THE ANALYTIC-SYNTHEI'IC DISTINCTION

Dissertation for the Degree of Ph; D. *
MICHIGAN STATE UNIVERSITY
DALE KENT
1976

 

 

   

This is to certify that the

thesis entitled

THE ANALYTIC-SYNTHETIC DISTINCTION

presented by

Dale Kent

has been accepted towards fulfillment
of the requirements for

Ph. D. Philosophy

degree in

%Qkéltff fj/ééz44ik/y

Major professor

Date August 9, 1976

0-7 639

 

 

MSU

LIBRARIES

 

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped beIow.

 

 

 

 

 

(I; /<./// 7/ ‘5’.

ABSTRACT

THE ANALYTIC-SYNTHETIC DISTINCTION

By

Dale Kent

The primary purpose of this work is to distinguish from
each other, as clearly as possible, the various arguments
which have been, or could be, made for and against the
analytic-synthetic distinction.

The arguments against the distinction which are considered
in this work do not presume the truth or falsity of Quine's
thesis of the indeterminacy of translation, although some
of these arguments can be viewed as arguments for Quine's
thesis.

Arguments are given for the claims that: l. The notion
of analyticity has no justificatory power; 2. Analyticity has
no explanatory power; 3. There are no analytic sentences;

4. The notions of meaning and synonymy are not needed to
describe successful translation and communication; 5. The
notion of analyticity is not clear enough.

Various defenses of the analytic-synthetic distinction
are examined and found to be wanting.

An important sense of 'obscure' is defined, in the
light of which the claim that 'analytic' is obscure is
examined. The attempt to explain away the obscurity of

'analytic' in terms of the indeterminacy of other expressions~

Dale Kent

is seen to involve the appeal to the notion of intensional
obscurity, a notion which is as problematic as analyticity.

The claim that the notion of analyticity has utility
in an ideal language is seen to depend on a confusion between
intrasubjective and intersubjective synonymy.

Carnap's empirical criterion of synonymy is examined
and found to be defective on the grounds that it is subject
to a partial circularity and on the grounds that it doesn't
provide a clarification of the notion of intralinguistic
synonymy.

One of Grice and Strawson's criteria of synonymy is
examined and found to indirectly presuppose the notion of
meaning and sameness of meaning, and therewith of analyticity.

The notions of analyticity by fiat and truth by con-
vention are shown to be incapable of guaranteeing analyticity
and of answering the criticism that analyticity has neither
explanatory nor justificatory power.

Quine's response to Grice and Strawson's proposal to
define synonymy along the lines of the verifiability theory
of meaning is seen to involve a misinterpretation of this
proposal. However, their proposal does not seem to hold the
promise of answering the criticism that analyticity has
neither explanatory nor justificatory power and the criticism
that there are no analytic sentences.

Two kinds of paradigm case arguments are distinguished.
It is shown that neither of them has the power to prove that
apparently analytic sentences have the properties traditionally

ascribed to them.

Dale Kent

Quine's notion of stimulus synonymy is examined and
found to provide a better reconstruction of traditional
synonymy than he claims. But it doesn't appear to provide
a complete reconstruction.

The idea that meanings may someday be made legitimate
by being discovered to be correlated with brain states is
examined. Several ways in which this might be accomplished
are discussed. It is argued that several dubious assumptions
are involved in such a program, as well as the fallacy of
thinking that if an expression is significant or meaningful
then there is a meaning which it has.

Finally, analyticity is examined in connection with
artificial languages. Conditions are specified under which
a predicate defined for the sentences of a formal language
can be said to be an analyticity predicate for that language.
A way is shown in which the use of an artificial language
might be used to clarify the notion of analyticity as it

applies to a natural language.

THE ANALYTIC-SYNTHETIC DISTINCTION

By

.-\ P‘"
\r

Dale Kent

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of
DOCTOR OF PHILOSOPHY

Department of Philosophy
1976

Copyright by
DALE KENT
1976

ACKNOWLEDGMENTS

I wish to thank Herbert Hendry for his kindness and

:for his valuable suggestions and comments, which resulted

:in improvements in this work.

iii

To argue against someone else's entire world view is a

rnost arduous sort of argument. Those within world view A

113ve a way of handling virtually every difficulty that comes

‘their way; and when absolutely necessary, they make adjust-
Inents in order to preserve their scheme. Those who perceive
21nd act through world view B cannot voice objections entirely

jgn the language and with the images familiar to them. For

‘tlne two systems, world View A and world view B seldom lie,
It

ssca to speak, on the same plane or along the same axis.
j.ss extremely difficult to "translate" from the language of

1:}1e one to the language of the other. The hermeneutical

p rob lems are severe .
Consequently, we tend to try new world views on for

:sciflze - at first, they hardly ever fit. Usually, we return

areazinfOrced, to our familiar world. Often, we do not "learn"

Intzczh.from our critics. They seem to us to miss the point.
Almost necessarily so.

Michael Novak

iv

CONTENTS

Introduction .

Chapter 1.
l.

2. The Justificatory Power of Analyticity.

3. The Explanatory Power of Analyticity.

4. The Clarity of 'analytic'

5. Whether the Notion of Meaning is Needed to
Describe Successful Translation and
Communication .

(Ihapter 2. Defenses of Analyticity.

l. Shifting the Blame for the Obscurity of
'analytic‘ onto Other Words . .

2. The Utility of Analyticity in an Ideal
Language. . . . .

3. Carnap's Empirical Criterion of Synonymy.

4. Grice and Strawson's Understanding-Believing
Criterion of Analyticity.

5. The Existence of Analytically True Sentences.

6. Analyticity by Fiat and Truth by Convention .

7. The Verifiability Theory of Synonymy.

8. Stimulus Synonymy as an Explication of
Synonymy. . . . . . .

9. Meaning, Significance, and Brain States
Chapter 3. Analyticity in Artificial Languages.
Chapter 4. Summary.

FOOtnotes.
Bibliography

Criticisms of Analyticity.

The Correspondence Problem.

15
18
27

48

53

53

62
66

75
78
92
98

.107
.118
.124
.150
.161
.166

INTRODUCTION

Anyone familiar with the literature on the analytic-
synthetic distinction cannot fail to be impressed by the

sheer number of apparently distinct arguments for and against

the legitimacy of this distinction. The principal purpose

of the present work is to distinguish from one another, and
to make as explicit as I can, the important arguments which
have been, or might be, made for and against the distinction.

It has sometimes been said that a dispute over a funda-

mental issue in phi1050phy, if pursued far enough, is seen

to ultimately involve a question begging, by both parties to

the dispute, of the very issues under dispute. The dispute

over the analytic-synthetic distinction often seems to be an

example of this. This work will have succeeded if it shows

Where such question begging occurs and if it shows what work

needs to be done, in those places where question begging

dOesn't occur, in order to move forward in clarifying the
Various issues involved in the debate over the notion of
analyticity and related notions.

The impression of neutrality in what I have just said
may appear to be contradicted by the body of the present

Work, which seems to be, for the most part, squarely against

the analytic-synthetic distinction. But the thrust of this

work is due merely to a historical accident, so to speak.

The historical accident is that because of the powerful
arguments against the notion of analyticity presented by

Quine the burden of poorf at the moment lies with the pro-

ponents of the analytic-synthetic distinction. One of the

aims of this work is to show as clearly as possible what it

is that the friends of analyticity have to prove.
I, for one, am not ready to give up on the notion of

zanalyticity. This is because, although Quine has offered a

Iaositive doctrine which is claimed to be incompatible with

the analytic-synthetic distinction, the specific parts of
this doctrine dealing with apparently analytic sentences
seem no better at explaining their "analytic appearance"
(i.e. their apparent necessity, immunity from revision, lack

of informational content, etc.) than the things said by the

prOponents of the analytic-synthetic distinction. It may,

in the end, turn out either that the appearances presented
by what we call analytic sentences need no explanation or
that it is not the phi1050pher's job to explain them. But

this remains to be seen.

Chapter 1
Criticisms of Analyticity

My main concern in this chapter is to consider various

things which could be meant by saying that someone rejects

or criticises the analytic-synthetic distinction, as applied

to natural language, and also to examine some replies to

“these criticisms.
There is one interpretation which can be ruled out

:immediately; i.e. an interpretation which focuses on the
vvord 'distinction' and which considers an attack on the

zinalytic—synthetic distinction to include a claim that the
(listinction is not sharp, or is not clear, because the
Ipcredicates 'analytic' and 'synthetic' overlap in the sense

‘t hat there are or could be sentences which are both analytic

2111d synthetic. We can rule out such an interpretation because

tzllose who have upheld the analytic-synthetic distinction,

Ifeegardless of how they define 'analytic', define 'synthetic'

to mean simply 'not analytic'. Thus, to this kind of criti-

Cism we can say that the analytic-synthetic distinction is

as clear as the word 'not'.

A more likely interpretation of what it means to
cllf'iticise the analytic—synthetic distinction is that such a

(ZIfiiticism consists of the claim that no sentence of any

nfiltural language is analytic. Harman has interpreted some

013 Quine's objections to the notion of analyticity in this

VVa)n He has also interpreted Quine as wanting to say that

"one cannot make sense of the analytic-synthetic distinction
in any way such that there turn out to be analytic truths".3
This last comment might tend to suggest that one of the things
wrong with the analytic-synthetic distinction is that one
cannot define or explain the notion of analyticity in such
a way that the claim that analytic sentences exist in some
sense indirectly follows from the definition of 'analytic'.
But any notion is such that it cannot be defined or explained
in any way such that there turn out to be things to which
the notion applies. (A counterexample to this claim might
seem to be provided by so-called ostensive "definition" and
its analogue, the paradigm case argument. However, these
will be discussed in Section 5 of Chapter 2, where it will
be seen that they are useless in proving whether there are
early sentences which have the properties traditionally invoked
in various definitions of analyticity.) Consequently, we
need to keep in mind the distinction between the question of
what it means to say that a sentence is analytic and the
question of whether there are any analytic sentences. One
Inight conceivably be encouraged to conflate these two questions
by Quine's comment that "...we have made no general experi-
Illental sense of a distinction between what goes into a
native's learning to apply an expression and what. goes into
his learning supplementary matters about the objects
C0ncerned".4 But if we don't make a distinction between
Whalt it means to say that something has a property and the

Clllestion of whether there are any things which have the

property then we will be free to "prove" that there are
analytic sentences by pointing to a sentence such as
'Everything is self identical' and saying that according to
some definition of analyticity this sentence is analytic.
There is a broad characterization of analyticity, upon
which Quine has commented, and which gives rise to what I

call the correspondence problem.
Section 1: The Correspondence Problem

One broad way of characterizing analytic truths, as well
as purely logical truths, is by saying that they are sentences
that are true purely in virtue of their meaning.5 This
characterization is intended to apply not only to those
sentences which can be turned into purely logical truths by
substituting synonyms for synonyms (such as 'All bachelors
are unmarried') but also to sentences which apparently cannot
be so transformed (such as 'If one thing is warmer than a
second, and the second is warmer than a third, then the first
is warmer than the third'). It also is intended to apply to
a truth such as 'Everything is identical with itself' and to
any propositional tautology. To say that these sentences are
true purely in virtue of their meaning can be taken to mean
that, given that they have the meanings which they do have,
it follows that they are true. This characterization needs
to be distinguished from a similar one which we will discuss
in Section 3; i.e. the definition of an analytically true

sentence as one which is either believed to be true or known

to be true purely in virtue of knowledge of its meaning.
The definition of the present section makes no reference to
knowledge or belief.

Now with respect to any definition of analyticity there
arises what might be called the existence problem, which is
the problem of whether there are any sentences which have
the property expressed in the definiens of the definition.
One way to show that there are no sentences whose truth
follows from the fact that they have the meaning which they
do is to take an arbitrarily chosen sentence which appears
to be analytic and show that it is possible for it to be
false. If this can be done then this is enough to show that
although it may be true, its truth does not follow merely
from the fact that it has the meaning which it does; i.e. it
is not true purely in virtue of its meaning. I will discuss
the existence problem in Section 5 of Chapter 2. I mention
it now only to distinguish it from what I have called the
correspondence problem. The correspondence problem is also
a problem which seems to arise out of the characterization
of an analytically true sentence as one which is true purely
in virtue of its meaning. While an analytically true sentence
is one which is supposed to be true purely in virtue of its
meaning, a synthetically true sentence is supposed to be one
which is true not only in virtue of its meaning but also in

virtue of the way the world is. As Quine puts it:

It is obvious that truth in general depends on
both language and extralinguistic fact. The
statement 'Brutus killed Caesar' would be false
if the world had been different in certain ways,
but it would also be false if the word 'killed'
happened rather to have the sense of 'begat'.6

But in objecting to the notion of analyticity Quine goes on
to say:

Thus one is tempted to suppose in general that
the truth of a statement is somehow analyzable
into a linguistic component and a factual compo-
nent. Given this supposition, it next seems
reasonable that in some statements the factual
component should be null; and these are the
analytic statements. But, for all its a priori
reasonableness, a boundary between analytic and
synthetic statements has not been drawn. That
there is such a distinction to be drawn at all is
an unempirical dogma of empiricists, a metaphy-
sical article of faith.

And in another place he says:

...true sentences generally depend for their truth
on the traits of their language in addition to the
traits of their subject matter; and...logical truths
then fit neatly in as the limiting case where the
dependence on traits of the subject matter is nil.
Consider however, the logical truth 'Everything is
self identical', or '(x)(x=x)'. We can say that

it depends for its truth on traits of_fhe language
(specifically on the usage of '='), and not on
traits of its subject matter; but we can also say,
alternatively, that it depends on an obvious trait,
viz., self identity, of its subject matter, viz.,
everything. The tendency of ou§ present reflections
is that there is no difference.

These remarks suggest a positive doctrine, in the light of
which the correspondence problem arises. The doctrine goes
as follows.

Eygry_true sentence depends for its truth on two factors,
the first one being the meaning which the sentence has, and
the second one being the way the world is. Even the

sentence 'Everything is identical with itself' depends for

its truth not only on the fact that the word 'identical'

has the meaning which it does, but also on a certain non-
linguistic fact about the world, namely the fact that
everything is identical with itself. Because a true
sentence has to correspond to the way the world is in order
to be true, there are no true sentences whose truth is
independent of the way the world is. Every true sentence
has its two-fold dependency on meaning and fact, on language
and the state of the world. There are no sentences in which
the contribution which the world makes to their truth drops
out.

In the light of this doctrine Quine has been viewed as
implicitly accusing Carnap of not taking the correspondence
theory of truth seriously when it comes to analytically true
sentences.9 Carnap seems to be ignoring the contribution
which the world makes to the truth of such sentences. He
seems to be saying that analytically true sentences are true
no matter how the world is; that an analytically true sentence
doesn't have to correspond to the facts in order to be true
because it is true in virtue of meaning and thus true no
matter what the non-linguistic facts about the world are.

He seems to be saying that even if the world were such that
there were things in it which were not self identical the

sentence 'Everything is self identical' would still be true
because its truth is not dependent on the way the world is

but only on its meaning.

Quine can be considered as saying that if one is to take
the correspondence theory of truth seriously then this sen-
tence, as well as all others, depends partly for its truth
on its correspondence with the world, in the sense that if
the world were different in relevant ways then the sentence
would be false.

Now the idea of one thing being independent of another can
be vaguely analyzed by saying that two things are independent
of each other if changes or variations in the one are not accom-
panied by changes or variations in the other. This is pretty
vague. But it seems clear enough to suggest a way in which
the friends of analyticity might attempt to extricate themselves
from the correspondence problem. They might define a sentence
to be analytically true just in case it is true, and moreover,
is such that if the world were different in relevant respects
then the meaning of the sentence would also be different in
such a way that the sentence would remain true. A synthetically
true sentence could then be defined as one which is true, and
moreover, is such that it is not the case that if the world
were different in relevant respects then the meaning of the
sentence would also be different in such a way that the sen-
tence would remain true.

Let us look at a possible problem hidden in the notion of
"relevant respects", in terms of the example 'Everything is self
identical'. The only aspect of the world which is clearly
relevant to the truth of this sentence is the fact that every-
thing is self identical. Now our counterfactual definition

of analyticity, as applied to the apparently

10

analytic sentence 'Everything is self identical', says that
if the world were such that not everything is self identical
then the sentence 'Everything is self identical' would still
remain true. But we are inclined to protest that, on the
contrary, if the world were such that not everything is self
identical then the sentence 'Everything is self identical'
would clearly be false. But this protest misses the intended
meaning of the definition. Although the definition is of
counterfactual form, what is intended by it can perhaps be
illustrated in terms of the notion of various distinct worlds.
The basic idea is that if 'Everything is self identical' is
analytically true then it is true in this world and, moreover,
is such that any world which we would from our world describe
as being a world in which not everything is self identical
would also be a world in which the sentence 'Everything is
self identical' has a different meaning than it does in our
world. In every such world it would have the meaning which
the sentence 'Not everything is self identical' has in our
world.

And, taking the sentence 'I have a green car' as an
example, the basic idea behind the definition of a synthetically
true sentence is that this sentence is true in our world, but
is not such that in every world which we would from our world
describe as a world in which I don't have a green car is a
world in which the meaning of the sentence 'I have a green
car' is different than it is in this world. In some, but not

all, of these worlds the sentence has the same meaning which

11

it does in this world; and in these worlds the sentence is
false.

Talking in terms of the notion of various distinct
worlds also helps us to see a way in which the basic idea
intended by our definition is an improvement over the notion
of an analytically true sentence as one which is true in all
(and presumably only) possible worlds. Assuming (as the
logical positivists did) that analyticity provides an expli-
cation of logical and necessary truth, the basic idea behind
our definition is that analytically true sentences are true

in all worlds simpliciter, including even "impossible" ones.

 

For example, a world which we, in our language, would describe
as a world in which not everything is self identical is
presumably an impossible world. Yet the sentence 'Everything
is self identical', if both true and analytically true in our
actual world, would be true in this impossible world also
because, according to the idea behind the definition, this
sentence would have a different meaning in this impossible
world than it does in our actual world. Thus, since the
truth of this sentence (assuming that it is analytically
true) is independent of the way the world is, our definition
of analyticity solves what I have called the correspondence
problem. However, the notion of an analytically true
sentence as one which is true in all (and presumably only)
possible worlds does not solve the correspondence problem.
For on this latter definition of analyticity the truth of

an analytically true sentence is not totally independent of

12

the way the world is. For if the world were such that not
everything is self identical then, since such a world would
be an "impossible" world, the sentence 'Everything is self
identical' would be false. According to the basic idea
behind our definition however, the truth of an analytically
true sentence is totally independent of the way the world is
since such a sentence is true no matter how the world is.

A question may seem to arise with respect to the world
which I said was a world which we, in our language, would
describe as a world in which not everything is self identical.
Assuming that 'Everything is self identical' is analytically
true, this sentence would, in that world, be true because in
that world it would have a different meaning which it has in
our world. The question then is: Is this weird world one
in which everything is self identical or not? What, in
other words, is the "real truth" about this world? The answer
comes to light when we consider the fact that a given string

of words is not true or false simpliciter, but true or false

 

in a language. A given string of words can be a true sentence
of one language and a false sentence of another language.

In our language this weird world is truly describable as a
world in which not everything is self identical. But in

this weird world some strings of symbols would have different
meanings than they do in our language. In the language used
by the inhabitants of this weird world the string. of symbols

‘Everything is self identical' would be a true sentence.

13

If any sentence "behaves", so to speak, in the way that
our notion of analyticity says that it does, it is still
hard to see how this "behavior” is due to the fact that the
sentence has the meaning which it does in this world. But
if any sentence does have the properties attributed to it by
our notion of analyticity, we may just want to take this as
a strange ultimate fact about those sentences, just as we
take it as a strange and ultimate fact about the world that
space is curved. At any rate, our notion of analyticity
does seem to provide an answer to the correspondence problem;
a problem which consists of the question of how it is pos-
sible for the truth of a true sentence to be independent of
the way the world is. I know of no other way to meet this
problem.

I might note parenthetically that our notion of analy-
ticity not only solves the correspondence problem but it also
has an additional advantage over the notion of truth in all
(and only) possible worlds. This latter notion never seemed
illuminating to me since a possible world can only be des-
cribed, it seems, as a world in which logical or analytic
truths hold true. But then to say that logical truths are
true in all and only possible worlds is just to say that
logical or analytic truths are true in all and only those
worlds in which they are true. But any sentence is true in
all and only those worlds in which it is true. Consequently,
a truth of physics could be defined as a sentence which is

true in all and only physically possible worlds, i.e. worlds

14

in which all the truths of physics hold true. But this, of
course, is equally unilluminating. However, if we construe
the basic idea behind our definition of analyticity in terms
of various distinct worlds we need not use the word 'possible'.
We can just say that an analytic or logical truth is one which
is true in all worlds, whereas a synthetic truth is not true
in all worlds.

Although the basic idea behind our definition of analy-
ticity seems easily construable in terms of the notion of
various distinct worlds the definition itself is of counter-
factual form. Counterfactuals, if they are supportable at
all, are supported by true laws pertaining to objects in our
actual world. Are there any statements of lawlike form which,
if true, would support the definition? It seems that there
are and that the friends of analyticity are appealing to such
lawlike statements when they say that if anyone disagrees
with us as to the truth of a sentence such as 'Everything is
self identical' then he is using the expressions in this
sentence to mean something different from what they normally
mean. The question arises, of course, of whether this latter
claim is true. In Section 5 of Chapter 2 I will consider an
argument which attempts to show that it is not true. But
for now I merely want to note that it is not irrelevant, as
Harman seems to think,10 for the friends of analyticity to
make such a claim in support for their claim that some
apparently analytic sentences are true independent of the

way the world is, and that they thus behave in the way that

15

our definition of an analytically true sentence says they
do.

