is’ (FLPV), Quine pursues this question: are there any forms of linguistic behaviour through which we do unambiguously commit ourselves to the existence of objects of different kinds? This prompts a search for a criterion of ontological commitment: a rule that we can apply to a body of assertions in order to determine what ontology must be accepted by anyone who is willing to make all those assertions. As Alonzo Church has remarked, Quine was the first philosopher to take seriously the search for such a criterion (Church 1958). ‘On what there is’ contains the results of this search. The remainder of this chapter provides an informal introduction to these views.


We begin with Russell’s theory of descriptions. Quine looks upon this as the paradigm use of paraphrase in philosophical analysis and, as we shall see, his views about ontological commitment serve as a kind of corollary to it.

In early work, Russell had accepted that every meaningful expression of a language stands for something. This led him to what Quine calls his ‘intolerably indiscriminate ontology’ (TT p. 74): alongside existents, were such entities as ‘numbers, the Homeric gods, relations, chimeras, and four-dimensional spaces’ (Russell, 1903 p. 449; quoted in TT p. 74). By the time he wrote his classic 1905 paper ‘On denoting’, Russell wished to avoid these excesses. His account of definite descriptions provides one of his means towards achieving this.

Russell claimed that neither of these two sentences

The King of France is bald

The King of France does not exist

has the logical structure that we have represented


Neither is really of subject–predicate form. Hence, it is not a precondition of either making sense that there ‘is’ (in some sense) a King of France. The surface appearance of each sentence conceals an underlying logical structure according to which neither sentence contains a singular term. Their structures can be represented more perspicuously thus:

There is one, and only one King of France, and he is bald

It is not the case that there is a King of France

Although these sentences do involve the predicate expression ‘is a King of France’, they do not contain singular terms which purport to refer to him. If there is no King of France, the first is straightforwardly false, and the second is true; neither contains a name or other singular term which fails to refer. Hence, ordinary definite descriptions are not genuine singular terms; so, arguments which urge the reality of different kinds of objects by assuming that definite descriptions are singular terms fail. We are fooled because the surface appearance of sentences of our language can conceal their underlying logical structure.

Russell’s theory provides a tool which enables us to escape the seductive force of the surface structure of our language. We do not commit ourselves to the existence of objects whenever we use a definite description. The paraphrases enable us to identify those descriptions which do commit us to the existence of objects; they are the ones which are paraphrased as saying that ‘there are’ objects of the appropriate kinds. Neither sentence requires there to be a King of France for it to be meaningful. Only the first requires that there be a King of France for it to be true. So, while the sincere assertor of

The King of France is bald

must believe that there is a King of France, the utterer of

The King of France does not exist

need not.