I do not consider the correspondence problem to be
particularly important. I have discussed it for the sake
of completeness because it is a problem which seems to be
suggested by the comments of Quine quoted earlier in this
section. A more important problem is the question (to be
discussed in Section 5 of Chapter 2) of whether there are
any analytic sentences. But an even more important problem
is the problem of how the claim that a sentence is analyti-
cally true justifies or explains the claim that the sentence
in question is true. This last problem is of paramount
importance because the primary interest in analyticity lies
in the epistemological role it is supposed to play. The
appeal to analyticity is supposed, in some sense, to account
for, or justify, or explain, our belief in or knowledge of
the truth of a sentence. In the next two sections I will

examine this question.
Section 2: The Justificatory Power of Analyticity

The ascription of analyticity to a sentence might be
considered to be a justification for the claim that it is
true. If, for example, I claim that all bachelors are
unmarried-and someone asks me to justify this claim (by
asking me how I know that it is true, or why I believe it)
then it might seem that I have provided a good justification

by saying that the sentence in question is analytic. But

16

is this a good justificatiOn?

Typically, the request for justification arises when
there is doubt or disagreement between two people over
whether a certain sentence is true. I will have provided
a good justification if I can show how the sentence in
dispute either inductively or deductively follows from
other beliefs which we both share. But in such a typical
situation the sentence in dispute means the same for both of
the disputants. It is just that one of them is privy, or
thinks he is, to some evidence for the truth of the sentence
which the other is not.

Let us go along, for the time being, with the claim of
the friends of analyticity that if someone doubts that all
bachelors are unmarried then he is using the words 'bachelor'
or 'unmarried' in a different sense then they are normally
used or, as it is sometimes put, he is Speaking a different
language. More to the point for the purposes of this section
is the case where I, in a philosophical spirit, ask myself
what my justification is for the claim that all bachelors
are unmarried. And the question is whether the claim that
this sentence is analytic is a good justification for the
claim that it is true. But here again I will have provided
myself a good justification if I can show myself that the
sentence in question follows from something which is more
certain, or believed by me to a greater degree, than the
claim that all bachelors are unmarried. However, the claim

that the sentence in question is analytic does not provide

17

a good justification for the claim that it is true for the
simple reason that the former claim is no more certain than
the latter. In order words, the degree of my belief that
the sentence in question is analytic is no greater than the
degree of my belief that it is true. Both beliefs are
equally certain. I provide no better justification for my
belief that the sentence in question is true by saying that
it is analytic than I do by saying that it is obviously true.

In order for the ascription of analyticity to a sentence
to constitute a good justification for the claim that the
sentence is true, it must be possible to believe in its
analyticity more than in its truth. But I know of no sen-
tence for which this is possible. And I am confident that
the friends of analyticity would agree that it is not pos-
sible to believe more strongly in the analyticity of a
sentence than in its truth. But if so, then in what sense
does the claim that a sentence is analytically true provide
a justification for the claim that it is true? Wouldn't it
just be better to say that sentences such as 'All bachelors
are unmarried' and 'Everything is self identical' either
need no justification or are obviously true (or perhaps even
that they are inductively supported by their instances)
instead of putting forward a theory which provides only a
pseudo-justification?

Let us now look at the claim that the ascription of
analyticity to a sentence constitutes part of an explanation

for our belief in its truth. There are some differences

18

between explanation and justification. In a justification
the justificandum, i.e. the thing to be justified, is more
in doubt than the justificans, i.e. that which does the
justifying. In an explanation the situation is reversed.
The explanandum, i.e. the sentence describing the event to
be explained, is often less in doubt than the explanans.
Perhaps this difference might permit us to attribute some

explanatory power to the notion of analyticity.
Section 3: The Explanatory Power of Analyticity

An analytically true sentence might be characterized
not as one which is true purely in virtue of its meaning
but as one which is, so to speak, believed true purely in
virtue of knowledge of its meaning. More precisely, an
analytically true sentence might be characterized as one
such that if you understand what it means then you believe
that it is true. This is equivalent to saying that an
analytically true sentence is one such that if you don't
believe that it is true then you don't understand it. Let
us call this a doxological characterization of analyticity
in order to distinguish it from the one discussed in Section
1; i.e. from an analytically true sentence characterized as
one which is true purely in virtue of its meaning.

“Let us also distinguish yet a third characterization of
analyticity which might, for lack of a better word, be called
epistemological. This third characterization says that an

analytically true sentence is one a knowledge of whose

19

meaning is sufficient for a knowledge of its truth, or that

 

it is a sentence whose truth is knowable in virtue of a
knowledge of its meaning.

I will attempt to show that the ascription of analyticity
to a sentence has no explanatory power under either the
doxological or epistemological definition of analyticity.
But before I do there is a defect in these definitions which
needs to be discussed lest it be thought that what is said
in the sequel hinges in some essential way on this defect.
The defect in these definitions consists in the fact that
they do not apply to sentences whose analyticity is not
obvious. There are several examples of sentences which would
incline us to the view that it is not always the case that
the analyticity of a sentence is obvious. One kind of
example consists of sentences of the form 'x is good if and
only if P' where 'P' is replaced by one of the various
expressions which, in the history of ethics, have been put
forward as an analysis of 'good'. G. E. Moore claimed that
none of these definitions of 'good' were adequate because it
was not contradictory to claim that something had the
property P but wasn't good. In other words he claimed that
every such definition which had been proposed was, even if
true, only synthetically true rather than analytically true,
and thus did not capture the meaning of 'good'. In fact
what he called the naturalistic fallacy could be explained
as the fallacy of confusing the analytic with the synthetic

with respect to biconditionals of the above form. Stevenson

20

claimed that the reason why pre—Stevenson analyses of

'good' had failed was because the various expressions which
had been put in place of 'P' did not have the same emotive
meaning as 'good' even though their descriptive meanings
might be the same. But aside from questions of emotive
meaning I have always felt that Moore was too quick to claim
that various proposed definitions of 'good' were at best,
synthetically true. It always appeared plausible to me that
some of these proposed analyses of 'good' were indeed
analytic, but not obviously so.

Another example of a kind of sentence whose analyticity
is not obvious is provided by the following kind of case
which most of us have come across. Suppose that someone
gives a political analysis of a society which consists, in
part, of the claim that certain evils in society are due to
the fact that the society is capitalistic. Such a claim can
be considered to be of the form 'If there is capitalism then
E', where in place of 'E' is put some description of various
evils. One often finds that when counterexamples to a claim
of this form are proposed the ideologue conveniently and
subtly changes the meaning of 'capitalism' and of the
expressions put in place of 'E' in such a way as to save the
claim from refutation. When this happens many times one
begins to suspect that the claim is, in spite of appearances
to the contrary, really analytic for the ideologue even

though its analyticity is not obvious even to the ideologue.

21

A third kind of sentence which inclines one to want to
countenance sentences whose analyticity is not obvious is
provided by mathematics. We may want to say that all, or
perhaps just most, of the truths of mathematics are analytic.
Consider, for example, Goldbach's conjecture, which says
that every even number is the sum of two primes. Here is a
sentence whose meaning we understand. Yet we may not believe
that it is true. We may believe it on the grounds that
every even number so far investigated is the sum of two
primes. But the point is that understanding what it means
is not sufficient for believing that it is true, since it is
possible to understand what it means without believing that
it is true.

Now in the case of mathematical statements, if not in
the other two kinds of cases, we can expand our definitions
in such a way as to accommodate them. We might amend the
doxological definition, for example, by saying that a sen—
tence is analytically true either if it is believed true in
virtue of meaning or if it is a logical consequence of
sentences which are believed true in virtue of meaning.

Thus, if Goldbach's conjecture, for example, is, unbeknownst
to us, a consequence of our favorite axioms of arithmetic
then it too is analytic, although not obviously so.

It turns out, however, that the ascription of analyticity
to a sentence has no explanatory power whether or not we
expand the definitions in the way indicated. Consequently

I will, for the sake of simplicity, use our definitions in

22

their original unamended form, indicating at the appropriate
places where their expansion to include non-obvious analy-
ticity would make a difference to the argument.

Suppose we have an analytically true sentence S whose
analyticity is obvious. For example, let S be the sentence
'All bachelors are unmarried' or 'Everything is self identical'.
A Hempelian type explanation of Jones' belief in S's truth
in terms of S's analyticity would look like this:

1. For any x, if x is analytically true then anyone who
understands the meaning of x believes that it is true.

2. S is analytically true.

3. Jones understands the meaning of S.

therefore

 

4. Jones believes that S is true.

(If 'x' ranges over sentences whose analyticity is not
obvious then the explanation is already defective on the
grounds that premise l is false. But let that go since
there are more serious problems.)

One of Hempel's requirements for a good explanation is
that at least one of the lawlike statements in the explanans
be synthetic. This explanation violates that requirement
because the only lawlike statement in the explanans is
premise l, and premise l is analytic. Part of what it means
to say that a sentence is analytic is that anyone who under-
stands it believes that it is true. (Recall that we are
limiting the range of 'x' to those sentences whose analyticity

is obvious.) What would it be like for premise l to be

23

false? This would require that there be an analytically
true sentence which someone understands but doesn't believe
to be true. But we can be sure that the friends of analy-
ticity, if allegedly presented with such a sentence, would
say either that the sentence is not really analytic, or

that the person in question really doesn't understand it
completely, (or understands it differently from the way it
is normally understood), or that he really does believe that
it is true.

Since no lawlike statement in this putative explanation
is synthetic it is not really an explanation at all. Such
an explanation is analogous to "explaining" the fact that
' someone is unmarried by pointing to the fact that she is a
spinster, along with the "law" that all spinsters are
unmarried. It would seem that the explanation of Jones'
belief in the truth of S in terms of S's analyticity is no
advance over "explaining" his belief in terms of the fact
that the truth of S is obvious to him.

Quine has made a very similar objection to what he calls
the linguistic doctrine of elementary logical truth.11 There
are some differences between Quine's objection and ours.
Quine seems to be denying the explanatory power of an expla-
nation which has as its explanandum not the claim that Jones
believes S to be true, but rather the claim that S is true.
But this difference is unimportant since we can easily
construct a putative explanation, in analogy with the one

given above, which has as its explanandum the claim that S

24

is true. We need merely change 'believes' in premise l and
statement 4 to 'knows' and.add the further premise that if
anyone knows that x is true then x is true. We could then
deduce the further conclusion that S is true. The additional
premise that if someone knows a sentence to be true then it
is true is also a lawlike statement. But it too is analytic,
as well as the amended premise 1. Thus, this amended
explanation of the fact that S is true would still be in
violation of Hempel's requirement.

Curiously enough, Carnap says that he agrees with Quine's
objection that what Quine calls the linguistic doctrine of
elementary logical truth is empty and without experimental

12 He agrees with it because of his view that

significance.
philosophical doctrines which are the result of explication
are analytic. Perhaps one can go along with this view.
Perhaps one of the jobs of a philosopher is to make the
analyticity of certain doctrines more obvious. But what is
puzzling, to say the least, is the sense of 'explanation' in
which the claim that a sentence is analytic explains the

fact that a person knows a sentence to be true or the fact
that he believes it to be true. It seems that appealing to
the analyticity of a sentence in order to explain Jones'
belief that it is true is no more of an explanation than

that given by Moliere's physician when he attempted to explain
the fact that opium puts a person to sleep by appealing to

its dormitive virtue. One might conceivably respond to

this by saying that there is a sense of 'explanation', going

25

beyond Hempel's sense of the word, which permits there to
be distinctly philosophical explanations. But then the
burden is on one who so responds to provide a criterion for
distinguishing between good and bad philosophical explanations
and to explain why the explanation given by Moliere's
physician is not a good philosophical explanation, and why
an explanation which utilizes the notion of analyticity is
a better phiIOSOphical explanation than one which utilizes
the notion of, say, essences or intrinsic natures. In what
sense is it a better explanation to appeal to the analyticity
of the sentence 'Everything is self identical' as an account
of Jones' belief in this sentence than it is to say that it
is in the nature of everything to be self identical and that
Jones believes this because he has an insight into the nature
of reality?

Let us call the objection of analyticity which points
to the fact that statements 1 through 3 violate Hempel's
requirement that "no synthetic law" objection. The question
arises of whether Quine can consistently put forward this
objection. Carnap has hinted that he cannot,13 and that
Quine's objection to the linguistic doctrine of elementary
logical truth (which is relevantly similar to the no
synthetic law objection) presupposes, in some sense, the
very notion of analyticity which is being objected to. Thus,
the question of whether Quine can consistently put forward
this objection is no mere issue of exegesis, but rather a
question of whether the notion of analyticity is so funda-

mental that it is, in some sense, unavoidable.

26

Quine has sometimes been interpreted as claiming that

14 He has also often claimed that

no sentence is analytic.
the notion of analyticity is obscure or unintelligible.
Quine's use of the expressions 'obscure' and 'unintelligible'
is itself quite obscure. Later on we will discuss in detail
an important sense of 'obscurity'. But for now we can, I
think, get by without analyzing the notion of obscurity

any further.

First of all, let us assume that someone claims that
no sentence is analytic, i.e. that every sentence is
synthetic. Then Hempel's requirement will be automatically
fulfilled by the potential explanans constituted by state-
ments 1 through 3, and so such a person cannot put forward
the "no synthetic law" objection. But he will still reject
the potential explanans as an explanans for a different
reason, i.e. because statement 2 is false.

Suppose however, that a person claims that the notions
of analyticity and syntheticity are, in some sense, unintel-
ligible. Hempel's requirement that there be at least one
synthetic lawlike statement in the explanans will then also
be unintelligble. And thus such a person cannot put forward
the "no synthetic law" objection either, since he cannot
object to the explanans on the grounds that it violates an
unintelligible requirement. But he can object to it for a
different reason. He can object to it on the grounds that
premises 1 and 2 contain the unintelligible predicate

'analytically true'. And presumably everyone would agree to

27

the requirement that every sentence in the explanans be
intelligible, as well as with the assumption that a sentence
is unintelligible if it contains an unintelligible predicate.

Thus, regardless of whether someone rejects the notion
of analyticity on the grounds that no sentence is analytic,
or on the grounds that the notion of analyticity is unintel-
ligible, he can still consistently put forward the argument
that the notion of analyticity has no explanatory power.

For those who accept the intelligibility of the notion
of analyticity, as well as the claim that some sentences
are analytic, there is a feature of the "no synthetic law"
objection which deserves mention. It seems that this objec-
tion will withstand any improvements which might be made in
the clarity of the notion of analyticity. The reason for
this is the fact that, since premise 1 seems clearly analytic,
the chances are minimal that a further clarification of the
notion of analyticity would show it to be synthetic; In
other words, since premise l is somewhere within the class
of paradigm cases of analyticity it could not turn out to
be synthetic without, as we say, "changing the meaning" of

the word 'analytic'.
Section 4: The Clarity of 'analytic'

The notion of analyticity and its cognates could be,
and has been, criticised on the grounds that it is not
clear. To say that the notion of analyticity is not clear

may just be a way of saying that it is difficult to

28

understand how a sentence can be true independent of the
way the world is (a problem discussed in Section 1). Or
it may be a way of saying that it is difficult to see how
the ascription of analyticity to a sentence justifies the
claim that it is true (see Section 2). Or it may be a way
of saying that it is hard to see how the claim that a sentence
is analytic explains anything about the sentence (see Section
3). However, there is a primary sense of 'unclear' in which
the notion of analyticity and its cognates could be said to
be unclear. The purpose of this section is to partially
explicate this primary sense of 'unclear', to examine the
claim that analyticity is unclear in this sense, and to examine
some of Quine's objections to the notion of analyticity in
terms of this sense of clarity.

I will limit myself to the characterization of clarity
as it applies to monadic predicates, not only because this
is sufficient for our purposes but also because of compli-
cations which arise when one attempts to define the notion
of the clarity of a sentence in terms of its well-formedness
and the clarity of its component predicates.

We first define the notion of a predicate F being clear
at an object a. Reflection on the way 'clear' is sometimes

used suggests that a piedicate F is clear s3 32 object 3 if

 

and only if either:

1. We believe it to be true that a is an F, or

2. We believe it to be false that a is an F, or

3. We know what to do to_gain evidence for the claim that a

is an F.

29

Strictly speaking, the clarity of a predicate at an
object is relative to both a person and a time. For a
predicate can be clear at an object for one person but not
for another, and it can be unclear for a given person at one
time but clear for that same person at a later time. But I
have omitted such explicit relativization for the sake of
simplicity.

Here are some examples to illustrate the definition.
If Jones has no hair at all then the predicate 'bald' is
clear at Jones since we believe (let us suppose) that it is
true that Jones is bald. And if Smith has a full head of
hair then 'bald' is also clear at Smith since we believe
(let us suppose) that it is false that he is bald. But if
a third man, Stevenson, has a small amount of hair such
that we don't want to say that it is true that he is bald
and also don't want to say that it is false that he is bald
then 'bald' is not clear at Stevenson. This is because we
don't know what to do to gain evidence for the claim that
Stevenson is bald. In fact, the claim that he is bald is
neither true nor false. It has no truth value at all. It
is no surprise that we don't know what to do to gain evidence
for the truth of a claim that has no truth value. With
respect to the question of Stevenson's baldness there is no
truth of the matter to be discovered. I don't want to
suggest that 'bald' couldn't be made more precise but only
that, as things now stand, this predicate is not clear at

Stevenson.

3O

Clause 3 in the definition of clarity reflects the fact
that in a case where we have no opinion about the truth
value of a sentence of the form 'a is an F' this lack of
opinion may be due to either one of two distinct factors.
It may be due to the fact that the predicate F is unclear
at the object a, in which case the sentence has no truth
value to be believed in. (It may also be due to the fact
that the singular term 'a' is not clear. But this need not
concern us. For our concern is with the clarity of semantical
predicates such as 'analytic', these predicates being appli-
cable to sentences. And sentences are typically named via
quotation mark names, these quotation mark names themselves
being clear since we know what they refer to.) On the other
hand this lack of opinion may be due simply to the fact that
although the sentence has a truth value we are not in pos-
session of any evidence for‘or against its truth, although
we know what to do to gain evidence for its truth. In this
latter kind of case the predicate is clear at the object a.
Consider some further examples. Many persons are such
that I don't have any opinion as to whether or not the pre-
dicate 'is elected president of the United States in the year
2000' applies to them. But this lack of Opinion is not due
to any unclarity in this predicate. The predicate is clear
at each object since for each object we know what to do to
gain evidence for the claim that the predicate applies to
that object; i.e. wait until the year 2000 and look in the

newspapers for sentences of the form ' has been elected

31

president of the United States', where the blank is filled
in by the name of the object in question.

By contrast consider, for example, the predicate 'is
a possible man'. Limiting our considerations to actual
objects we can ask whether this predicate applies to the
pencil on my desk, or to any actual non-man for that matter.
Or consider the predicate 'equals the number of possible
men'. Of what number is this true; i.e. how many possible
men are there? The trouble with these questions is not just
that we don't know the answer to them but that we don't know
what to do to gain evidence for an answer. Indeed there is
no answer to be had. And this is what makes these last two
predicates unclear.

In the case of an unclear predicate it is not sufficient
to say that what we need to do to gain evidence for whether
it applies to an object is to "reason" or "think about it".
This kind of injunction is not specific enough. It can
apply to most any unclear predicate. By contrast, with the
clear predicate 'is elected president of the United States
in the year 2000' we have not only the injunction "wait and
see", but also directions which are specific to this parti-
cular predicate.

It might be wondered why clauses l and 2 are required
in the definition of clarity. The answer is that they are
there to cover certain paradigm cases of clear predicates,
the observational ones. Take, for example, the predicate

'red'. It might be thought that 'red' fulfills clause 3,

32

and that therefore clauses 1 and 2 are not required. It
might be claimed that we know what to do to gain evidence for
the claim that a given object is red; i.e. look and see.
But this is also what we do to gain evidence for the claim
that it is green. And if we don't require the directions
for gaining evidence for the application of a predicate to
be specific to the predicate in question in the case of
observational terms but do require such specificity in the
case of non-observational terms then we could be justifiably
accused of a philosophically unjustified bias in favor of
observational terms when it comes to their clarity. This
would not sit well with those philosophers who consider the
process of "seeing with one's mind" that something is so
(e.g. "seeing" with one's mind that two words have the same
meaning) to be just as legitimate as seeing with one's eyes
that something is so (e.g. seeing that the object in front
of me is red). We don't want our definition of clarity to
be tied to a philosophical doctrine which says that gniy
observational terms are clear. Rather we want our definition
of clarity to be an explication of a sense of clarity which
we all use, regardless of our philosophical doctrinal
differences. Since most philosophers would agree that
observational terms are clear we want our definition to
reflect this fact.

Now it might be thought that clause 3 is sufficient to
capture the clarity of observation terms on the grounds that

the directions for gaining evidence for the claim that an

33

object is red is specific to the predicate 'red'. It might be
claimed that the directions consist of looking to see whether
the object in question has the same color as, say, a fire
engine. But then the question arises of whether the predicate
'red' is clear at that fire engine. What are the directions
for gaining evidence for the claim that the fire engine is
red? It seems clear that if we are to avoid both an infinite
regress and a kind of circularity which just brings us back
to the original object in question we have to include something
like clause 1 and 2 in order to take care of the agreed upon
clarity of observational terms. It might be thought that
clause 3 is sufficient to capture the clarity of observational
terms on the grounds that the directions I have for gaining
evidence for the claim that the object in front of me is red
consist of asking suitably placed members of my linguistic
community whether this object is red. (Notice that the
reference to my linguistic community is necessary since we
would not want to claim that 'red' is unclear to ms on the
grounds that some German speaker who knows no English neither
assents to nor dissents from the claim that the object in front
of us is red.) But this suggestion is actually equivalent
to clause 1 in the definition since I will not regard anyone
to be a member of my linguistic community unless he agrees
with my belief that 'red' applies to the (red) object in
front of us.

There is another feature of the definition which perhaps
deserves mention. Why not, it may be asked, replace the word
'believe' in clauses l and 2 by the word 'know'? The answer

is that if we did so then the clarity

34

of a predicate F at an object a would depend on the truth
(clause 1) or the falsehood (clause 2) of the sentence 'a

is an F'. Suppose, for example, that I see an object which
appears red to me at time t1. Suppose also that later on,

at time t2, I discover that I was the victim of some practical
joke which involved the use of trick lighting or a secretly
administered drug which made an object which was not red
appear to be red to me. Since knowledge implies truth I will,
at t2 claim that since the object was not red at t1 I didn't
know at t1 that it was red. But on the suggested amendment
of the definition I would have to conclude that 'red' was

not clear to me at t . But intuitively we want to say that

1
'red' was clear to me at t , even though my belief that the

object was red was a falselbelief. The clarity of a predicate
F at an object a should not depend on the truth value of the
sentence 'a is an F'.

Some further definitions: If a predicate is not clear
at an object then we will say that it is obscure at that

object. If a predicate is clear at an object then we will

also say that there is a search piecedure for the predicate

 

 

s: that object. If a predicate is clear at every object we

 

will say that there is (are) a search procedure(s) for the

 

predicate, and that the predicate is completely clear.

 

Some predicates are such that they have a search proce-
dure which is the same for each object at which they have a
search procedure at all. And this might lead one to believe

that the search procedure for a predicate is its meaning.

35

Consider, for example, the predicate 'chair'. Its search
procedure is the same at every object at which it is clear,
and consists of directions telling us to see whether the
object in question is used or could be used to sit on. And
this example makes it sound like the search procedure for a
predicate is its meaning. Perhaps it is in this case. But
it isn't in the general case. As a counterexample consider
the predicate 'true'. This predicate has a different search
procedure at the sentence 'snow is white' than it does at the
sentence 'grass is green'. That is, we do different things to
gain evidence for the claim that grass is green than we do to
gain evidence for the claim that snow is white. But the
predicate 'true' doesn't have a different meaning depending
on the sentence to which it is applied.

Since clarity is a matter of degree we would like to be
able to compare two predicates and say that one predicate
is clearer than another. If the sets of objects at which
predicates were either clear or obscure contained only a
finite number of members then we could simply define a
predicate F as being clearer than a predicate G just in case
the number of objects at which F is clear is greater than
the number of objects at which G is clear. Moreover, the
notion of clarity would also be a metric concept which would
allow us to define what it means to say, for example, that
one predicate is three times as clear as another. But
typically, predicates are clear or obscure at an infinite
number of objects. However, we can specify a sufficient

condition for the relative notion of one predicate being

36

clearer than another. We can say that a predicate F is
clearer than a predicate G if the set of objects at which

G is clear is a proper subset of the set of objects at
which F is clear. However, this proper subset condition
doesn't seem to be also necessary for F being clearer than
G. Intuitively it seems that 'bald' is clearer than 'heavy'
because there seem to be "more" objects at which 'heavy' is
obscure than at which 'bald' is obscure. Yet there are
people who are clearly bald but at whom 'heavy' is obscure.
And thus, if we were to make the proper subset condition a
necessary one we would then be led to the counterintuitive
conclusion that 'bald' is not clearer than 'heavy'.

Our sufficient condition for F being clearer than G is
meant to apply to many kinds of predicates. But since we are
interested here only in how it applies to the notion of analy-
ticity I will just concentrate on those applications which
are relevant to our purposes.

The basic idea is that if every object at which G is clear
is also an object at which F is clear, but that F is clear at
objects at which G is obscure, then F is clearer than G. This
can be illustrated by the fact that 'true' is clearer than
'analytically true' (as will be argued in Section 1 of Chapter
2). Every sentence at which 'analytically true' is clear is
also a sentence at which 'true'is clear. However, there are
sentences at which 'analytically true' is obscure~but at
whiCh 'true' is clear. One such sentence is mentioned by

Quine. He says:

37

I do not know whether the sentence 'Everything

green is extended' is analytic. Now does my

indecision over this example really betray an

incomplete understanding, an incomplete grasp of

the "meanings”, of 'green' and 'extended'? I

think not. The trouble is not with 'green' or

'extended', but with 'analytic'.

The trouble is the obscurity of the predicate 'analytic'
at the sentence 'Everything green is extended'. On our
analysis of obscurity this means that we believe it to be
neither true nor false that the sample sentence is analytic,
and we don't know what to do to gain evidence for the claim
that it is analytic. A sentence of the form 'All A's are
B's' can be said to be analytically true just in case the
meaning of the term in place of 'B' is included in the meaning
of the term in place of 'A'. Quine has often rejected the
notion of meaning on the grounds that the identity conditions
for meanings are not clear. This is as good an example as
any of what Quine can be taken to mean, although we have to
adjust the complaint, for the sake of this example, to inc-
lude not only obscurity of identity conditions for meanings
but also the obscurity of conditions for "partial" identity
of meanings, i.e. meaning inclusion.

In the light of the fact that we don't know what to do
to gain evidence for the claim that the sample sentence is
analytic Grice and Strawson16 miss the point when they say
that the indecision would still remain if we replace the
problematic word 'analytic' with the presumably unproblematic

word 'true'. Their claim is that we would be just as

undecided over whether the sample sentence is true as we

38

are over whether it is analytic. Maybe 50. But the point
that they miss is that 'true' is clear at the sentence
'Everything green is extended', while 'analytic' is not.

In order to increase our evidence for the claim that the
sample sentence is true we need only gather up more and more
instances of things which are both green and extended. But
what are we to do to gain evidence for the claim that the
sample sentence is analytic. Needless to say, we cannot
answer by saying that we need only ask ourselves whether it
is logically possible for something to be both green and
extended, since the relevant notion of logical possibility
is just as obscure as the notion of analyticity.

There is an important question concerning Quine's
complaint that the notion of analyticity is obscure. And
that question is this: should we interpret his complaint
as meaning that the predicate 'analytic' is completely
obscure, i.e. obscure at every sentence? Or should we
interpret it to mean that while it is clear at a few sen-
tences, such as the sentence 'All bachelors are unmarried',
it is not clear enough to be a candidate for scientific
utility? I interpret his complaint in the second way, and
for the following reasons.

First of all, if the predicate 'analytic' were completely
obscure then it would be like the genuinely meaningless
predicate 'glug'. And if this were the case then Quine
wouldn't be so good at distinguishing synonymy from mere

coextensiveness, nor at distinguishing logical truth from

39

the wider concept of analytic truth, nor at distinguishing
the theory of meaning from the theory of reference, nor at
seeing how various responses to his objections merely assume
the notion of analyticity in a disguised form, not at being
able to define the notions of analyticity, meaning, and
synonymy in terms of each other. Quine knows very well what
is intended by that notion of analyticity which is defined
as "turnable into a purely logical truth by substitution of
synonyms for synonyms". It is just that his standards of
clarity are very high.

In the second place, my interpretation of some of Quine's
complaints against the notion of analyticity allows him to
acknowledge the soundness of a very powerful argument of
Grice and Strawson,l7 for the claim that 'analytic' is
meaningful at least to some extent. They point out that
philosophers apply and withhold 'analytic' to more or less
the same sentences, and that they are able to do this for a
potential infinity of new sentences which they haven't been
explicitly taught to characterize as either analytic or
synthetic. The idea is that in order for this to be possible
there must be some rule, no matter how implicit, and no matter
how poor philosophers may be at making it explicit, guiding
our use of the word 'analytic'. And this is enough to show
that 'analytic' is not completely meaningless. In other
words, this shows that there is a rule for its use.

Harman has responded to this argument by saying that it

is merely an application of the paradigm case argument.18

40

He doesn't think much of such arguments because a similar
argument would show that there were once witches. That is,
since there were once people to whom most people (let us
suppose) applied the word 'witch', witches once existed.

Now it is true that paradigm case arguments applied to
a predicate F are usually taken to prove that there exists
something to which the predicate applies. When G. E. Moore
held up his famous hand, he did so to prove that there
existed something to which the predicate 'material object'
applied. But this is not the kind of argument which Grice
and Strawson have given. Their goal is not to show that
analytic sentences exist, but rather to show that the predi-
cate 'analytic' is meaningful. The crucial part of their
argument is not the claim that there are sentences to which
philosophers apply the word 'analytic' but rather that they
are able to apply this word, in a non-arbitrary way, to new
sentences which they have not been taught to call analytic.
What we have to keep in mind here is the distinction between
the existential claim that there exist things to which a
predicate applies, and the different claim that the predicate
is meaningful. Referring back to Harman's example, although
we now believe that witches never existed, the predicate
'witch' remains meaningful to this day.

Harman responds to this kind of argument by saying that
all it shows is the meaningfulness of the predicate 'seems
to be analytic', which does not mean the same as the predicate

19

'is analytic'. These two expressions do indeed mean

41

something different. But the point is that the expression
'analytic' appears in both of them and means the same in

both cases. I don't mean something different by the word
'green', for example, when I claim that a certain object

seems green than I do when I claim that it is green. And

it just begs the question to admit this in the case of 'green'
and disallow it in the case of 'analytic' by claiming that
'seems analytic' should be taken as a simple unanalyzable
predicate not dividable into 'seems' and 'analytic'.

Thus, Grice and Strawson's argument still stands. And
we can grant the soundness of this argument while still
maintaining the view which I have attributed to Quine, i.e.
the view that the concept of analyticity, while meaningful
to some extent, is not clear enough to be useful.

A third reason in favor of my interpretation of some of
Quine's remarks is that it removes a serious apparent incon-
sistency in his writings. The apparent inconsistency is
that while Quine has often said that he doesn't understand
the concept of synonymy he also says that he understands it
when it is created by fiat in the case of explicit conven-

tional notational abbreviation.20

Grice and Strawson point
out that Quine is like a person who says that he understands
what it means to say that two things fit together when they
have been expressly made for the purpose of fitting together,

but that he doesn't understand what this means when two

things just accidentally happen to fit together.

42

When Quine says that he understands synonymy when it is
created by fiat in the case of explicit notational abbre-
viation, you can take his comment at face value, or you can
regard it as a slip of the mind on his part. I prefer to
take it at face value, for the first of the reasons which I
gave above, i.e. Quine's obvious understanding of what is
intended by the concept of synonymy.

Quine's admission that he understands synonymy when it
is created by fiat is consistent with his other objections
if we interpret some of these other objections as amounting
to the claim that theeconceptof'synonymy is clear to some
extent but not clear enough to be useful.

I might remark that the kind of analyticity which Quine
admits to understanding is limited to 'analytic' defined as
'turnable into a purely logical truth by substitution of
synonyms for synonyms' where these synonyms are limited to
those created by fiat, as in the case of explicit conventional
notational abbreviation. He has never admitted to understanding
the notion of synonymy when it is a case of pre-existing
synonymy, although the argument of Grice and Strawson,
referred to above, shows that if he is to be consistent then
he ought to admit to an understanding of 'synonymous' when
it applies to sentences exemplifying pre—existing synonymies
as well. But if he made this further admission he is still
free to complain, as I have interpreted him as doing, that
'synonymous' (and therewith 'analytic' defined as 'turnable

into a purely logical truth by substitution of synonyms for

43

synonyms') is too obscure (in our sense) to be useful. He
would also still be free to complain that he doesn't under-
stand 'analytic' when it is defined as 'believed true in
virtue of meaning' and when it is defined as 'true in virtue
of meaning'. Moreover, it might be remarked that while one
can admit to understanding 'analytic' defined as 'turnable
into a logical truth by substitution of synonyms for synonyms'
it doesn't follow that analytic sentences, so defined, are
true in virtue of theirmmeaning. No doubt they are true in
virtue of meaning ii the purely logical sentences into which
they are transformed are true in virtue of ihsis meaning.

In Section 1 we have discussed the difficulty involved in
claiming that any sentence is true purely in virtue of its
meaning.

Obscure words can be useful for some purposes. They
are useful to the advertiser because they permit the ever
hopeful consumer to interpret the advertisement in accord
with his desire to get high quality merchandise at a fair
price, and at the same time are too obscure to be the basis
for a successful lawsuit against the manufacturer on the
grounds of false advertising.

The purposes relative to which the obscurity of analy-
ticity is to be judged are presumably the purposesof science,
i.e. explanation and prediction; But since not even the
predicates of science are completely clear, the question
arises of how clear a predicate has to be to be scientifi-

cally useful. I don't know the answer to this question, and

44

neither does anyone else. We can ask, however, what Quine
would take to be an acceptable standard of clarity. We
might seek an answer by considering what it would take,
according to him, to remove the obscurity.

Also we find it argued that the standard of

clarity that I demand for synonymy and analy-

ticity is unreasonably high; yet I ask no more,

after all, than a rough characterization in

terms of dispositions to verbal behavior.21

The expression 'verbal behavior' is pretty vague. But

we know from Quine's work in Word and Object that what he

 

has in mind here are dispositions to assent and dissent from
queried sentence tokens. Now I doubt that this penchant for
verbal behavior, as opposed to meanings, is due to some kind
of implicit naive ”materialism" or "sensationalism". Rather
it is due to the implicit (and correct) assumption that the
predicate 'x dissents from (or assents to) token S at time
t' is less obscure (in our sense) than the predicate 'S is
analytic for x at time t'.

The reference to verbal behavior in connection with
synonymy and analyticity pops up again when he says,

One quickly identifies certain seemingly trans-

parent cases of synonymy, such as 'bachelor'

and 'man not married', and sense the triviality

of associated sentences such as 'No bachelor is

married'. Conceivably the mechanism of such

recognition, when better understood, might be

made the basis of a definition of synonymy and

analyticity in terms of linguistic behavior.22

Now I think that we can all agree that the mechanism of
such recognition is not understood by anyone at present.

Moreover, in ordinary (i.e. unphilosophical) contexts we do

use the word 'obscure' in reference to at least some

45

processes whose workings we don't understand. Thus it
would seem that epithets such as 'obscure', 'unclear',
'dim', and 'mysterious' as applied to the process of recog-
nizing analyticity is not out of place and is not the result
of some idiosyncratic standard of clarity on Quine's part.
But if this is what he means by calling synonymy and
analyticity obscure then I am puzzled as to why he doesn't
find to be equally obscure and mysterious those behavioral
criteria he is constantly hankering for. Consider some
paradigm cases of "behavioral" or "empirical" terms; terms
such as 'soft', 'round', and 'green' (about which, you recall,
there was no trouble). To the best of my knowledge no one
understands the mechanism by which we recognize two objects
as having the same color (e.g. the same shade of green). Why
should this process be considered less mysterious than the
process by which we recognize that two predicates have the
same meaning (and are thus synonymous)? Why should the
process of seeing with one's eyes that two objects have the
same color be considered as less obscure than the process of
"seeing" with one's mind that two terms are synonymous (this
latter process being one which we call 'understanding')?
There is the bare possibility that Quine's penchant for
behavioral terms is merely a matter of taste and philosophical
temperament. But, methodologically speaking, this is an
explanation of last resort. Moreover, there is a more
promising explanation of why empirically minded philosophers,

including Quine, favor empirical terms; viz. these terms are

46

amongst the clearest (in our sense of clarity) that we
have.

Thus, I can see no clear and plausible way of inter-
preting some of Quine's complaints about 'analytic',
definable as "turnable into a logical truth by synonym sub-
stitution", other than to say that his charge consists, at
bottom, of the claim that the notion of synonymy is very
obscure (in our sense), and in particular, more obscure than
behavioral or empirical terms.

(A friend of mine, in an epistemology class, handed out
a list of sentences to his students with the instructions to
determine, for each of them, whether they were analytic or
synthetic. Every sentence on the list was extremely proble-
matic with respect to this characteristic. The implicit
purpose of this exercise seems to have been to exhibit the
obscurity of the notion of analyticity.)

The question arises, however, of at what point along a
scale of obscurity a predicate becomes too obscure to be
acceptable. Perhaps Quine's standards of clarity are too
high. This is, I think, an open question at this stage of
philosophical progress (and a good one for further research).
However, there does seem to be a presumption in favor of the
claim that 'analytic' is too obscure to be a candidate for
scientific utility. There are an embarrassingly large
number of important sentences with the following character-
istics: we believe them to be true, we don't want to justify

them by appeal to experience, we don't know what to do to

47

gain evidence for the claim that they are analytic. In
other words, 'analytic' is obscure at all these sentences.
As examples I need only mention the axiom of infinity, the
axiom of choice, and the claim, at least in the mouths of
some, that people seek only what they think will give them
happiness. The reader can doubtless conjure up his own
favorite examples. We cannot explain our belief in the
truth of these sentences by appealing to the fact that they
are analytic because, since 'analytic' is obscure at these
sentences, the claim that they are analytic is neither true
nor false. Since so many interesting sentences are of this
kind it is no wonder that the concept of analyticity is
suspected of not being clear enough to be useful. Notice
that this criticism would remain even if the difficulty
discussed earlier were to be overcome (i.e. the difficulty
involving the sense of 'explanation' used in the claim that
the ascription of analyticity to the sentence "All bachelors
are unmarried” explains either our belief in, or knowledge
of, its truth).

A favorite way in which the friends of analyticity
respond to the criticism that the notion of analyticity is
obscure (in our sense) is to blame the obscurity of this
notion on the obscurity of the words in the sentence at
which analyticity is obscure. We will examine this response
in Chapter 2. But before we do there remains another

criticism of analyticity which needs to be looked at.

48

Section 5: Whether the Notion of Meaning is Needed to Describe

Successful Translation and Communication

Another objection against the notion of meaning (in terms
of which synonymy and analyticity can be defined) is that
this notion is simply not needed to describe the successful
translation of one language into another, and that it is
also not needed to describe the process of successful com-
munication. If this objection holds up; then it is but a
short step to the further conclusion that meanings don't
exist. For if a kind of entity is not needed in the service
of our descriptive or explanatory purposes then nothing more
is required as evidence for the claim that entities of that
kind don't exist. Our only justification for the claim that
electrons exist is the fact that appeal to electrons is
required to describe and explain what we want to. (Notice,
incidentally, that the claim that meanings don't exist is
different from the claim that all sentences are synthetic,
since one might hold the view that although meanings exist,
no two expressions of the same language have the same
meaning.)

The view that meanings are not needed is at the base of
Quine's philOSOphy of language, as expressed in ngd 33g
Object, and elsewhere throughout his work. On this view
the process of translation is one which is best described
not as preservation of meanings but as preservation of
reference. More specifically what is aimed at in a good

translation from one language into another is not

49

intensional isomorphism but, at most, simply extensional
isomorphism of a kind to be illustrated below. (Even
extensional isomorphism may, for Quine, be too strong as an
ideal (given his views on the inscrutability of reference)
implying, as it does, the autonomy of syntax, i.e. the
possibility of an independently ascertainable claim about

the syntactical structure of a radically foreign language;
"independently ascertainable" in the sense that the syntactic
structure of the radically foreign language can be determined
independently of theories about the reference of its compo-
nent words and sentences. But the issue of the inscrutability
of reference is not my concern here.)

An old fashioned pragmatic argument against the claim
that meanings are needed to describe what goes on in the
process of translation is implicit in Quine's view on the
indeterminacy of translation. For example, in discussing
propositions as the meanings of sentences he says this:

...if the posit of propositions is to be taken
seriously, eternal sentences of other languages
must be supposed to mean propositions too; and
each of these must heidentical with or distinct
from each proposition meant by an eternal sentence
of our own, even if we never care which. Surely
it is philosophically unsatisfactory for such
questions of identity to arise as recognized
questions, however academic, without there being
in principle some suggestion of how to construe
them in terms of domestic and foreign dispositions
to verbal behavior. ...For insofar as we take such
a posit seriously, we thereby concede meaning,
however inscrutable, to a synonymy relation that
can be defined in general for eternal sentences of
distinct languages as follows: sentences are
synonymous that mean the same proposition. We
would then have to suppose that among all the
alternative systems of analytical hypotheses of
translation which are compatible with the totality

50

of dispositions to verbal behavior on the part

of two speakers of two languages, some are

"really" right and otherswrong on behaviorally

inscrutable grounds of propositional identity.

...The very question of conditions for identity

of propositions presents not so much an unsolved

problem as a mistaken ideal.2

A premise of the pragmatic argument implicit here is that
a difference in meaning, in order to be a difference, has to
make a difference in verbal behavior. Another implicit
premise which Quine has often argued for is that alleged
differences in meaning really don't show up in differences
in verbal behavior.

On Quine's view, what a good translation preserves is
not sameness of meaning but sameness of reference. As he
puts it (here Quine is talking about an artificial language,
although the point made applies equally well to either a
foreign natural language or to expressions used by two
speakers of the same language):

Being a new invention, the language has to be

explained; and the explanation will proceed by

what may be called formation and transformation

rules. These rules will hold by arbitrary fiat,

the artifex being boss. But all we can reasonably

ask of these rules is that they enable us to find

corresponding to each of his sentences a sentence
of like truth value in familiar ordinary language.

24

In the View of those who believe in meaning, what a good
translation preserves is meaning. An expression of a foreign
language should have the same meaning as its translation
into say, English. But, in line with Quine's view, one
can make the claim that the most that need be required for

one expression to be a good translation of another is that

their extensions be the same. Take, for example, a simple

51

atomic sentence of the form 'a is an F'. A good translation
of this sentence need be only an English sentence of the form
'b is a G', where 'b' refers to the same thing that 'a' does,
and 'G' applies to the same set of things that 'F' does.

A good translation is, by definition, one which will
enable us to communicate with the foreigner. The reason why
sameness of reference is sufficient is because this is all
that is needed to achieve successful communication. This can
be seen in terms of an example. Suppose the foreigner tells
me that all F's are dangerous. Suppose further that I have
two translations of F which are different in meaning according
to our intuitive semantics. According to the one translation
F is to be translated as 'creature with a kidney', while the
other one translates it as 'creature with a heart'. Supposing
these last two English predicates to be coextensive, it then
makes no difference which of these two translation I choose.
If I take the foreigner's information to heart and flee from
a creature with a kidney because I believe that it is dangerous,
I will also be fleeing from a creature with a heart, and vice
versa. And so the intuitive different in meaning between the
coextensive predicates 'creature with a kidney' and 'creature
with a heart' is a difference which makes no difference. I
see no difficulty in the way of extending this basic idea to
complicated expressions and to expressions which don't refer
to observables.

Another interesting point to bring out in this connection
is that ordinary talk about a lack of communication or a

misunderstanding between two people as being due to a

52

difference in meaning which the two parties attach to a word
will not support the philosopher's notion of meaning. If
you look more closely at cases of what is called a purely
verbal dispute you will find that what is really happening
is merely that one person is using a word in such a way that
it refers to something different than it does for the other
person.

For example, if the ordinary person is disagreeing with
the biologist's claim that no fish is a mammal he may even-
tually come to discover that the disagreement is a merely
verbal one, which both the plain person and the biologist
will describe by saying that they are using the word 'fish'
to mean two different things. But all that this amounts to
is that they are using the word 'fish' to refer to different
things. The plain person is using the word in such a way that
it includes whales in its extension, while the biologist is
not. The moral is that ordinary language cannot be used to

support the philosopher's more refined notion of meaning.

Chapter 2
Defenses of Analyticity

The purpose of this chapter is to examine various things
which have been, or could be, said in defense of the analytic-

synthetic distinction.

Section 1: Shifting the Blame for the Obscurity of

'analytic' onto Other Words

A prominent response to the criticism that 'analytic'
is obscure is to ”explain away" this obscurity in terms of
the obscurity of the words in the sentence about whose
analyticity we are in doubt. If, for example, we are in
doubt as to whether the sentence 'All men are mortal' is
analytic, and we don't know what to do to gain evidence for
the claim that it is analytic, then it sounds plausible to
say that this is due to the fact that 'man' (and perhaps
'mortal') is not completely clear, and that since most words
of ordinary language are unclear to some extent it is no
wonder that 'analytic' will be obscure at those sentences
which contain such words.1 This kind of response is made by
Grice and Strawson in responding to Quine's complaint about
the obscurity of 'analytic' at the sentence 'Everything green
is extended'.

The indecision of 'analytic' (and equally, in this

case, the indecision over 'true') arises, of course,

from a further indecision: viz., that which we feel
when confronted with such questions as "Should we

count a psiss of green light as extended or not?"

As is frequent enough in such cases, the hesitation

53

S4

arises from the fact that boundaries of

application of words are not determined by

usage in all possible directions. But the

example Quine has chosen is particularly

unfortunate for his thesis, in that it is only

too evident that our hesitations are not here

attributable to obscurities in 'analytic'. It

would be possible to choose other examples in

which we would hesitate between 'analytic' and

'synthetic' and have few qualms about 'true'.

The crucial part of their response comes when they go
on to say:

But no more in these cases than in the sample

case does the hesitation necessarily imply any

obscurity in the notion of analyticity; since

the hesitation would be sufficiently accounted

for by the same or similar kind of indeterminacy

in the relations between the words occuring within

the statement about which the question, whether

it is analytic or synthetic, is raised.
In contrast to what Grice and Strawson say in the first part
of this quotation we have already seen the sense of 'clear'
in which 'true' is clear at the sentence 'Everything green
is extended' while 'analytic' is not. What is relevant for
our purposes here, however, is what they say in the second
part of this quotation. When they say (without argument)
that the hesitation over whether Quine's sample sentence is
analytic is sufficiently accounted for by the indeterminacy
in the relations between the words occuring in the sample
sentence, what relations do they have in mind here? If the
relation in question is that of meaning inclusion then they
are just begging the question of the intelligibility of this
relation. And if the relation is that of referent inclusion

then indeterminacy in this relation is not sufficient to

account for the indecision over whether the sample sentence

55

is analytic. For even if the relation of referent inclusion
were not indeterminate there might still be hesitation over
whether the sample sentence is analytic. In other words we
lwould, in such a case, be sure of the truth of the sample
sentence but be undecided over its analyticity.

This question deserves further discussion. For if we
can successfully blame our lack of a search procedure for
'analytic' (i.e. at those sentences at which it is obscure)
solely on the obscurity or unclarity of the words which
occur in the sentence then we will be able to save the notion
of analyticity from the charge of obscurity (in our sense
of obscurity).

In Section 4 of Chapter 1 we defined what it means to
say that a predicate is obscure at an object without regard
to a distinction which it is now necessary to make, i.e. the
distinction between extensional and intensional obscurity.

A predicate is extensionally obscure to some extent if and
only if there exist actual objects for which we don't know
what to do to reduce an indecision about whether or not the
predicate applies to these things. Here we use 'exist' in
the timeless sense in which it means 'has existed, does
exist, or will exist'.

A predicate is intensionally obscure to some extent
just in case there are possible but non-actual objects for
which we don't know what to do to reduce an indecision about

whether or not the predicate applies to them.

56

Since non-actual objects don't exist, this definition,
if taken literally, is useless. It is useless because the
definiens says that there exist objects which don't exist.

And because of this contradictoriness we can use the defini-
tion to conclude that no predicate is ever' intensionally
obscure. And if this were really the case then there would
never be any indecision over the analyticity of any sentence.
So let us not take the definition literally and instead try
to paraphrase it into a more useful form.

The basic idea behind intensional obscurity is that a
predicate is intensionally obscure to some extent if and only
if there might be an object for which we don't know what to
do to reduce an indecision over whether or not the predicate
applies to the object.‘ But to say that there might be such
an object is just to say that it is possible that there is
such an object. And the relevant sense of the expression
'it is possible that' is just as obscure as the notions of
synonymy and analyticity which we are trying to rescue. This
is no surprise since the relevant notion of 'possibility'
is just the notion of logical possibility (to be distinguished
from the narrower notions of physical and technical possibility).

Thus, since the notion of intensional obscurity
ultimately presupposes a notion which is just as obscure as
the notions of synonymy and analyticity it is hard to see
what is being accomplished by appealing to the former notion

as a way out of difficulties with the latter notions.

57

The question arises of whether the relatively unpro-
blematic notion of extensional obscurity is ever sufficient
to account for the obscurity of 'analytic'. (Notice,
incidentally, that the obscurity of 'analytic' itself is
extensional obscurity. The sentences at which 'analytic'
is obscure are not merely possible sentences but actual
sentences. This is so because we consider a sentence of
spoken language as a sequence of phonemes; a sequence which
exists even though no token of the sentence is ever uttered.5)

If 'analytic' is obscure at a sentence then we cannot
conclude that the predicates in the sentence are extensionally
obscure. However, if the predicates in a sentence are
extensionally obscure to some degree or other then 'true'
will also be obscure at that sentence. (As Quine puts it,
"Attribution of truth...to 'Snow is white'...is every bit
as clear to us as attribution of whiteness to snow.")6 In
such a case 'analytic' will also be obscure at the sentence
in question. But in general we cannot explain away the
obscurity of 'analytic' at a sentence in terms of the exten-
sional obscurity of the predicates in the sentence since
there are sentences whose predicates are extensionally clear
but at which 'analytic' is obscure.

In order to illustrate this point consider the sentence
'All men are mortal'. Let us assume, for the sake of this
example, that the two predicates 'man' and 'mortal' are
both extensionally clear. Let us further assume that we have

amassed so many instances of things which are both men and

58

mortal that we are firm in our conviction that the sentence
'All men are mortal' is true. Yet it may happen that we

are undecided over whether this sentence is analytically
true. Suppose you are someone who is thus undecided. A
defender of analyticity might want to blame this indecision
on the obscurity of one of the two predicates in this
sentence. He might say, for example, that yourindecision

is due to the fact that the predicate 'man' is not completely
clear to you.7 But under our assumption that this predicate
is extensionally clear, the kind of obscurity which the
defender of analyticity is appealing to is intensional
obscurity. This can be seen by considering the kind of
things which a defender of analyticity might say to you in

an attempt to decrease your indecision about whether it is
analytically true that all men are mortal. He might say
something like the following: ”I know you don'tbelieve that
there actually are any immortal men but if there were an
object which was in all respects like a man except that it
was immortal would it still be a man?" To this question you
might reply that if there were such an object then it would
exist, and since all existent men are mortal it wouldn't be

a man. At this point, the defender of analyticity, seeing
that he didn't say what he intended to say, will try again.
He might say something like the following. "Surely you can
understand talkabout non-existent objects. You do it all

the time when you are reading fiction and telling fairy tales

to your children. Now what I want to ask you is whether

59

there is a possible, but non-actual, object which has all
the characteristics of a man except mortality but is
nevertheless still a man (i.e. would you still call it a
man)? Or, to put it another way, is it possible for there
to be such an object which you would call a man?" To this
question there are three answers you could give. You could
say 'no', in which case the defender of analyticity will
conclude that the sentence 'All men are mortal' is analytic
for you. If you answer 'yes' then he will conclude that
this sentence is synthetic for you. But you may say that
you don't know whether you would call such an object a man.
The defender of analyticity would then probably say that the
reason you don't know is due to the fact that the concept
of being a man is not completely clear to you. But clearly
what he has in mind here is intensional obscurity, a notion
which is no clearer than that of analyticity.

Now it is true that there are words which are exten-
sionally obscure (witness 'bald'). The existence of such
words will make the notion of truth extensionally obscure
also. I don't want to deny that the notion of truth is
obscure to some extent. At the moment I am concerned only
to illustrate that the kind of obscurity which must be
appealed to by those who attempt to explain away the exten-
sional obscurity of 'analytic' is, in general, a different
and much more problematic kind of obscurity than the kind of
obscurity required to explain away any extensional obscurity

of 'true'.

60

We have already mentioned that the question of how
clear a predicate has to be to be acceptable is an open
one. But since both 'analytic' and 'true' are predicates
applicable to existent sentences we can ask the more manage-
able question of whether 'true' is (extensionally) clearer
than 'analytic'. To say that it is is to say that every
sentence at which 'analytic' is clear is also a sentence at
which 'true' is clear, but that there are sentences at which
'true' is clear but at which 'analytic' is not. The first
part of this claim can be shown by showing that there is no
sentence whose truth we are undecided about (with no pro-
cedure for decreasing the indecision) but whose analyticity
we are either sure about or have a procedure for decreasing
the indecision. This can be shown in the following way.
Most of those who attack, and most of those who defend, the
notion of analyticity would, I think, agree to the principle
that if a sentence is analytically true (false) then it is
true (false). Thus, if we believe that a sentence is analy-
tically true (false) then we will believe that it is true
(false). And if we know what to do to gain evidence for
the claim that a sentence is analytically true (false) then
we will know what to do to gain evidence for the claim that
it is true (false). Also, it seems that if we believe it
to be false that a sentence is analytic then either we will
believe that it is true or believe that it is false or

know what to do to gain evidence for the claim that it is

61

true. A counterexample to this last claim may seem to be
provided by the sentence 'Stevenson is bald' (See Section 4
of Chapter 1 for the construction of this example.) It may
seem that we believe it to be false that this sentence is
analytic (in which case 'analytic' is clear at this sentence).
And it is the case that we believe it to be neither true nor
false that 'true' applies to this sentence and that we have
no procedure for reducing the indecision (in which case
'true' is not clear at this sentence). But this is not really
a counterexample since we don't believe it to be false that
this sentence is analytic. What we do believe is that it is
neither true nor false that this sentence is analytic. But
from this we cannot conclude that we believe it to be false
that this sentence is analytic. If someone asked us to give
our opinion as to the truth value of the claim that this
sentence is analytic we would give neither 'true' nor 'false'
as an answer. Rather we would simply say that such a claim
is neither true nor false. Thus, the sentence 'Stevenson is
bald' is a sentence at which neither 'analytic' nor 'true'
is clear.

In order to show the second part of the claim that 'true'
is clearer than 'analytic' we need only to look at those sen-
tences (such as 'Everything green is extended', as well as

many others) at which 'true' is clear but 'analytic' is not.

62

Section 2: The Utility of Analyticity in an Ideal

Language

There is a move which might be considered to be a
defense, in some sense, of the notion of analyticity. The
basic idea behind this move seems to be implicit in much
of Carnap's work in the construction of artificial languages.
Consider in particular the following passage in which Carnap
is considering the construction of a formal language system.

Suppose he (i.e. the author of the system) wishes

the predicates 'BI' and 'R' to correspond to the

words 'black' and 'raven'. While the meaning of

'black' is fairly clear, that of 'raven' is rather

vague in the everyday language. There is no point

for him to make an elaborate study, based either on

introspection or on statistical investigation of

common usage, in order to find out whether 'raven'

always or mostly entails 'black'. It is rather his

task to makeup his mind about whether he wishes the

predicates 'R' and 'BI' of his system to be used in

such a way that the first logically entails the

second. If so, he has to add the (meaning) postulate

'(x)(Rx.D le)‘ to the system, otherwise not.

A preliminary comment that needs to be made here is that
in constructing a formal language system the use of artificial
symbols (such as 'BI' and 'R') which are not predicates of a
natural language is theoretically unnecessary. They are only
used for expository and pedagogical purposes. In this
passage 'B1' and 'R' could be replaced by 'black' and 'raven'
without affecting its content. I make this preliminary comment
because it seems not to be understood by some philosophers
who, not understanding it, proceed to dismiss formal philo-

SOphy as mere "symbol juggling" which has no relevance to

philosophical problems. (Ryle, for example, in a review of

63

Carnap's Meaningand Necessity practically fulminates over

 

Carnap's use of the word '1anguage' to describe his formal
languages, and insists instead that they be called 'codes'.9
(If it is not the use of artificial symbols which bothers
Ryle then what is the point of his insistence that Carnap's
formalized languages be called 'codes'?))

The basic idea behind the thoughts in the quoted passage
is that since natural languages have certain undesirable
features, an ideal language bereft of these features would
better serve philosophical purposes. One of the undesirable
features of natural languages is the fact that there are too
many sentences at which 'analytic' is obscure (in our sense).
In an ideal language however, 'analytic' would be clear at
every sentence of this language. A typical way to achieve
this clarity is to give a recursive definition of 'analytic'
for the sentences of this artificial language. This defini-
tion will be true by fiat, the artifex being boss. Thus
there is no need to apply behavioral or sociological tests
to discover pre-existing synonymies. Assuming then that
every term used in the definition of 'analytic' for the
formal language is itself clear, 'analytic' will also be
clear at every sentence of the constructed language.

But the question I want to raise is this: why should
there be any analytic sentences in an ideal language at all?
Another way to put this question is to ask why an ideal
language which has meaning postulates is better than one

which doesn't? (It might conceivably be replied that this

64

question is irrelevant since it is the philosopher's task
merely to explicate concepts, not to show why they are
needed. But certainly if a concept is not philosophically
or scientifically useful then there is not much point in
explicating it.)

One answer to our question is that if we all spoke a
language which was ideal to the extent that 'analytic' was
clear at all its sentences this would eliminate the possi-
bility of misunderstanding, miscommunication, and purely
verbal differences of opinion. The way this would work
would be as follows. Suppose you and I disagree over the
truth of a sentence of the form 'All and only F's are G's'.
If, upon calculation, it turned out that this sentence was
analytic in the ideal language then any disagreement with
it would indicate that the party who doubted its truth was
speaking a different language (i.e. using one or the other
of the predicates 'F' or 'G' in a different sense) than the
one who believed in its truth. And thus the difference of
opinion would be merely a verbal difference and not a dis-
agreement about the "facts". To be more specific, suppose
that I am speaking a language in which a biconditional of
the above form is analytic, and that you disagree with the
truth of this biconditional. On the view we are considering
the situation would be described by saying that while we
are perhaps both using 'G' in the same sense, you must be
using 'P' in a different sense than I am. (This fits in

well with our earlier doxological definition of an analytically

65

true sentence as one which is believed true in virtue of
knowledge of meaning.)

But now how does this view square with our earlier
admission that in order for successful communication to
take place it is sufficient that you and I use 'F' in such
a way that its extension turns out to be the same, and that
sameness of sense which you and I attach to 'F' is not
required? One answer is that these two views are compatible.
It might be claimed that since sameness of sense guarantees
sameness of reference, the requirement that you and I attach
the same sense to a given word in our dispute merely guarantees
that you and I are using 'P' in such a way that its extension
is the same for both of us. On this View, while intersub-
jective coextensiveness of 'F' might occur without 'F' being
intersubjectively synonymous this is not something which
should be left to chance. The intersubjective coextensiveness
should be guaranteed beforehand by the intersubjective
synonymy. And this intersubjective synonymy is guaranteed
in turn by the fact that we speak the same language, i.e.
by the fact that the language I am speaking has all and only
the same analytic sentences as the language which you are
speaking.

The flaw in this view is contained in the previous sen-
tence; i.e. in the assumption that if the language I speak
has all and only the same analytic sentences as the language
you speak then this is enough to guarantee the igisisubjective

synonymy of each of the terms we use. But as a matter of

66

fact it doesn't provide such a guarantee. From the fact
that, e.g. a sentence of the form 'All and only F's are G's'
is analytic in both of our languages it simply doesn't
follow that I am using 'P', say, in the same sense in which
you are using it. In other words, from the fact that 'F'
and 'G' are isiissubjectively synonymous for both you and me
it doesn't follow that 'F', say, means the same for me as it
does for you. isiissubjective synonymy of 'F' with 'G'
doesn't guarantee isisisubjective synonymy of either 'F' or
'G'. Thus, the purposes of increasing the effectiveness of
communication and reducing misunderstanding are not served
even if we spoke an ideal language at all of whose sentences
'analytic' is clear. And it is difficult to see what other

purposes might be served by speaking such a language.
Section 3: Carnap's Empirical Criterion of Synonymy

The purpose of this section is to examine an empirical
criterion of synonymy given by Carnap in his article "Meaning
and Synonymy in Natural Languages"10 In this article Carnap
asks us to imagine a situation in which we have two field
linguists investigating the linguistic reggnses of a German
speaker by the name of Karl. The dispute between the two
linguists is whether Karl's word 'Pferd' should be trans-
lated as 'horse' or as 'horse or unicorn'. Carnap then
says:

Suppose, for example, that one linguist, after an

investigation of Karl's speaking behavior, writes
into his dictionary the following:

67

(l) .BISEQ: horse

while another linguist writes:

(2) BIEEQ’ horse or unicorn

Since there are no unicorns, the two intensions

ascribed to the word 'Pferd' by the two linguists,

although different, have the same extension. If

the extensionalist were right, there would be no 11

way for emp1r1cally dec1d1ng between (1) and (2).
Carnap then goes on to argue for the claim that there is a
way for empirically deciding between (1) and (2) as trans-
lations of 'Pferd'.

But before we go any further there is already a dif-
ficulty for a Quineian. In the very description of the
testing procedure the intelligibility of the notion of
intralinguistic synonymy is presupposed by the claim that
'horse' is not synonymous with 'horse or unicorn'. Does
Carnap think that those who demand a behavioristic criterion
for synonymy require it only for the case of interlinguistic
synonymy? Perhaps not. Perhaps Carnap set up the situation
in this way only for pedagogic purposes.

(Whatever Carnap's intent it may be worthwhile to
mention, in passing, that the one "kind" of synonymy is
neither more nor less problematic than the other. To say,
for example, that synonymy of two terms within a natural
language is less problematic than synonymy of terms between
two natural languages is like saying that we can understand
what it means to say that two cars have the same weight but

that we don't understand what it means to say that some car

has the same weight as some boat. The fact that most pe0ple

68

implicitly consider intralinguistic synonymy to be less pro-
blematic than interlinguistic synonymy is illustrated by the
fact that Quine has found it pedagogically useful to argue
for the unintelligibility of any kind of synonymy by using
the extreme example of translation from a language like

English into a radicaliy foreign language.)

 

At any rate we can avoid the presumption of the intel-
ligibility of intralinguistic synonymy by changing the
testing procedure from the way Carnap has described it. We
can just consider the test as a criterion for whether or not
'Pferd' is synonymous with horse. Let us suppose then that
the two linguists are agreed that 'Pferd' is coextensive with
'horse', but that they disagree over whether these terms are
synonymous.

Since, as Carnap points out, the extension of 'Pferd'
and 'horse' is the same, no response by Karl, affirmative or
negative, with respect to any actual thing will reveal whether
or not these two predicates are synonymous. We must; he says,
take into account not only the actual cases, but also possible
cases. We do this by getting Karl to consider kinds of things
which have no instances and then ask him whether he would be
willing to ascribe the predicate 'Pferd' to things of those
kinds. As Carnap says, the linguist must investigate Karl's
responses not only to cases believed by Karl to be physically
possible, but also to cases believed by Karl to be logically
possible (though perhaps not physically possible, i.e. ruled

out by the laws of nature which Karl believes to hold).

69

Suppose that one of the things that we get Karl to
consider is a thing like a horse but with a horn in the
middle of its forehead. What the linguist wants to find
out is whether Karl considers it logically possible for a
thing of this kind to be a Pferd. If he does consider it
to be logically possible then this is evidence that 'Pferd'
is not synonymous with 'horse', and if he doesn't consider
it to be logically possible then this is some evidence that
'Pferd' is synonymous with 'horse'. But how is the linguist
supposed to discover Karl's views on whether he considers it
logically possible for a thing of the described kind to be
a Pferd? Carnap suggests that the linguist should accom-
plish this by putting to Karl the German equivalent of the
question 'Is it logically possible for a thing of this kind
to be a PferdI. But this has the consequence that the
linguist must have evidence that some particular German
expression is synonymous with the English expression
'logically possible'. And how will he gather this evidence?
It is hard to see how he can use the criterion under
discussion since this criterion presupposes that we already
know what German expression is synonymous with 'logically
possible'.

Carnap seems to provide an answer to this objection
when he says that it is not necessary to put these modal
questions to Karl in German. He says that the linguist
can merely get Karl to consider things, by the use of

pictures, for example, which the linguist knows to be

7O

logically possible, and then ask Karl the question
'Pferd?'. But the question 'Pferd?' is ambiguous. It
can mean either 'Is it physically possible for a thing of
this kind to be a Pferd?" or 'Is it logically possible for
a thing of this kind to be a Pferd?'. (Perhaps Carnap is
implicitly assuming the false principle that if you show a
picture of something to Karl then Karl will not believe that
it is physically impossible for the thing pictured to be a
Pferd.) So in order to avoid the ambiguity of the question
'Pferd?' it seems that the linguist has to put the appro-
priate modal question to Karl in German after all. And in
order to do this the linguist must learn a great deal of
German. I say 'a great deal' because the notion of logical
possibility (i.e. analytic impossibility), being a technical
notion, must be taught to Karl. And in order to teach him
this we must be able to give him examples (e.g. the German
equivalents of 'bachelor' and 'unmarried man', etc.) along
with the German equivalents of all the other things we say
when we are trying to explain to English speaking students
the notion of analytic impossibility. But Carnap's prOposed
test for synonymy is, in effect, supposed to be a method for
learning German (i.e. it is a method for finding out what
German expressions are synonymous with what English expres-
sions).

Thus Carnap's proposed criterion for synonymy appears
to be subject to a kind of circularity. In order to apply

the synonymy criterion we must already know what it is

71

supposed to give us, i.e. knowledge of a great many of the
English-German synonymy relata.

This circularity can be avoided if Carnap's criterion
is applied only to those German expressions knowledge of
whose English synonyms is not required in order to apply
the criterion. If we thus limit the criterion by excluding,
on pain of circularity, its applicability to a certain class
of German expressions, and if the linguist must know the
English synonyms of the expressions in this excluded class
in order to apply the criterion to German expressions out-
side of this class, then the question arises of how the
linguist is to gain knowledge of the English synonyms of
the German expressions in the excluded class.

One answer is that the linguist can "go native", i.e.
he can learn the use of the expressions in the excluded
class in the same way that a native born German child would.
Afterwards he can then, without circularity, apply Carnap's
criterion to the non-excluded class of German questions by
putting, in German, the appropriate modal questions to Karl
(utilizing, of course, only those German expressions which
the linguist learned by going native, i.e. those expressions
in the excluded class).

The upshot of restricting Carnap's criterion in the
way we have described is that the linguist must somehow come
to know some of the English-German synonymy relata before
he can use Carnap's test for getting evidence of other

English-German synonymy relata.

72

It is instructive to compare Carnap's test for synonymy
with intelligence tests. One respect in which they are
similar is that just as the linguist must understand the
concept of analyticity so also must the designer of the
intelligence test understand the concept of intelligence.

A point of dissimilarity is that while Karl must understand
the notion of analyticity in order for Carnap's test to be
successfully applied to him, the one who takes an intelligence
test need not understand the concept of intelligence.

The fact that the linguist must understand the notion
of analyticity does not constitute a good objection to
Carnap's criterion. This can be seen by analyzing the
situation with respect to intelligence tests, these tests
being ones which we normally consider to be empirical or
behavioristic. While the factors which constitute what we
call intelligence are complex we do routinely make estimates
of people's intelligence. Moreover, a good estimate of
intelligence requires that a great deal of time be spent, in
many different contexts, with the person whose intelligence
we want to estimate. So called I.Q. tests are intended as
just a short cut method for getting this estimate. The
effectiveness of the test is judged by the extent to which
it gives the same output (i.e. who is intelligent to what
degree) as the more time consuming method. If the tests
were completely accurate then the expressions 'is intelligent'
and 'gets a high I.Q. score' would indeed be coextensive,

but they still wouldn't be synonymous, i.e. they wouldn't

73

mean the same thing. (This point is not understood by those
who, in their eagerness to be "scientific”, want to define
'is intelligent' as 'gets a high I.Q. score'. They sometimes
express this view by saying that intelligence is that which
is measured by I.Q. tests (a statement which, under the
assumption of complete accuracy of the tests, is true if the
first 'is' is interpreted as synthetic identity, but false

if it is interpreted as an analytic identity).)

The moral of this comparison between Carnap's criterion
and I.Q. tests is that not all empirical tests for the
application of a predicate are supposed to provide us with
the meaning of the predicate, or to make an originally
meaningless predicate meaningful. If we did view such tests
as procedures which make an originally meaningless predicate
meaningful then we would invite the justifiable complaint
that while we understand what objects the result of the test
attributes the predicate to, we don't understand what is
being attributed to them.

From a behavioristic standpoint, however, it might be
considered to be objectionable that Carnap's criterion for
synonymy requires that Karl also understand the notion of
analyticity. This kind of objection is characteristic of
those who like to style themselves as behavioristic. The
behavioristic ideal often seems to presuppose that the
subjects of an ideal behavioristic theory be considered as
being as bereft of human understanding as an electron or a

stone, and that consequently concepts such as that of

74

understanding are not legitimate parts of an ideal
behavioristic theory.

But fortunately for our purposes the question of
exactly what it means to be behavioristic, along with
the pros and cons of behaviorism, can be avoided. The
reason is that Quine's request for a behavioristic
cirterion of synonymy is, as I have interpreted it,
merely a request for what I have in Chapter 1 called a
search procedure for the application of the predicate
'synonymous', which will enable us to gather evidence
for synOnymy claims which we didn't have before the
introduction of the criterion.

Now it might seem as though Carnap has achieved this
goal at least to some extent, i.e. with respect to those
words which do not belong to what I have called the
excluded class. We have, for example, the word 'Pferd',
and we have no idea whether or not it is synonymous with
'horse'. We then apply Carnap's criterion to gain evi-
dence for or against the claim that it is so synonymous.

But it seems to me that this is somewhat of a hollow
achievement, and that Carnap's criterion doesn't really
provide what Quine is asking for. What he is asking for
is greater clarity of the predicates 'synonymous',
'analytic', and the like, as applied to expressions in our
own language as well as that of other languages. This
means that Carnap's criterion should be evaluated on the

basis of this question: given that one knows the language

75

(or at least the meanings of the words to which the
criterion is to be applied) as well as anyone, does the
criterion help us to gather evidence for or against a
synonymy claim which we didn't have before the introduction
of the criterion? As applied to Carnap's criterion the
answer to this question is clearly negative. Quine would
point out that application of Carnap's criterion to the
sentence 'Everything green is extended' will not help in
deciding whether or not this sentence is analytic. And so
also with any other sentence, of a language we know, about
whose analyticity we are in the dark.

So, in spite of the fact that we can partially avoid the
circularity criticism mentioned earlier, we have to conclude
that Carnap's criterion doesn't provide the increased clarity

that Quine is looking for.

Section 4: Grice and Strawson's Understanding-Believing

Criterion of Analyticity

Let us turn now to another proposed criterion for the
application of analyticity to the sentences of a natural
language. This criterion, as given by Grice and Strawson,
turns on the distinction between not understanding a sentence

12 The basic idea is that an analytically

and not believing it.
true sentence is one whose negation we don't understand,

while a synthetically true sentence is one whose negation
we do understand even though we may not believe, or assent

to, it. This definition, as it stands, needs amendment.

76

Otherwise every sentence of a language foreign to us will
be analytically true since every such sentence is one whose
negation we don't understand. So let us say that an analy-
tically true sentence is one which we understand but whose
negation we don't understand.

Grice and Strawson have this to say about the distinction
between not believing a sentence and not understanding it.

It would be rash to maintain that this distinction

does not need clarification; but it would be absurd

to maintain that it does not exist. In the face of

the availability of this informal type of explana-

tion for the notions of the analyticity group, the

fact that they have not received another type of

explanation (which it is dubious whether any expres-

sions ever receive) seems a wholly inadequate ground

for the conclusion that the notions are pseudo-

notions, that the expressions which purport to

express them have no sense.

Another advantage claimed for this explanation by
Grice and Strawson is that it breaks out of the family
circle of terms which Quine objects to; terms like 'possible',
'meaning', 'synonymous', 'necessary', 'semantical rule',
'meaning postulate', and 'analytic', all of which, according
to Quine, are equally obscure. Grice and Strawson don't
clearly explain what they mean by breaking out of the family
circle of terms. If they merely mean that their informal
explanation doesn't explicitly utilize any of the terms
which Quine has "officially" objected to, then their claim
is true but uninteresting. The more interesting question
is whether their explanation implicitly utilizes the disrepu-

table notions. At first glance it doesn't seem to. To not

understand a sentence is, on one view, to not know what to

77

do to gain evidence for its truth. To understand a sentence
but not believe it is to know what to do to gain such evi-
dence but to expect that if one attempts to gather such
evidence it will show that the probability of the sentence
being false is equal to or greater than that of its being
true. (To simplify matters I am ignoring the fact that
both evidence and belief is a matter of degree.) But the
distinction between not understanding a sentence and not
believing it does implicitly presuppOse the notions to which
Quine objects. This can be seen as follows.

Let us apply what I will hereafter call the understanding-
believing criterion to the sentence 'All bachelors are
unmarried'. Its negation is equivalent to 'There is a
married bachelor'. To understand this latter sentence is to
know what to do to gain evidence for its truth. Do we know
what to do to gain evidence for its truth? One answer is
'Yes, look for a married bachelor'. An answer to the question
of how we are to go about looking for a married bachelor,
which is in line with positive aspects of Quine's phiIOSOphy
of language, might go something like this. Use a questionaire
method and ask people both whether they are a bachelor and
whether they are married. If perchance we should find someone
who says that he is both married and a bachelor what should
we conclude? That we have found a married bachelor, or that
our informant is using the words 'bachelor' and/or 'unmarried'
in a different sense than most of us do? No doubt we would

say the latter, as Grice and Strawson have done in a similar

78

example.14 But then we are invoking the notion of sense

or meaning. Moreover, what can we say about the sense in
which most of us use the word 'bachelor'? All we can say,
it seems, is that we use the word 'bachelor' in the sense
that all bachelors are unmarried. This amounts to truth by
fiat as applied to the sentence 'All bachelors are married'.
For emphasis we might add that we are using the word
'bachelor' in the sense which implies being unmarried, or
which has being married as part of its very meaning, or in

the sense that it is impossible for a person to be both

 

married and a bachelor.

Thus, in spite of appearances to the contrary, it seems
that the understanding-believing criterion doesn't really
break out of the circle of terms which Quine objects to after

all.

Section 5: The Existence of Analytically True Sentences

In this section I would like to examine various things
which have been, or might be, said in answer to the question
"What guarantee do we have that any given sentence of a
natural language is analytically true?”.

One guarantee that there are analytically true sentences
might seem to be provided by the paradigm case argument.

There are two uses of such an argument. One of them was
discussed in Section 4 of Chapter 1. This use of the argument,
which we considered to be sound, consisted of showing that

'analytic' is meaningful; i.e. that there is a rule, no

79

matter how implicit, guiding our use of the word 'analytic'.
Let us call this kind of argument a 'proof of meaningfulness
paradigm case argument'. There is however, another and more
traditional use of the paradigm case argument which we may
call a 'proof of existence paradigm case argument'. This
kind of argument, as applied to the predicate 'analytic' is
designed to show that analytic sentences exist. I will be
concerned here only with the latter kind of argument.

As Quine points out sentences such as 'All bachelors
are unmarried', 'everything is self identical', and every
sentence of the form 'p or not p' have a feel that everyone

15 and which seem to be different in kind from

appreciates,
sentences such as 'All bachelors are over two feet tall' and
'Everything is self loving'. If we want to, we can invent

a word, say 'analytic', which we apply to sentences of the
first kind, and if someone asks us what 'analytic' means we

can point to sentences of that kind as examples of what we

take 'analytic' to refer to. If this is all we do then the
question of whether there are analytic sentences is out of
place. But then it will be a further, and distinct, question
of whether analytic sentences have the properties traditionally
attributed to them; i.e. the properties of being true purely

in virtue of meaning, or of being believed true in virtue of
knowledge of meaning, or of being true in all worlds, or even
of being true. And whether or not sentences such as 'Every-
thing is self identical' have these properties is something

that the proof-of—existence paradigm case argument cannot

prove.

80

Alternatively, we can define 'analytic' as meaning, say,
'true in virtue of meaning', or 'believed true in virtue of
knowledge of meaning'. But then it will be a further ques-
tion of whether there are any sentences which have the properties
invoked in the definition.

But we can't have it both ways. That is, we can't use the
proof-of—existence paradigm case argument to simultaneously
prove both that analytic sentences exist and that they have
the properties traditionally ascribed to them. Nor can this
be accomplished by using the proof-of—existence paradigm case
argument in tandem with the proof-of—meaningfulness paradigm
case argument. For all that the latter kind of argument shows
is that there is some rule, not matter how implicit, guiding
our use of the word 'analytic', not that analytic sentences
have the properties traditionally attributed to them.

The situation with 'analytic' is similar to that of
'witch'. We can use the proof-of—existence paradigm case
argument to prove that there were once witches. But we don't
thereby prove that they have the powers traditionally ascribed
to them. Alternatively, we can define 'witch' to mean 'someone
with supernatural powers etc.' But then it will be a further
question whether there ever were any witches.

I prefer to take the second kind of alternative with
respect to 'analytic'; i.e. to take 'analytic' to be defined
by the various things which have been said about analytic

sentences, and then consider it a further question as to

81

whether there are any sentences of the kind so defined.

One of the properties ascribed to analytically true
sentences by the definition considered in Section 1 of
Chapter 1 is the property of being true purely in virtue of
meaning. This means that for such a sentence, given that
it has the meaning which it does, it follows that it is
true. In other words, its meaning guarantees that it is
true. Let us try to see whether the sentence 'Everything
is self identical' has this property. Since this sentence
has, of course, the meaning which it does have the question
is then whether this sentence is therefore guaranteed to be
true. We have no such guarantee if it is possible for some
person to come up with a theory which has great explanatory
and predictive power and which has as one of its components
the claim that there are some things, strange theoretical
particles for example, which are not self identical. The
claim that this is not possible, and that therefore we have
a guarantee that 'Everything is self identical' is true
might be considered to be supported by the claim that the
sentence in question has a property attributed to it by what
we have called in Section 3 of Chapter 1 the doxological
definition of analyticity; i.e. the property of being believed
true in virtue of knowledge of meaning. In other words it
might be claimed that the sentence in question is guaranteed
to be true since if anyone did come along with a theory of
the kind mentioned above this would merely show that he is

using the predicate 'is identical with' in a different sense

82

from, i.e. he doesn't understand it in the same way as, the
sense which this predicate now has. And since this is so,

so the claim goes, this guarantees that the sentence
'Everything is self identical', with the meaning it now has,
is true. Moreover this shows that it is immune from revision,
provided only that its meaning doesn't change.

But now the question shifts to whether or not the sen-
tence in question does have the property attributed to it by
the doxological definition of analyticity? That is, if
someone believed and put forward the strange particle theory
would we have to conclude that he is using 'is identical
with' in a different sense, or with a different meaning,
than it now has? Here we have a question of the intersub-
jective synonymy of the predicate 'is identical with'; i.e.
does the proponent of the strange particle theory attach a
different meaning to this predicate than we do? If the
answer is 'yes' then we can conclude that the sentence
'Everything is self identical' is guaranteed to be true and
that it is analytically true. If the answer is 'no' then
we can conclude not only that the sentence in question is
not analytically true, but if there is a great deal of
evidence (albeit indirect) to support the strange particle
theory then we will also conclude that the sentence in
question is false. Well, what is the answer to this question
of intersubjective synonymy? Quine's answer, to paraphrase
him from a slightly different context, is that this question

represents not so much an unsolved problem as a mistaken

83

ideal.l6 This Quineian view seems to be butressed by our
earlier considerations involving what we have called the
obscurity of the notion of synonymy. For a likely answer
to the queStion of the intersubjective synonymy of 'is
identical with' is neither 'yes' nor 'no' but rather 'I
don't know and I have no way of finding out', in which case
the claim that 'is identical with' is intersubjectively
synonymous is neither true nor false.

But suppose that someone claims that the proponent of
the strange particle theory must be using 'identical with'
in a different sense than we do. He might even say, in
support of this claim, that by 'identical with' we just
msss that relation which holds between everything and
itself, and that anyone who says that there are things which
are not self identical must mean something different by
'identical with"than we normally do.

But suppose that the strange particle theory turns out
to be very successful in explaining and predicting what we
want it to. We would have to conclude that the strange
particle theory is true. Under these assumptions 3 question
would then seem to arise. If the proponent of the strange
particle theory is not using 'identical with' in the same
sense that we do then in what sense is he using it? Another
way to put this question is to ask whether his utterance is
translatable into a sentence whose meaning we already under-
stand. Suppose it isn't. Then we will have to say that

there are truths which we don't understand. Since as both

84

Carnap and Quine stress, analytic truth as well as truth

is relative to a language we can still say that in 92;
language the claim that everything is self identical is
analytically true, and therefore true, and that the claim
that 'Some things are not self identical' is analytically
false, and therefore false, in our language. But I don't
think that we, ever seeking the truth, would take much solace
in this. Instead, we would be more inclined, I think, to
take a cue from Carnap when he says that languages are to be
judged on the basis of their fruitfulness and utility. And
if so, then we will be inclined to say that the language of
the scientist is more fruitful than our language. But this
counsel will be just a source of frustration unless it is
possible for us to learn the language of the scientist. I
am willing to grant that since we have assumed that the
scientist's language (i.e. the relevant part of it) is not
translatable into ours, we do not, as of now, understand the
language of the scientist. But the question is whether we
could come to learn his language. I see no reason why not.
Just as the plain man starts out not understanding the
negation of the parallel postulate in geometry, (indeed he
might even claim that it is impossible for it to be true),
nor the claim that space is curved, nor the claim that some
things are neither waves nor particles, but can come to
understand them by learning the language of physics-mathematics
so also we can come to understand the language used by the

scientist within which he claims that some things are not

85

self identical. To understand a sentence is to know what

to do to gain evidence for it. In the case of the highly
theoretical sentence in question this means that we know how
to use it, in combination with other sentences, as part of

a theory from which we can derive testable consequences.

And if the proponent of the strange particle theory can do
this then so can we come to learn to do it.

Since we have no guarantee that the events described
in this little story won't happen so also we have no
guarantee that our present language (in which we have, for
the sake of argument, assumed that it is analytically true,
that everything is self identical) will remain adequate
and fruitful as a means for explaining and predicting events
in the world.

There is a third possibility yet to consider. Suppose
that the utterance of the scientist is translatable into a
language we already understand but that its translation is
not any of the sentences in our language which we consider
to be analytically false. This would not require us to give
up any sentence of our language which we consider to be
analytically true, in particular the sentence 'Everything is
self identical'.

Now if we are to have a guarantee that this sentence is
true then we would have to argue that this third possibility
is the only real possibility with respect to the story of the
strange particle theory. If, on the other hand, the utterance

of the scientist (which, we have assumed, has a great deal

86

of evidence in favor of its truth on the grounds of its
explanatory and predictive power) is translated into our
language as 'There are some things which are not self-
identical' then the claim that everything is self identical
is not guaranteed to be true. And if his utterance is not
translatable into any sentence of our language then there is,
I suppose, a sense in which it could be said that in our
language the sentence 'Everything is self identical' is
guaranteed to be true. But then since we would have to
conclude that our present language is inadequate such a
guarantee would, I suspect, lose its interest.

I know of no good argument for the claim that in the
context of our little story the third possibility is the
only real possibility. I know of no good argument for the
claim that our present language is guaranteed to be adequate
to any possible future development in our attempts to explain
and predict phenomena. I alSo don't know of any good argument
for the claim that if the scientist of our story disagrees
with the claim that everything is self identical then he
mgsi be using these words to mean something different than
we do. Indeed we can even imagine a situation in which the
scientist claims that his strange particles are not self
identical and in which he is not using these words to mean
something different than we do. Consider the following
conversation. Scientist: My successful theory has, as a
part, the claim that certain strange particles are not self

identical. Friend of analyticity: Well then you must mean

87

something different by 'identical with' than I do. You
must be speaking a different language than I am, since by
'identical with' I just msss that relation which exists
between everything and itself. Scientist: You are over-
working the word 'mean'. When you first learned the notion
of logical or numerical identity your teachers had to dis-
tinguish between this notion and that of, say, genidentity
or between logical identity and qualitative identity. And
in order to do this they said that logical identity is
"defined" as that relation which exists between everything
and itself. But they too were overworking the words 'mean'
and 'define'. All they were really doing was teaching you
the notion of logical identity by reference to their belief
that everything is self identical (a belief that I have good
reason to think is false). But since the truth of their
belief was not necessary in order for you to come to under-
stand the notion of logical identity you came to understand
this notion in spite of the falsity of their belief. I too
originally came to understand the notion of logical identity
in the way that you did. And when I claim that not every-
thing is identical I am using 'identical with' in the same
sense that you are. Indeed I can even say that I am using
'identical with' in the sense in which you (falsely) believe
it to hold of everything.

FOA: All right. I will, at least for the sake of argument,
grant that you are using these words in the same sense that

I am. But still there is a sense in which I don't understand

88

what it means to say that these strange particles of yours
are not self identical.
8: But if you understand what it means to say that every-
thing is self identical, and you also agree with the principle
that if you understand a sentence then you understand a
sentence which is equivalent to its negation then you will
have to agree that you understand what it means to say that
some things are not self identical.
FOA: Well, in this case I think that I might give up the
applicability of the principle to analytically true sentences.
For such sentences I think that even though I understand them
I don't understand their negations.
S: You are right. There is a sense in which you don't yet
understand what it means to say that these strange particles
are not self identical. This sense is hard to pinpoint, but
it seems to be hinted at in Wittgenstein's dictum that the
meaning of an expression is its use in-the language. And
admittedly you do not, as yet, understand the use of the
claim that strange particles are not self identical. That
is, you do not understand how to use this claim, in conjunc-
tion with other sentences, in order to derive successful
predictions of phenomena. But I can teach you how to do this.
Since we have no guarantee that the events described in
this dialogue won't happen we also have no guarantee that the
sentence 'Everything is self identical' is analytically true,

and thus true.

89

(Incidentally, since the events described in this
dialogue are not incomprehensible, and since the scientist
in the story could be described, by our present lights, as
an "illogical" person, if anyone could, Quine seems to be
wrong when he says that the doctrine of there possibly being
illogical peoples (he uses the word 'prelogical') is meaning-
less. Quine says that illogicality is a trait injected by

bad translators.17

Moreover, if we do accept Quine's view
that illogicality is a trait injected by bad translators, and
if we also assume that our language will always be adequate,
then we can be guaranteed that no obviously analytically true
sentence of our present language will be shown to be false.
For if every utterance of any possible future scientist is
destined not to be translated into a sentence of our present
language which we consider to be obviously analytically false
(on the grounds that such a translation would be the best
evidence that we could have that the translation is a poor
one) then those sentences which we consider to he obviously
analytically true are immune from refutation. And thus we
could be assured that such sentences are true purely in
virtue of their meaning; i.e. we could be assured that their
truth follows from the fact that they have the meaning which
they do have.)

We have been assuming that, at least for some sentences
S, the friends of analyticity require that sentences of the
form '8 is analytically true' be guaranteed to be true, and

thus that S be guaranteed to be true (provided, of course,

90

that its meaning remains constant). But have they accepted
such a requirement? And do they need to? Historically, it
would seem that the friends of analyticity have implicitly
accepted this requirement. Analytically true sentences were
supposed to be, in some sense, necessary. They were also
supposed to be known a priori. (And thus, since they
could be known independently of experience, no experience
could refute them. And since no possible experience could
refute them they were guaranteed to be true.) But all of
this could conceivably be accomodated to the view that no
given sentence of the form '8 is analytic' is guaranteed to
be true. One could imagine a philosopher holding the view
that all the friends of analyticity Essd be committed to
whether they realized it or not) are the hypothetical claims
that ii a sentence is analytically true then it is known
a priori, and ii a sentence is analytically true then it is
necessary, and ii a sentence is analytically true then it has
no content, etc., etc. Such a philosopher might express the
view that no sentence of the form 'S is analytic' is guaranteed
to be true by saying that no sentence of this form is
necessary, or analytic, or known a priori. He might say
that all a philosopher should be concerned with is analyzing
the notion of analyticity and relating it to other notions,
but not with making or proving any claim of the form '8 is
analytic'.

I can imagine someone taking this 'if p then q' view

of philosophical activity. It has its metaphiIOSOphical

91

attractions, not only with respect to the notion of analy-
ticity but also with respect to other notions philosophers
are prone to analyze. Whether it could be maintained as a
general view of what philosophical activity should be,
without ultimately invoking spurious distinctions and ques-
tion begging assumptions, is a difficult question. For our
purposes however, it is sufficient to point out that if the
friends of analyticity are not concerned with maintaining
the view that apparently analytic sentences are guaranteed
to be true then it is hard to see what motivates the philo-
sophical interest in analyticity.

Historically speaking, there is little doubt that the
friends of analyticity have considered apparently analytic
sentences such as 'Everything is self identical' and sen-
tences of the form 'p or not p', for example, to be guaranteed
to be true. This is reflected in the fact that such sentences
were said to be necessarily true and couldn't possibly fail
to be true. The claim that such sentences couldn't possibly
fail to be true was considered to be justified by the claim
that these sentences were analytic. But, as we saw in an
earlier section, the claim that a sentence is analytic is
not a good justification for the claim that it is true.
Moreover, the friends of analyticity will have to deal not
only with this negative criticism but also with the positive
criticism, contained in the story of the strange particle
theory, which claims that it is possible for an apparently

analytic sentence to be false. They would also have to deal

92

with the possibility that our present language might turn
out to be inadequate (in the sense described earlier) to

future develOpments in science.
Section 6: Analyticity by Fiat and Truth by Convention

The purpose of this section is to see to what extent
the notion of analyticity by fiat and truth by convention
can be used to defend the notion of analyticity against the
objections already mentioned in this work.

Suppose we introduce 'E' as an abbreviation for 'the
smallest prime number between 12 and 100'. We can view this
as introducing, by fiat, a relation of synonymy between these
two expressions. And since sameness of sense implies same-
ness of reference it follows that these two expressions have
the same reference, and thus that the sentence formed by
placing '=' between these two expressions is true. Its
truth is guaranteed by the fact that the synonymy was created
by fiat, along with the truth of the principle that sameness
of sense implies sameness of reference. Alternatively we
could guarantee the truth of the identity directly in the
same manner that the truth of meaning postulates are guaran-
teed in an artificial language, i.e. by just declaring by
fiat that it shall be true. This would be a case of what
Quine calls "legislative postulation". The former move
would be a case of what he calls "legislative definition",
to be distinguished from what he calls "discursive definition",

the latter notion being a case of a statement of definitional

93

form which "sets forth a pre-existing relation of inter-

changeability or coextensiveness between notations in

already familiar usage".18

In "Two Dogmas" Quine had seemed to endorse the notion
of synonymy when it is created by fiat in the case of
explicit conventional notational abbreviation.19 In his
later work, however, he set forth a view in the context of
which his earlier endorsement is seen to give little aid to
the friends of analyticity. Quine's later views are best
set forth in his own words.

Definition, in a properly narrow sense of the word,
is convention in a properly narrow sense of the
word. But the phrase 'true by definition' must be
taken cautiously; in its strongest usage it refers
to a transcription, by the definition, of a truth
of elementary logic. Whether such a sentence is
true by convention depends on whether the logical
truths themselves be reckoned as true by convention.
Even an outright equation or biconditional connecting
the definiens and the definiendum is a definitional
transcription of a prior logical truth of the form
'x=x' or 'p :5 p'.

Definition commonly so called is not thus narrowly
conceived, and must for present purposes be divided,
as postulation was divided, into legislative and
discursive. Legislative definition introduces a
notation hitherto unused, or used only at variance
with the practice proposed, or used also at
variance, so that a convention is wanted to settle
the ambiguity. Discursive definition, on the other
hand, sets forth a pre-existing relation of inter-
changeability or coextensiveness between notations
in already familiar usage. A frequent purpose of
this activity is to show how some chosen part of
language can be made to serve the purposes of a
wider part. Another frequent purpose is language
instruction.

It is only legislative definition, and not discursive
definition nor discursive postulation, that makes a
conventional contribution to the truth of sentences.
Legislative postulation, finally, affords truth by
convention unalloyed.

94

Increasingly the word 'definition' connotes the
formulas of definition which appear in connection
with formal systems, signalled by some extrasys-
tematic sign such as '= '. Such definitions are
best looked upon as correlating two systems, two
notations, one of which is prized for its economical
lexicon and the other for its brevity or familiarity
of expression. Definitions so used can be either
legislative or discursive in their inception. But
this distinction is in practice left unindicated,
and wisely; for it is a distinction only between
particular acts of definition, and not germane to
the definition as an enduring channel of inter-
translation.

The distinction between the legislative and
discursive refers thus to the act, and not to its
enduring consequence, in the case of postulation

as in the case of definition. This is because we
are taking the notion of truth by convention fairly
literally and simple-mindedly, for lack of an
intelligible alternative. So conceived, conven-
tionality is a passing trait, significant at the
moving front of science but useless in classifying
sentences behind the lines. It is a trait of events
and not of sentences. ‘

Might we not still project a derivative trait upon
the sentences themselves, thus speaking of a sentence
as forever true by convention if its first adoption

as true was a convention? No; this, if done
seriously, involves us in the most unrewarding
historical conjecture. Legislative postulation
contributes truths which become integral to the

corpus of truths; the artifiicality of their origin
does not linger as a localized quality, but suffuses
the corpus. If a subsequent expositor singles out
those once legislatively postulated truths again as
postulates, that signifies nothing; he is engaged
only in discursive postulation. He could as well
choose his postulates from elsewhere in the corpus, 20
and will if he thinks this serves his expository ends.

With respect to our example about 'E', some of these
views of Quine can be illustrated as follows. After
introducing 'E' as an abbreviation it will start appearing
in other statements which we believe to be true. Its
abbreviatory nature and the artificiality of its origin will

be forgotten. Later on we may find it useful, in our

95

attempts to explain the world, to retain some of the
statements in which 'E' occurs but to reject, as false,

the sentence 'E= the smallest prime number between 12 and
100'. Here we have an expression of the Quineian view even
if an expression is initially introduced as an abbreviation
it will become subject to Quine's "law" that an expression
does not have meaning in isolation from its fellows.

The distinction between the discursive and the
legislative is a good one. But the substantive views of
Quine, as expressed in this quotation constitute merely a
begging of the question of the desirability of making a
sharp distinction between the analytic and the synthetic.
These views gain their plausability because they seem to
be accurate descriptions of people's actual linguistic
behavior. The friends of analyticity are just as willing
as anyone else to admit that as a matter of fact scientists,
as well as others, have not, in the past made a sharp dis-
tinction between discursive and legislative definition, and
that even if they were to institute a truth via legislative
definition this fact would be lost sight of in the subse-
quent deve10pment of their science. But to say, as Quine
does, that it is wise to continue this practice is just to
make the question begging claim that it is unwise to make a
sharp distinction between the analytic and the synthetic
and also unwise to adhere to it once it has been made in a

particular c ase .

96

I am concerned, at the moment, only to show that in the

quoted passage there are no new arguments against the desire-

 

ability of making, and adhering to, a sharp analytic synthe-
tic distinction. Quine's comments add nothing to the argument,
which we gave in Section 2 against the utility of making a
distinction between the analytic and synthetic. However,

his comments, viewed as descriptive statements about people's
actual practice, can help us in the main task of this section;
i.e. to see whether the notion of analyticity by fiat can

help defend the notion of analyticity against the objections
brought up earlier in this work.

Consider first the problem of whether there are any
analytically true sentences. Let us suppose, at least for
the sake of argument, that if at one time an expression such
as a predicate 'F' has been introduced as an abbreviation for
another, say 'G', and declared to be synonymous with it,
then the universal closure of the biconditional of these two
predicates is analytically true by fiat and guaranteed to be
true. But even within these assumptions the application of
this line of reasoning to the question of, say, whether the
sentence 'All bachelors are unmarried adult males' is analy-
tic would involve us in the unrewarding historical conjecture
that at some time in the past someone explicitly declared
'bachelor' to be synonymous with 'unmarried adult male'.

Consider next the problem about justification discussed
in Section 2 of Chapter 1. Here it is pretty clear that the

notion of analyticity by fiat will be of no help, since any

97

historical conjecture to the effect that some apparently
analytic sentence 8 was at one time declared to be true by
fiat will not be more certain than the claim that S is true.
And thus the historical claim cannot be a good justification
for either the claim that is is true or for the claim that

S is analytically true.

Consider next the question, discussed in Section 3 of
Chapter 1, of whether the notion of analyticity has any
explanatory power. One might attempt to explain our belief
in the truth of a sentence in terms of the notion of analy-
ticity by fiat by constructing an explanation whose explanans
contains, as the only lawlike premise, the statement that if
a predicate F has at one time been declared to be synonymous
with another predicate G then provided that the meanings of
F and G do not change, we believe the universal closure of
the biconditional connecting F and G. But a ”no synthetic
law" objection can be made against this explanation as well
since the only lawlike statement in this explanans is analytic.

Thus, it seems that the notion of analyticity by fiat
will not help in answering the objections brought up in this
work to the notion of analyticity as it applies to the sentences
of our present language.

A weakness in the notion of analyticity or truth by fiat
shows up clearly even in the case of artificial languages.
The inability of fiat to institute or guarantee truth is
clearly seen in a case where we have a set of sentences whose

non—logical terms have not yet been interpreted and each

98

sentence of which is declared true by fiat. In such a case
there is, because of Church's theorem, no guarantee that we
won't be able to later on derive a contradiction from this
set of sentences. And if we were to derive such a contra-
diction then not all of the sentences in the set would be
true, in spite of the fact that all of them were originally
declared true by fiat. Thus, truth by fiat doesn't guarantee

truth.
Section 7: The Verifiability Theory of Synonymy

The purpose of this section is to examine a definition
of sentential synonymy put forward by Grice and Strawson in
response to one of Quine's criticisms, and to examine Quine's
response to this definition.

In "Two Dogmas" Quine had said that ”If the verification
theory (of meaning) can be accepted as an adequate account
of statement synonymy, the notion of analyticity is saved
after all."21 Now there is more than one way of defining the
notion of analyticity via the verification theory of meaning.
Whichever way we take however, it must be kept in mind, in
order to avoid needless confusion, that definitions of analy-
ticity must always be considered as containing a relativization
to time and speaker since a sentence can be analytic at one
time but not at another, and analytic for one person but not
for another. Sometimes this relativization is implicitly
taken care of when analyticity is defined with reference to

an ideal language, the implicit assumption being that such

99

a language does not change over time in any respect since if
it did it would not be.the same language after the change as
before. But when we are not defining analyticity with respect
to an ideal language the relativization to time and speaker
must be more firmly kept in mind. However, in the defini-
tions to be considered I will, for stylistic reasons, usually
not explicitly include such relativization, although the
reader should keep them in mind.

One way to define the notion of analyticity is to say
that a sentence is analytically true just in case any experience
would confirm it to some degree or other and that no experience
would disconfirm it to any degree. In order to make clear
the intent of this definition it should be pointed out that
from it we cannot conclude that if at some future time an
experience occured which we then took to be disconfirming
evidence for a sentence we could then conclude that the sen-
tence never was analytic. For the intent of the definition
is that a sentence is analytically true at time t just in case
any experience, if it occured at time t, would confirm at
time t the sentence (and similarly with disconfirmation).

The definition of analyticity is of counterfactual form (in
this it seems no worse than Quine's definition of stimulus
meaning) and it makes no claim to the effect that if a sen-
tence is analytic at one time then it must be analytic at

other times.

100

Another way to define analyticity via the verification
theory of meaning is to say, first of all, that two sentences
are synonymous just in case any experience which would confirm
or disconfirm one of them would confirm or disconfirm the
other to the same extent. We can then define an analytically
ture sentence as one which is synonymous with some logical
truth.

These two definitions do not have the same virtues and
defects. The second definition, but not the first, has the
defect of not defining analyticity as it applies to purely
logical truths and of presupposing the notion of logical
truth. However, it does have a virtue which the first defini-
tion doesn't. The first definition implicitly utilizes the
notion of analyticity. Consider the first definition as it
applies to, for example, the sentence 'Nothing is both round
and square'. Certainly if we were to experience something
which is both round and square this experience would discon-
firm the sentence in question. You can, if you want to, say
that it is logically impossible to have such an experience
(i.e. that the sentence describing this experience is analy-
tically false). But then this shows that you are implicitly
utilizing the notion which you are supposed to be defining
and that the first definition actually needs to be amended
to say that a sentence is analytically true just in case every
logically possible experience would confirm it and no logically
possible experience would disconfirm it.

Now the second definition doesn't suffer from this

defect. We can just say 'any experience' simpliciter, without

 

101

restricting the experiences referred to in the second
definition to those that are "logcically possible”. If any
given experience, including even "logically impossible ones"
confirms or disconfirms 81 to a certain extent, and also
confirms or disconfirms S2 to the same extent, then S1 and
S2 are synonymous.

Both of these definitions however, are claimed by
Quine to suffer from the fact that they presuppose the claim
that a sentence, taken in isolation from its fellows, can
admit of confirmation or disconfirmation at all. He then
claims that this assumption is false. His countersuggestion
is that "our statements about the external world face the
tribunal of experience not individually but only as a cor-

"22 and that it is an illusion to suppose that

porate body

"to each statement, or each synthetic statement, there is

associated a unique range of possible sensory events such

that the occurence of any of them would add to the likelihood

of truth of the statement, and that there is associated also

another unique range of possible sensory events whose occurence

would detract from that likelihood”.23
Quine's countersuggestion seems quite plausible, eSpecially

with respect to the highly theoretical sentences of science.

For example, an experience describable as being evidence for

the truth of Newtonian mechanics doesn't seem to confirm

Newton's second law any more than, say, Newtpn's first law.

Moreover, Quine's view can even be made to look plausible

with regard to relatively non-theoretical sentences. This

102

can be seen in terms of an example. Consider the claim
that there is a pencil on my desk at time t. Suppose that,
at time t, I look carefully on my desk and see no pencil.
Quine's view is not merely that I can continue to accept the
claim that there is a pencil on my desk by making changes
elsewhere in my system of beliefs (e.g. by pleading hallu—
cination, or by claiming that an invisble Martian put an
invisible anti-pencil filter in front of me at time t, or...)
but that we need not consider this recalcitrant experience as
evidence against 221K the claim that there is a pencil on
my desk. We can also consider it as evidence equally against
the claim that I am not having a hallucination, and that
there is no invisible Martian..., and .... We can "smear
out" and distribute this disconfirming evidence against
these sentences in any way we please, subject only to pragmatic
considerations of simplicity, conservatism, etc. The same
point is perhaps more readily seen not in terms of a recal-
citrant experience but in terms of a confirming one. If I
do see a pencil on my desk at time t, this need not be
considered as evidence in favor of only the claim that there
is a pencil on my desk but also equally in favor of the claim
that my senses were in working order, and that no invisible
Martian put an anti-pencil filter in front of me, and....

Now both of the definitions we haVe mentioned seem to
presuppose that a sentence, taken in isolation from its
fellows can admit of confirmation and disconfirmation. Grice

and Strawson, in an attempt to save the verification theory

103

of sentence synonymy, respond to this view of Quine's in the
following way. (What they refer to as 'the second doctrine'
is the view of Quine's just described, i.e. the doctrine that
it is an illusion to suppose that an individual statement,
taken in isolation from its fellows, can admit of confirmation
or disconfirmation at all.)

It is easy to see that acceptance of the second
doctrine would not compel one to abandon, but only

to revise, the suggested explanation of synonymy.
Quine does not deny that individual statements are
regarded as confirmed or disconfirmed, are in fact
rejected or accepted, in the light of experience.

He denies only that these relations between single
statements and experience hold independently of our
attitudes to other statements. He means that experi-
ence can confirm or disconfirm an individual state-
ment, only given certain assumptions about the truth
or falsity of other statements. When we are faced
with a "recalcitrant experience", he says, we always
have a choice of what statements to amend. What we
have to renounce is determined by what we are anxious
to keep. This View, however, requires only a slight
modification of the definition of statement synonymy
in terms of confirmation and disconfirmation. All

we have to say now is that two statements are synon-
ymous if and only if any experiences which, 93 certain
assumptions about the truth-values of other statements,
confirm or disconfirm one of the paif, also, on the
same assumptions, confirm or disconfirm the ofhef—fo
the same degree. More generally, Quine wishes to
substitute for what he conceives to be an oversimple
picture of the confirmation-relations between parti-
cular statements and particular experiences, the idea
of a looser relation which he calls "germaneness"

(p. 43). But however loosely "germaneness" is to be
understood, it would apparently continue to make

sense to speak of two statements as standing in the
same germaneness-relation to the same particular
experiences. So Quine's views are not only consistent
with, but even suggest, an amended account of statement
synonymy along these lines.

 

 

 

 

Since "Two Dogmas" Quine's Word and Object was published, a

 

book in which the rough notion of two statements standing

in the same germaneness relation to each of all possible

104

experiences is explained as being, on an optimum modulus,
the relation of sentential stimulus synonymy. In the same
book he also responds to the above attempt to save the veri-
fication theory of synonymy. As we shall see, Quine mis-
interprets Grice and Strawson's proposal. But first let us
look at Quine's response and at some preliminary problems
with it.

Grice and Strawson...(define) 51 and 52 as synonymous
when, for every assumption as to the truth values of
other sentences, the same experiences confirm (and
disconfirm) S1 on that assumption as confirm (and dis-
confirm) S2 on that assumption. Now instead of

'every assumption as to the truth values of other
sentences' we can as well say simply 'every sentence
8': for S can be the logical conjunction of those
"other sentences" in question or their negations.

So 8 and S are defined to be synonymous when, for
every S, the same experiences confirm (and disconfirm)
S1 on the hypothesis S as confirm (and disconfirm)

82 on S. The notion of confirmatory and disconfirma-
tory experiences had a behavioral approximation in our
notion of stimulus meaning; but can we relativize it
thus to a hypothesis 8? I think we can; for confir-
mation or disconfirmation of S on S is presumably
confirmation or disconfirmation of the conditional
sentence consisting of S as antecedent and S as
consequent. Then the proposed definition of synonymy
becomes: S and 52 are synonymous if for every S the
conditional compound of S and S and that of S and 82
are stimulus-synonymous. But now it is apparent that
the definition fails to provide a tighter relation
between S and S than stimulus synonymy. For, if

S and S are stimulus-synonymous then s fortiori

t e conditionals are too. ‘

Now Quine thinks that stimulus synonymy is not (even
if we limit our considerations to intrasubjective synonymy)
an adequate reconstruction of synonymy in the traditional
sense. (In the next section I will try to show that, in
the intrasubjective case, it is more adequate than Quine

claims. But for now we have other things to consider.)

.I’\|Illlh

105

Quine seems to be presupposing something like the false
principle that the probability of q given p is the same as
the probability of 'if p then q'. He seems to make this
mistake when he says ”for confirmation or disconfirmation of
S1 on S is presumably confirmation or disconfirmation of the
conditional sentence consisting of S as antecedent and S1
as consequent". However, this mistake of conflating condi-
tional probability with the probability of the conditional
is not part of Quine's misinterpretation of Grice and Strawson.
The reason it is not is that it "cancels out”, so to speak,
in Quine's calculations. To illustrate what I mean by Quine's

mistake cancelling out consider the following two principles

(the second of which is false).

a. If S1 and S2 have the same probability then so do
'if S then 81' and 'if S then 82'.
b. If 'if S then 81' and 'if S then 82' have the same

probability then so do S1 and S2.

Now when Quine finally gets down to his version of Grice and
Strawson's proposal he doesn't seem to depend on the mistaken
conflation of conditional probability with probability of the
conditional, although he does seem to depend on something
like a stimulus synonymy "analogue" of the two principles (a)
and (b). So this particular mistake of Quine's does not, by
itself, make his reply to Grice and Strawson inappropriate.

Let us now consider the way in which Quine's reply mis-
interprets or is inappropriate to Grice and Strawson's
prOposal. First of all, Quine pretty clearly seems to mis-

interpret Grice and Strawson (hereafter G and S) when they

106

say "on certain assumptions about the truth values of other
statements", and he says "on EXEEX assumption about the
truth values of other statements". However, in spite of
this misinterpretation, and in spite of the fact that G and
S don't amplify what they had in mind when they used the
expressions 'certain assumptions' and 'other statements'
Quine's reply is inapprOpriate for an even more fundamental
reason. In order to see what this more fundamental reason
is, let me try to get Quine and G and S as close together as
possible, so to speak.

G and S:

S1 means the same as S2 for person P at time t just in case
if certain other sentences (whose conjunction we will call S)
are true, then S1 and S2 are stimulus synonymous for person
P at time t.

Quine:

S means the same as 82 for person P at time t if and only

1

if the conditional compound of S and S and that of S and 82

l
are stimulus synonymous for person P at time t.
(Here '8' refers to the same conjunction of sentences that
it did in the G and S definition.)

Now Quine claims (correctly, as far as I can see) that
if 81 and S are stimulus synonymous then so are 'if S then

2

81' and 'if S then 82'. I take it that Quine also holds to

the converse of this (see bottom two lines on p. 64 of Word

and Object, where his 'if' is apparently meant as 'if and

 

only if'). So Quine's definition reduces to the claim that

two sentences are intrasubjectively synonymous just in case

107

they are intrasubjectively stimulus synonymous.

But now it is clear that Quine's definition is not
equivalent to G and 8's definition, since the respective
definientia are not equivalent. That is, the claim that S1
is stimulus synonymous with S2 is not equivalent to the claim
that if S is true then S1 and S2 are stimulus synonymous.

Grice and Strawson's definition of sentential synonymy
has the virtue of not implicitly presupposing the very
notion which is being defined. Perhaps this is why Quine
claims that if the verification theory of meaning is
adequate as an account of sentential synonymy then the notion
of analyticity is saved after all. Grice and Strawson's
definition certainly seems worthy of development and filling
in of details. Nevertheless, at present I cannot see how
such a further development along these lines would help in
answering the objections to the notion of analyticity mentioned
in earlier sections of this work, especially the objections
that the notion of analyticity has neither explanatory nor

justificatory power.
Section 8: Stimulus Synonymy as an Explication of Synonymy

The purpose of this section is to see how far the notion
of stimulus synonymy, as this notion is developed by Quine

in Word and Object, can be used to reconstruct the notion

 

of synonymy in the traditional sense. I will argue that
even though it won't do the job completely it goes farther

than Quine claims.

108

For our purposes we need to make distinctions among
various subspecies of synonymy. In this section I will
consider only intrasubjective synonymy of expressions.
Intersubjective intralinguistic and intersubjective inter-
linguistic synonymy will not be discussed. What we are
concerned with in thissection then is the notion of two
expressions being synonymous for a person P at time t.
Moreover, the expressions whose synonymy we will first
consider are sentences.

When the sentences in question are occasion sentences
Quine admits that stimulus synonymy, especially as socialized,
pretty well reconstructs traditional synonymy.26 It is even
adequate enough "when the sentences are standing sentences
which, like 'The Times has come', closely resemble occasion
sentences in the variability of assent and dissent".27
The trouble comes, he claims, when we consider standing sen-
tences with sparse stimulus meanings. I will argue that
stimulus synonymy adequately reconstructs traditional synonymy
even in the case of standing sentences with sparse stimulus
meanings. But before I do I think it worthwhile to point
out a possible misunderstanding of Quine which might arise.

It might appear to some that Quine is wrong when he
says that for intrasubjective synonymy of occasion sentences,
their intrasubjective stimulus synonymy, especially as
socialized, is adequate as an account. It might appear
that Quine's own examples 'It is a rabbit' and 'It is a

rabbit part' are counterexamples. Here we have two occasion

109

sentences which are stimulus synonymous but apparently

not synonymous. But Quine's rabbit-part example is meant

to show that the isims 'rabbit' and 'rabbit part' (and also,
of course, 'rabbit stage' etc.) are not synonymous. This

is pointed out by Quine on the bottom half of page 51 and

page 52 of Word and Object. It is also pointed out by him

 

in his article ”On the Reasons for Indeterminacy of Transla-

28 In this article he says, among other things

tion".
relevant to this point, that "The gavagai example was at
best an example only of the inscrutability of isims, not of
the indeterminacy of translation of sentences."

To make his thesis of the indeterminacy of meaning
easier to see Quine uses the example of a radically foreign
language and expression (such as 'Gavagai'). But if the
thesis of the indeterminacy of meaning is to be taken seriously
then it seems to me that its lessons apply equally well at
home, i.e. to a speaker's own language. So let me try to
apply these lessons to a speaker's own language. I have to
do this because I am concerned only with the case of intra-
subjective intralinguistic sentence synonymy when I claim
that this kind of synonymy is more adequately captured
than Quine thinks by the notion of intrasubjective intra-
linguistic stimulus synonymy of sentences.

In Word and Object Quine says this about the case of

 

intrasubjective intralinguistic synonymy, both of terms and

of sentences.

110

Yet surely the main interest of the synonymy of

'Bachelor' and 'Unmarried man' as occasion sen-

tences was the line it seemed to give on the

synonymy of 'bachelor' and 'unmarried man' as

terms. Now within English the situation is not

beyond saving. To get synonymy of terms from

synonymy of the corresponding occasion sentences

we need only add a condition that will screen out

such pairs as 'bachelor' and 'part of a bachelor';

and this we can do by requiring that the subject

be prepared to assent to the standing sentence

'All F5 are Gs and vice versa', thinking of 'F'

and 'G' as the terms in question. The definition

becomes this: 'F' and 'G' arestimulus synonymous

as terms for a speaker at t if and only if as

occasion sentences they have the same stimulus

meaning for him at t and he would assent to 'All

F5 are Gs and vice versa' if asked at t.2

Now if we apply this definition we see, for example,
that 'rabbit' and 'rabbit part' are not stimulus synonymous
terms for me, for example, because when asked, I would not
assent to 'All rabbits are rabbit parts, and vice versa'.
They are also not synonymous as terms for me. But these
expressions are stimulus synonymous for me as sentences. More
precisely, the sentences 'This is a rabbit' and 'This is a
rabbit part' are stimulus synonymous for me. But now we come
to the crucial question, which is this: Are these sentences
synonymous for me? If they aren't then indeed Quine is wrong
in his claim that sentential synonymy of occasion sentences
(for the intrasubjective intralinguistic case) is adequately
reconstructed by sentential stimulus synonymy.

Now this crucial question is difficult to answer
because we aresnaat home with the syntax of our own language.
Moreover, we automatically assume that the 'This' which occurs
in the first of these two sentences is coextensive with the

'This' which occurs in the second of these two sentences.

111

(Notice that if the 'This' which occurs in 'This is a
rabbit' refers to the rabbit in front of me, and the 'This'
which occurs in 'This is a rabbit part' refers to the rabbit

part in front of me, then the two sentences (or perhaps,

 

more appropriately, sentence tokens) might be more easily
said to be synonymous.)
Now my strategy is to answer the crucial question of

whether these two sentences are synonymous for me in the

 

affirmative. I do this because my intuitions about whether
these two sentences are synonymous fail me here. That is,

my intuitions don't clearly give a negative answer to this
crucial question. This move is analogous to the move where
we let an explication decide a question for us in those cases
where our intuitions have no answer one way or the other.
Thus we can conclude that the rabbit-part example is not a
counterexample to Quine's claim that intrasubjective synonymy
of occasion sentences is pretty adequately reconstructed by
the notion of their stimulus synonymy.

Let me now return to my main concern, which is to argue
that Quine is wrong when he says that stimulus synonymy falls
short when it comes to those standing sentences with sparse
stimulus meanings. Of these kinds of sentences he says:

"But the less variable the standing sentences are in
point of assent and dissent, the sparser their stimulus
meanings will be and hence the more poorly stimulus synonymy

will approximate to synonymy of the envisaged sort."30

112

The situation is best illustrated in terms of a pair
of highly theoretical sentences of some science. To say
that the stimulus meaning of a sentence is sparse means
that most of the stimulations we might come across are
causally irrelevant to our assent to or dissent from the
sentence. If, for example, upon being queried I assent to
or dissent from Newton's second law while looking at my
desk, the sight of my desk is causally irrelevant to my
assent or dissent. The sight of my desk does not prompt my
assent or dissent. Thus the sight of my desk belongs to
neither the affirmative nor the negative stimulus meaning of
Newton's second law for me at that time.

However, the sparseness of stimulus meanings for highly
theoretical sentences is not a sufficient ground for the
claim that stimulus synonymy of such sentences is a poor
reconstruction of their synonymy in the traditional sense.
For even if 51 and 82 were to each have, only one member in
each of their affiramtive and negative stimulus meanings,

S1 would not be stimulus synonymous with S2 if their
respective stimulus meanings were not identical. A "near
miss" is as good as a mile. Stimulus synonymy may come in
degrees, but synonymy in the traditional sense does not.

Thus, as long as there is at least one possible experi-
ence which would, at time t, confirm or disconfirm 81’ but
not confirm or disconfirm 82, then 81 and S2 are not completely
stimulus synonymous.

113

Thus Quine's talk of sparseness of stimulus meanings,
and his talk about ways of "tightening" the relation of
stimulus synonymy, is irrelevant to the question of whether
stimulus synonymy is an adequate reconstruction of traditional
synonymy.31 All that this talk shows is that for sentences
with sparse stimulus meanings it is much more difficult to
gain evidence for the claim that two of them are not stimulus
synonymous than it would be if they were occasion sentences.

What E2219 be relevant to the question of whether stimulus
synonymy adequately reconstructs traditional synonymy for
sentences of this kind is an example of two sentences which
are completely stimulus synonymous but not synonymous in
the traditional sense. But since Quine's definition of
stimulus synonymy requires us to consider all possible
stimulations it seems very unlikely that we would be able to
find such a pair of sentences. Indeed, what we would be
more likely to find, in such a search, is a pair, or more,
of sentences which are "pretty nearly" stimulus synonymous,
and about whose synonymy in the traditional sense we are
puzzled.

This last kind of situation is well illustrated in
cases where mounting recalcitrant experiences result in the
simultaneous overthrow of a whole set of theoretical sentences.
For example, mounting recalcitrant experiences resulted in
the overthrow of a whole batch of sentences of Newtonian
physics which were strongly connected to each other. And
it is just these kinds of sentences which, to a thoughtful

observer, always seemed suspiciously related to each other

114

by devious ties of synonymy. Think, for example, of the

set consisting of Newton's three laws of motion. It is

just such kinds of sentences which tend to support Quine's
contention that sentences are not confirmed or disconfirmed
in isolation from their fellows. But it is also just such
cases in which we are inclined to suspect that the relations
among overthrown sentences are relations of near synonymy

in the traditional sense, in spite of superficial appearances
to the contrary. And this fact supports the claim that
stimulus synonymy is an adequate reconstruction of traditional
synonymy even for this kind of sentence.

There is a further difficulty involved in trying to get
stimulus synonymy to reconstruct the traditional notion of
synonymy, a difficulty which Quine attempts to solve with
his notion of socialized stimulus synonymy. Socializing
intrasubjective stimulus synonymy is supposed to make
stimulus synonymy a more adequate reconstruction of intra-
subjective synonymy by cutting out the effects of information
which is idiosyncratic to the individual. Thus, the sentences
'Indian nickel' and 'Buffalo nickel', while stimulus synon-
ymous for the numismatic expert, are, Quine assumes, not
synonymous for the expert. But then we repair the situation
by pointing out that these two sentences are not socially
stimulus synonymous since they are not stimulus synonymous
for the novice. So then we say that two sentences mean the
same for person P just in case they are stimulus synonymous

for P and also stimulus synonymous for every member of P's

115

"community". The desired result is that the two sentences

in question are not synonymous for the expert because while
they are stimulus synonymous for him they are not socially

stimulus synonymous.

I have always been unimpressed by this move of Quine's
because it seems to me that his implicit and indirect use of
the traditional notion of meaning is hidden in his use of
the notion of "community". I take it that by 'community'
he means 'linguistic community'. To stick to the example,
Quine seems to presuppose that 'Indian nickel' means the
same (in the traditional sense of 'means the same') for one
member of the community as it does for any other member of
the community. Without this presupposition the socialization
of stimulus synonymy loses its point. On the other hand I
don't see how Quine can justify this uncritical reliance on
the supposedly disreputable notion of inisisubjective synonymy.
(Incidentally, it might be said that if 'Indian nickel'
doesn't mean the same for me as it does for you then we are
not speaking the same language. But see Quine's own misgivings

about the notion of 'a language' on page 214 of Word and Object.)

 

Thus, if we are to avoid implicitly utilizing in our definitions

the notions against which Quine has so often inveighed then

we must do without the notion of socializing stimulus synonymy.
Now since I want to claim that intrasubjective stimulus

synonymy is, all by itself, (i.e. without socializing it) an

adequate reconstruction of intrasubjective synonymy of

occasion sentences how do I handle the case of the numismatic

116

expert? I handle it by not agreeing with Quine's implied
claim that 'Indian nickel' and 'Buffalo nickel' are not,
for the expert, synonymous in the traditional sense. I
don't claim that they are synonymous for him either. I
agree that for the non-expert they are not synonymous. But
it is not clear whether or not they are synonymous for the
expert. It doesn't seem to me that for the expert 'Here

is a coin with an Indian head on one side' is a better analysis
of 'Indian nickel' than is 'Here is a coin with an Indian
head on one side and a buffalo on the other'. Furthermore,
we cannot settle this question by appealing to the numis-
matic experts because they, not being philosophers, are
like dictionaries in not making fine enough distinctions
between coextensiveness and synonymy, or between synthetic
truths and analytic ones.

The upshot is that since our intuitions don't come down
clearly on the side of the claim that 'Indian nickel' and
'Buffalo nickel' are not synonymous for the expert we can let
the definition decide the case for us. And the definition
says that, as sentences, they are synonymous because they
are stimulus synonymous.

It might be worthwhile to mention, in passing, that
from this conclusion it doesn't follow that stimulus meaning
is, even for the sentences so far considered, an adequate
reconstruction of meaning in the traditional sense. To
think that it is adequate is to embrace a dogma of reduc—
tionism, i.e. the dogma that the meaning of a sentence is

nothing but the set of all those experiences which would

117

confirm it (or perhaps the ordered set, the first member of
which is the set of all those experiences which would confirm
it and the second member of which is the set of all those
experiences which would disconfirm it). Two sentences of

the sort so far considered have the same meaning if and only
if they have the same stimulus meaning; but it doesn't follow
from this that the meaning of a sentence is nothing but its
stimulus meaning. Berkeley made the mistake of embracing

the above mentioned dogma by claiming, for example, that the
meaning of the sentence 'This table exists' is, in effect,
nothing but the set of sentences describing experiences

which would verify it (sentences such as 'I see the table',
'I feel the table', etc.). His mistake seemed plausible
because we seem hard pressed to come up with an answer to

his implicit question 'If 'exists', as applied to physical
objects, doesn't mean to be perceived or to be perceivable
then what else does it mean?'. (My answer is that 'exists'
is cognitively primitive in the sense that if you really
don't know what 'exists' means then I can't tell you.)

There remains a further problem for me if I want to
claim that intrasubjective stimulus synonymy is an adequate
reconstruction of intrasubjective synonymy. The problem is
illustrated by the example of the sentences 'There are black
dogs' and 'There are black cats'. The affiramtive stimulus
meanings of these sentences are presumably empty. And their
negative stimulus meanings, if any, are not capturable in

stimulations occurring within an optimum modulus. Thus these

118

two sentences are presumably stimulus synonymous, yet seem

to be non-synonymous. It is clear that the terms 'black dog'

 

and 'black cat' are not synonymous. The above sentences can
be said to be non-synonymous only if we assume that the two
occurrences of 'There are' are synonymous (for if they
weren't then the two sentences might be synonymous even though
the terms 'black dog' and 'black cat' are not). Since it is
also clear that these two occurrences are synonymous we have
to conclude that the two sentences are non-synonymous.
(Note, incidentally, a difference between this example and
the earlier one about rabbits and rabbit parts. No one would
want to claim that words like 'it' and 'this thing' refer to
the same thing in all their occurrences (and thus, since
sameness of sense implies sameness of references, they don't
have the same sense, if they have a sense at all, in all their
occurrences). But even those who, like Ryle, would want to
claim that 'there are' (or 'exist') is ambiguous surely would
agree that at least 'there are' means the same as applied to
cats as it does when applied to dogs.)

Thus we have to conclude that stimulus synonymy doesn't
adequately reconstruct traditional synonymy for sentences of

the above kind. I simply see no way out of this difficulty.
Section 9: Meaning, Significance, and Brain States

The purpose of this section is to examine some ways in
which it might be thought that the notion of meaning might,

in the future, be made more "legitimate" by being linked up

119

with some internal mechanisms supposedly responsible for
the process which‘we call understanding an expression.

In its present state the notion of meaning might be
compared with the notion of a dormitive virtue, as used by
Moliere's physician. The claim that opium puts people to
sleep because it has a dormitive virtue was non-explanatory
because the only criterion for whether or not something had
a dormitive virtue was that it put people to sleep. The
same kind of defect is involved if we say that a person
understands an expression, or knows how to use it (in the
sense of applying it in a non-arbitrary way), because there
is a meaning which the expression has and which the person
"grasps". For the only criterion for whether or not a person
grasps the meaning of an expression is that he understands
it, or knows how to use it.

Now the notion of a dormitive virtue can be seen as a
promissory note which was eventually cashed in. To say that
opium put people to sleep because it had a dormitive virtue
can be viewed as a statement to the effect that there is an
unknown something in opium which is responsible for the fact
that opium puts people to sleep. This unknown was eventually
discovered by chemistry, and thus the promissory note was
cashed in. Similarly it might be claimed that someday the
notion of meaning will become "legitimate" by being cashed

in in terms of neurophysiological states.

120

Let me paint a picture, so to speak, suggested by this
kind of move, and then see what scientific sense might be
made of it.

According to this picture to each expression which we
understand there corresponds an as yet unknown brain state.
In order to screen out irrelevant issues let us also assume,
for the sake of argument, that the notion of synonymy is clear
at each pair of expressions which we understand; i.e. that
for each such pair we either know that they are synonymous
(as with 'bachelor' and 'unmarried man'), or know that they
are not synonymous, or know what to do to gain evidence for
the claim that they are synonymous. According to the picture,
the claim that two expressions are synonymous is tantamount
to the claim that when the brain state correlates of expres-
sions are eventually found it will be seen that for any
given person P the brain states correlated with two expressions
which are, for P, intuitively intrasubjectively synonymous
(such as 'bachelor' and 'unmarried man') are identical.
Moreover, on this view, the analogue of the fact that opium
puts people to sleep is the fact that there are expressions
which we understand, in the sense that we apply them in a
non-arbitrary way to new objects which we have not been told
to apply the expression in question. Thus it is the fact
of what might be called our linguistic competence with
respect to these expressions which is the fact to be eventually
explained by appeal to the brain states correlated with these

expressions. When these brain states are discovered, so the

121

story goes, agenuine explanation of our linguistic compe-
tence will be available since the explanation will have the
form 'A person understands expression E if and only if he
is in brain state B'. The explanation will be genuine
because the criterion for whether a person is in brain state
B will be independent of the criterion for whether he
understands expression E. Another way to put this is to
say that the connection between the fact that a person is in
brain state B and the fact that he understands E is a con-
tingent or factual connection rather than a necessary one.
Things might have been different in the sense that brain
state B might just as well have been correlated not with E
but with some other expression B'. Let us now see how much
scientific sense can be made of the picture I have just
constructed.

For the sake of simplicity let us limit our considerations
to those expressions which are monadic predicates. Since a
brain can simultaneously be in many different states the
question arises of how we are to discover which of these
states is the one correlated with some expression E. One
might try to answer this question by saying that we could,
in principle, use a before-and-after procedure in which we
would take a child learning his native language and compare
his brain states before and after he has learned the use of
B. What is wrong with this method is not merely the fact
that, in practice, there will probably be many "new" brain

states from which to pick. There is the more serious

122

objection that it is highly doubtful whether it is even
possible in principle for someone to learn the use of any
but a few of the expression of his first language in the
stringent one-by-one fashion which the before-and-after
method requires.

We might try another approach to the problem of dis-
covering what the brain state correlate of an expression is.
We might consider the class of all those people who under-
stand some expression E and then look for a brain state B
common to all and only those people. We then might be
tempted to conclude that this brain state B is the one
correlated with E. But this would be a mistake. It would
be a mistake because from the fact that all of these people
understand E, in the sense of applying it in a non-arbitrary
way to new cases, it doesn't follow that they are all using
E in the same sense or with the same "meaning". In order to
be able to conclude that B is "the" brain state contingently
correlated with B we would need a criterion for the inter-
subjective synonymy of E which is independent of the criterion
for whether they are all in the brain state B. But this is
just what we don't have. We may, by noticing the objects
to which these people are willing to apply or withhold E,
gather inductive evidence for the claim that E is, or is not,

intersubjectively coextensive. But what we are looking for

 

is something else; i.e. a way of getting evidence for the

claim that E is intersubjectively synonymous. However, I

 

see no way of getting such evidence which does not arbitrarily

123

and implicitly assume the intersubjective intralinguistic
synonymy of expressions other than E in a manner similar to
the way Carnap assumed the intersubjective interlinguistic
synonymy of various expressions, as mentioned in the cir-
cularity criticism of Carnap which we brought out in
Section 3 of this chapter. Consequently, I feel obliged to
conclude that it is difficult to see how the brain state
picture can be empirically supported in such a way as to
support the view that for each meaningful expression there

is a meaning which it has.32

Chapter 3
Analyticity in Artificial Languages

It is an understatement to say that Carnap considered
artificial languages to be of help in the clarification of
concepts, including the concept of analyticity. Quine, on
the other hand, has not considered the use of an artificial
language to be of any help in the clarification of analyticity.1
The purpose of this chapter is to examine some issues connected
with the attempt to clarify analyticity with the help of
artificial languages, to show in what sense the notion of
analyticity is thus clarified, and to provide conditions
under which a predicate, defined with reference to an
artificial language, can be said to be an analyticity
predicate.

At the outset it should be mentioned that, strictly
speaking, the objects which are obscure or clear, and whose
clarification we seek, are not concepts but linguistic
expressions. (Although I have, in this work up till now,
used the word 'concept' the reader may have noticed that its
use was never essential because it could have been replaced
by 'linguistic expression'.)

The Encyclopedia 3i Philosophy, in conformity with

 

 

philosophical usage, defines a formal language as an
interpreted formal system.2 The ambiguity of 'interpreted
formal system' raises many interesting questions about what

the relationship should be between an interpreted formal system

124

125

and a natural language in order for the formal system to
properly merit the title 'language'. For example, a parti-
cular interpretation can be given in more than one way. It
can be given directly by specifying the extensions of the
various expressions which are being interpreted. Or it can
be given indirectly by specifying the intensions of these
expressions. Suppose that we had an object language predicate
'F'. One way to interpret it would be to lay down the
following rule in the metalanguage: 'F' is coextensive with
'salesmen' (or: 'F' denotes the class of salesmen). Another
way would be to lay down in the metalanguage the following
rule: 'F' is synonymous with (i.e. has the same intension
as) 'salesmen', (or perhaps 'F' designates the property of
being a salesman). There may be some who might want to argue
that only the second way of giving the interpretation would
make the formal system a language because only this way is
the meaning (as opposed to merely the extension) of the
predicate 'F' given.

A question also arises as to whether an interpreted
formal system should contain axioms in order to qualify as
a language.

Still others might want to claim that an interpreted
formal system is not a language if it is not actually used
for the purposes of communication.

I will ignore these and other interesting questions
about the notion of language since I can do so without

detriment to the goals of this chapter.

126

Those who like clarity, exactness, and precision like
formal languages because some questions and answers con-
cerning them are, to some extent, more clear, exact, and
precise than they are for natural languages. E. W. Beth
says, for example, (to use an illustration chosen at random):

It is one of the main advantages of formalised
languages as compared to natural languages that
their syntax and semantics can be constructed as
deductive sciences. Thus numerous questions which
cannot be settled satisfactorily with regard to
natural language can be answered in a completely
rigorous manner for formalised languages. For this
reason it is most regrettable that British analytic
philOSOphy has recently turned away from using and
investigating formalised languages and now develops
what I have described in section VI as a "mystical"
attitude with regard to natural language.

An example of the clarity made possible via formalised
languages is provided by the fact that we can define
'sentence of L' in such a way that there is what may be
called a decision procedure for this predicate; i.e. we
can mechanically determine, in a finite number of steps,
whether or not a given object is a sentence of L. Thus,
the predicate 'sentence of L' is clear in the sense that we
have a search procedure for it (see Section 4 of Chapter 1
for the notion of a search procedure). (If we have a
decision procedure then we automatically have a search
procedure (although the converse doesn't hold) since a
decision procedure is a search procedure which is guaranteed
to be successful.) By contrast, it seems that we have

neither a decision procedure nor a search procedure for the

predicate 'sentence of English'.

127

Not all syntactical or semantical predicates of a
formalised language have decision procedures. For example,
the syntactical predicate 'provable in L' has no decision
procedure for formalised languages rich enough to include
truth functions and quantification. Likewise the semantical
predicate 'logically true' has no decision procedure for
such languages. But since we have search procedures for
these predicates they can be said to be clear, whereas the
corresponding predicates for a natural language are not clear
because they have no search procedures. We have, for example,
a search procedure for the predicate 'logically true in L'
(where L is some appropriate formalised language). One of
these procedures would consist in having a person construct
a truth tree for the negation of some given sentence whose
logical truth we want to test for.

Similarly 'analytic in L' can be made clear for a
formalised language. But the question arises of how this
clarifies the predicate 'analytic in English'. It doesn't
seem to provide any clarification at all. A definition of
'analytic in L', for some formalised L, can be as exact,
precise, and clear as one could want and yet we are still
left with the same English sentences whose analyticity we
are undecided about and about which we don't know what to
do to gain evidence for or against their analyticity as we
had before the definition of 'analytic in L'. The cause of
this situation is quite simple. It consists in the fact

that typically all, or at least most, of the sentences of

128

a formalised language are not identical with any English
sentence. Because of this the extensions of 'analytic in
L' and 'analytic in English' hardly overlap, if at all.
Consequently, if we take, for example, the English sentence
'Everything green is extended' we won't be helped in
deciding whether or not it is analytic by being told that
according to the definition of 'analytic in L' some pips:
sentence (i.e. a sentence of L) is or is not analytic in

L.

The situation is not beyond saving. But in order to
save it we will have to look closely at the question of
what the relations must be between some formalised language
and English in order for a definition of 'analytic in L'
to throw light on the predicate 'analytic in English'.

I think one can easily be misled by concept talk into
thinking that because we may define, in a clear and exact
way, the predicate 'analytic in L', for some formalised L,
the "global" concept of analyticity is thus clarified, even
though the predicate 'analytic in English' is not thus
clarified. But even those who would grant the legitimacy of
concept talk would doubtless agree to the principle that a
concept is clarified if and only if every linguistic expres-
sion which is known to express it is clarified. And if some
procedure doesn't clarify the expression 'analytic in
English' (an expression which presumably expresses the
"global" concept of analyticity) then it can't be said to
clarify the global concept of analyticity, in spite of the

fact that it makes clear other predicates which express this

129

global concept; i.e. predicates such as 'analytic in L'
(where L is some definite formalised language). These
remarks are a way of explaining Quine's request for a
definition of 'analytic in L', where 'L' is now not just a
name of some language but a genuine variable ranging over
languages. For if we were able to provide an exact and
clear definition of the two place predicate 'x is analytic
in L' then we would presumably be able to get an equally
clear one place predicate from it by replacing the variable
'L' with the name 'English' (provided that 'English' was at
least as clear as 'x is analytic in L') and thus get a clari—
fication of 'x is analytic in English', which could then be
of use in deciding, for example, whether 'All green things
are extended' is analytic.

Quine's request for a definition of 'analytic in L',
for variable 'L', has provoked a response by Bohnert which
brings up an important issue relevant to our present concerns
and which needs to be discussed. In discussing the binary
predicate 'S is analytic for L' Bohnert says this:

As Quine pictures it, it would seem that this

relational predicate would have, itself, to be a

logically determinate concept since a logically

determinate one place predicate is to be gotten

from it by simple substitution of 'L ' for 'L'.

This would require complete specific1ty as to

the structures of all languages in the range of

'L', obtainable, presumbaly, only by recursive

methods, which would be hard to reconcile with

Quine's desire for extreme generality. Presum-

ably a better way of posing the metalogical

problem would be to find a definition of 'x is

an analyticity concept for L' in terms of

adequacy conditions to be met by the definition

of x in the metalanguage of L. Even here,
however, the range of language would have to

130

be exactly delimited by fixing certain general

features essential to being a language and the

cry would go up again that the philosophically

essential features lay in the non—formally

expressed reasons for specifying these features

in just the way they were.

Quine is portrayed here as asking for the impossible.
On the one hand Quine asks for a clarification of analyticity
and on the other hand he is portrayed as refusing to accept
a procedure which would make it perfectly clear. But this
dilemma into which Quine is being placed is a false one.
As far as I know Quine doesn't require that the predicate
'S is analytic for L' to be logically determinate. The most
that can be attributed to Quine on the basis of his writings
is a request that the predicate 'S is analytic in English',
which would be gotten from the binary predicate by substi-
tuting 'English' for 'L', be clearer than it is at present.
But the fulfillment of this request doesn't require that we
use recursive methods in the definitions of the terms
involved, or that we require complete specificity as to the
structures of all languages in the range of 'L'. Quine seems
comfortable enough (in the context of his complaint) with
the notion of language, or at least with the notion of
sentence of English. And presumably he would be happy if
the notion of analyticity were to be made as clear as the
notion of sentence of English, even though this last notion
could be claimed to be not completely clear. Since '8 is
a sentence of English' is not a logically determinate

predicate there is no reason to suppose that Quine is

requesting 'S is analytic in English' to be one either.

131

If we want to claim that a definition of 'analytic in
LO', for some formalised L0, throws light on the predicate
'analytic in English' then the main question we have to
concern ourselves with is the question of what the relations
should be between these two predicates in order for light
to be so thrown. The way Quine pictures it, the only thing
that these two predicates have in common is the joint use

5 This doesn't seem to be

of the syllables 'analytic'.
enough.

We obviously can't require that the two predicates in
question he coextensive (unless, of course, LO were English
itself). And since synonymy guarantees coextensiveness we
also can't require that these two predicates be synonymous.
But we can require that the extensions of these two predi-
cates be related in a way to be indicated below.

Bohnert has emphasized the autonomy of meaning of a
semantical predicate within the framework of a formalised
language.6 This autonomy means that we do not have to have
a previous understanding of, or familiarity with, a semantical
predicate defined for a formal language in order to say that
we understand the predicate.as applied to the formal language
in question. One way to emphasize this autonomy would be to
use, as the defined predicate, not a previously familiar
expression (such as the 'analytic' in 'analytic in LO') but
a totally unfamiliar one. In order to emphasize the autonomy

of a predicate such as 'analytic in L ' Quine suggests

0
that we should instead "rewrite" it as 'K' so as not to

132

misleadingly suggest that by clearly defining 'analytic in
L0' we have thrown any light on the predicate 'analytic in
English'.7 80 let us use a predicate such as 'K', or
'sentence of kind K', and see what can be said about the
power of such a predicate, defined for a formal language,
to clarify the predicate 'analytic in English'.

Suppose we have a formalised language L0 and a definition
of 'sentence of kind K' such that this predicate is clear at
each sentence of L0. Suppose further that there is a known
function (perhaps only a partial one) which takes sentences
of English as arguments and sentences of L0 as values such
that a sentence of L0 is the value of the function for a
given sentence of English as argument if and only if the
sentence of L0 is what may be called "the best paraphrase"

of the English sentence into L (The best paraphrase should

0'
not be thought of as preserving complete synonymy between
some English sentence and its correlate in L0; see below.)

To say that 'sentence of kind K' is clearer than
'analytic in English' is to say that if 'analytic in English'
is clear at some English sentence S then 'sentence of kind
K' is clear at the best paraphrase of S, and that there are
English sentences at which 'analytic in English' is not clear
but at whose paraphrases 'sentence of kind K' is clear.

Consider a division of all English sentences into two
classes; those at which 'analytic in English' is clear, and

those at which it is not. There may be those who might want

to claim that the predicate"analytic in English' is clear

133

at x' is itself unclear at some English sentences. This

is equivalent to saying that there are borderline cases of

the predicate"analytic in English' is clear at x' or,

metaphorically speaking, that clarity and obscurity them-

selves have fuzzy edges. We could even permit these fuzzy

edges as long as "analytic in English' is clear at x' is

clearer than 'analytic in English'. And indeed it is clearer
since there obviously are English sentences (such as 'Everything
green is extended') at which the predicate 'analytic in

English' is clearly unclear. However, let us assume for the

sake of simplicity that 'analytic in English' is, at each

sentence of English, either clear or not clear. That is,

let us assume that clarity and obscurity don't have fuzzy

edges. The set of English sentences at which 'analytic in

English' is clear will contain the following kinds of

sentences.

1. Those which are definitely believed to be analytically
true.

2. Those which are definitely believed to be analytically
false.

3. Those which are definitely believed to be synthetic.

4. Those which do not belong to categories 1 through 3 but
which are such that we know what to do go tOgain
evidence for the claim that they are analytic or synthetic.

An example of the kind of sentence in subclass 1 might
be the sentence"All bachelors are unmarried'. An example

of subclass 2 might be the sentence 'No bachelor is unmarried'.

134

Examples of subclass 3 are the sentences 'I have a green
car' and 'I don't have a green car'. An example of sub-
class 4 might be the sentence 'All men are mortal'. (We
have to consider sentences of kind 4 because the analyticity
or syntheticity of a sentence is not always immediately
obvious.) At the outset we may not know whether this sen-
tence is analytic or synthetic for a person at a given time.
But if, upon examination, it is discovered that the person
is using the words 'man' and 'mortal' in such a way that he
considers it logically impossible for an object to be both
a man and mortal, or if he doesn't consider it to be logically
impossible, then we will have discovered that the sentence
is analytic or synthetic for him, as the case may be. If
however, it is determined that the person just doesn't know
whether or not he considers it logically impossible for
something to be both a man and immortal, then we will classify
the sentence as belonging to none of the four classes described
above but rather to the class of sentences at which 'analytic
in English' is unclear. (More precisely, (see Section 7 of
Chapter 2) the predicate in question is 'analytic for English
speaker P at time t' or, if we want to consider English as
a set of closely related languages, 'analytic in that dialect
of English which the speaker is using'. However, I will omit
such explicit relativization for the sake of perspicuity.)

Let us assume that to each sentence we can apply either
the kind of procedure which we imagined being applied to the

sentence 'All men are mortal' or some other procedure which

135

will help us to determine whether the speaker believes a
sentence to be analytic. (Another procedure which might be
used to help a speaker realize that he believes some sentence
to be analytic is to show that the sentence is a logical
consequence of some sentences which he already believes to
be analytic.) If, upon application of some appropriate
procedure, a sentence would be believed by our speaker to
be analytic (or synthetic) then this sentence will belong to
the class of sentences at which 'analytic in English"is
clear. But if, upon application of some appropriate
procedure, our speaker says that he just doesn't know whether
or not he believes that the sentence is analytic (or synthetic)
then this sentence will belong to the class of sentences at
which 'analytic in English' is unclear. (I emphasize 'is'
because it is possible for a predicate to be clear at an
object without the speaker believing that it is clear at
that object (because he has never considered the question of
whether the predicate applies to that object). It is also
possible for a predicate to be unclear at an object without
a person believing that it is unclear at that object.
Because we, with our limitations, typically consider so
relatively few objects (compared to the many that there are)
it is also possible for a predicate to be more unclear than
we think it is.)

Another feature of the sets of sentences of kind 1
through 4 is that these sets are not necessarily mutually

exclusive; i.e. we must leave open the possibility that our

136

speaker's analyticity intuitions are inconsistent. One way

in which they might be inconsistent is as follows. Suppose

the speaker firmly believes some sentence S to be synthetic,
but that it is then shown that S is a logical consequence

of sentences which he believes to be analytic. Assuming that

our speaker believes that only analytic sentences can be

logical consequences of analytic sentences and that he believes
the claim that S is such a logical consequence we will end

up with S being believed to be both analytic and synthetic.

In what follows I will, in order to avoid the tiresome
parenthentical expression 'or would be believed to be, upon
appreciation of some appropriate procedure', coin the
expression 'believesw' an expression which will be considered
to be equivalent to 'believes, or upon application of some
appropriate procedure, would believe to be'.

Let us assume that the predicate 'sentence of kind K'

is defined for a formalised language L0 in such a way that

it is clear at each sentence of that language. In order for

this predicate to be called an analyticity predicate it will

have to fulfill the following conditions.

1. If a sentence of English is believedw to be analytically
true (false) in English then its best LO paraphrase is
believedw to be true (false) and a sentence of kind K.

2. If a sentence of English is believedw to be synthetic in

English then its best L paraphrase is believedw to be

0
not a sentence of kind K.

137

Notice that the question of whether a sentence is belivedw
analytic in English is independent of the question of whether
a sentence of L0 is believedw to be a sentence of kind K.

We have some predicate defined for L0. We then look to see
whether this predicate satisfies conditions 1 and 2. Whether
or not a sentence of L0 is believedw to be a sentence of

kind K is a direct or indirect result of linguistic conven-
tion, or linguistic fiat. That is, it is the result of the
definition of 'sentence of kind K', a definition which is
itself true by fiat. But after this definition is laid down
it is no longer a matter of convention or fiat whether or

not the relations between English and L0 are such that condi-
tions 1 and 2 are satisfied.

However, when we come to consider those English sentences
at which 'analytic in English' is not clear, convention or
fiat enters the picture again, although in a different way.
For these sentences we stipulate, by fiat, that they shall
be believed analytically true (false) if their best para-
phrases are believedw true (false) and of kind K, and also
that if their best paraphrases are believedW to be not of
kind K then they shall be believed to be synthetic. Let us
call this stipulation the explication stipulation.

The basic idea behind the way in which this procedure
clarifies 'analytic in English' is as follows. If the
predicate 'sentence of kind K', defined for the sentences

of L0, fulfills conditions 1 and 2 then the structure of L0

mirrors that of English well enough at those English sentences

138

at which 'analytic in English' is clear to warrant the
claim that the predicate 'sentence of kind K' is an analy-
ticity predicate. It is hard to see what more could be
required to support such a claim. And if any less were
required then it is hard to see what warrant there would be
for calling 'sentence of kind K' an analyticity predicate.
(See below for a discussion of the suggestion that we might
want to require less in order to permit the definition of
'sentence of kind K' to ”educate our intuitions".) Now for
those English sentences at which 'analytic in English' is
not clear the explication stipulation, together with the
fact that 'sentence of kind K' is clear at each sentence of
L0, provides us with a search procedure for 'analytic in
English' at each sentence of English, or at least at "more"
of those English sentences than we had before. I

Thus, 'analytic in English' is clearer after the intro-
duction of the procedure described than it was before its
introduction. I stop short of saying that this procedure
makes 'analytic in English' completely clear (i.e. clear at
each English sentence) because of the fact that the predicate
'x is the best paraphrase of y' may not be clear at each
ordered pair, the first memeber of which is a sentence of L0
and the second member of which is a sentence of English.
In other words there may be English sentences which we just
don't know how to paraphrase into L0. In this case the
function corresponding to 'x is the best L0 paraphrase of y'

will be only a partial function.

139

(I should take this opportunity to mention that I am
assuming that if 'analytic in English' is clear at some
English sentence then we will know how to paraphrase this
sentence into L0' Without this assumption it would seem
difficult to support the claim that the use of an artificial
language can clarify the predicate 'analytic in English'.

Now I think that this assumption is true on independent
grounds. For, if we know the meaning of an English sentence
well enough to believew that it is analytic or synthetic then
it would seem that we ought to know it well enough to know
how to paraphrase it into L . But if it should turn out
that this is not so then weocan still repair the situation
by just arbitrarily taking any L0 sentence believed to be of
kind K to be the paraphrase of an English sentence we don't
know how to paraphrase and which is believedw to be analytic,
and also arbitrarily taking some L0 sentence which is
to be not of kind K to be the-paraphrase of any English
sentence we don't know how to paraphrase and which is believedw
to be synthetic. This procedure would, of course, still
not make the function associated with 'x is the best L0
paraphrase of y' a total function. It would still be a
partial function because there still might be English
sentences which are neither believedw to be analytic nor
believedW to be synthetic and which we don't know how to
paraphrase into L0.)

It ought to be remembered, at this point, that our task

is not to clarify 'x is the best paraphrase of y' but to

answer the question of how the use of an artificial language

140

can clarify 'analytic in English' gixsp the assumption that
this latter predicate is not as clear as 'x is the best
paraphrase of y'. (But see below for one condition which I
want to place on the paraphrasing function.) Moreover, this
assumption certainly seems to be true, for there seem to be
many English sentences we know how to paraphrase (into some
appropriate L0) but at which 'analytic'is not clear. The
complete clarity of every notion is probably an impossible
goal. But the clarification of some notions in terms of
others which are clearer not only seems possible but consti-
tutes one of the important goals of philosophy.

The general procedure outlined above can be said to
fill in analyticity gaps, in analogy with the process of
filling in truth value gaps provided by the process of para-
phrasing English sentences into an artificial language.

The one procedure seems neither more nor less acceptable
than the other.

I would now like to take up several points merely
alluded to earlier. The first of these points concerns a
condition which I would want to place on the paraphrasing
function. The condition is that paraphrasing should pre-
serve belief. What this means is that if a sentence of
English is believed to be true (false) then its best L0
paraphrase is believed to be true (false), (although not
necessarily conversely). It also means that if a sentence
of English is such that it would be believed to be true
(false) upon presentation of all the available evidence then

its best L0 paraphrase is also believed to be true (false).

141

Notice that this condition does not require that the relation
between some English sentence and its best paraphrase be one
of synonymy.

A reason for laying down the paraphrasing condition is
that if we don't do so then we will be blurring an important
distinction between clarification and rational reconstruction.
If, for example, someone paraphrases the sentence 'Whales
are fish' into a scientific language in which the paraphrase
of this sentence is (believed) false then he should not be
considered as clarifying the ordinary notion of fish but
rather as attaching a new meaning to 'fish' so as to be able
to more easily come up with new and simple laws about nature.
But there is a big difference between clarifying the meaning
which we presently attach to a word and attaching a new
meaning to that word. In the one case we could be said to be
sharpening up the present meaning of a word. But in the
other case we are just ignoring or riding roughshod over this
present meaning. (Ordinary language philosophers seem to
appreciate this point more than some of their more formalis-
tically inclined bretheren.) Rational reconstruction has its
place. But it should not be confused with clarification.8

The thoughts in the previous paragraph bring me to a
similar point about the clarification of analyticity which
was alluded to earlier. It might be suggested that conditions
1 and 2 are too strong because they don't permit the para-
phrasing of a sentence into L to "educate our intuitions".

0
According to this suggestion we should permit at least some

142

English sentences which we believew to be synthetic to be
paraphrased into an L0 sentence which is believedw to be of
kind K, and thus to "educate our analyticity intuitions.
There seems to be at least two things wrong with this
suggestion. First of all, I suspect that it may issue from
a confusion between Clarification of 'analytic' and a
rational reconstruction of it. Secondly, and more
importantly, this suggestion seems to issue from a neglect
of the fact that 'sentence of kind K' is completely auto-
nymous within the framework of the artificial language. One
way to appreciate this autonymy is to recall that the
question of whether an L0 sentence is believedw to be of
kind K is independent of the question of whether an English
sentence is believedw to be analytic. Now certainly if
'sentence of kind K' fulfills conditions 1 and 2 then we
can say that it is an analyticity predicate for L0. And
under these conditions someone might even say that 'sentence
of kind K' means, for all intents and purposes, 'analytic
in LO'. But the point is that if conditions 1 and 2 are
psi fulfilled (as the suggestion about "educating our analy-
ticity intuitions says they may not be fulfilled) then
'sentence of kind K' is simply not an analyticity predicate
for L0 and thus cannot be said to mean 'analytic in LO'.

I can imagine someone responding to this by saying
that if there are "strong analogies" between 'analytic in
English' and 'sentence of kind K' then this is enough to

show that this latter predicate is an analyticity predicate

143

for L0 and that it means 'analytic in L0' even though there
may be some English sentences which we believew to be syn-
thetic but whose L0 paraphrases are believedw to be of kind
K.9 My response to this is that if we accept this "strong
analogies" criterion of adequacy the result is that we are
not clarifying our intuitions about 'analyticity in English'
but are just ignoring them.

But, it may be objected, what of our earlier claim that
we must leave Open the possibility that our intuitions about
'analytic in English' are inconsistent? Doesn't this show
that conditions 1 and 2 are too strong? Consider, for
example, the possibility that we have very strong reasons for
believing that some English sentence is analytic and equally
strong reasons for believing that it is synthetic. (To help
the imagination think of 'analytic' in analogy with 'is a
member of itself', 'synthetic' in analogy with 'is not a
member of itself', and some English sentence in analogy with
the class which contains all and only those classes which are
not members of themselves.) Wouldn't this show that we
should ride roughshod over our analyticity intuitions in
this case, at least, by paraphrasing this sentence which we
strongly believe to be synthetic in English into an L0
sentence which is believedW to be of kind K? And if we did
so then, although we would be satisfying condition 1, we
would be violating condition 2, but with good reason.

The flaw in this reasoning is contained in its implicit

assumption that 'sentence of kind K' is consistent in the

144

special sense that it is impossible for one to have
evidence that some L0 sentence is of kind K and equal
evidence that it is not of kind K. But we have no guarantee
that 'sentence of kind K' is consistent in this sense. And
if it is inconsistent in this sense then it is possible
for both conditions 1 and 2 to be satisfied.

Moreover, and this is an important point, if we did
have a guarantee that 'sentence of kind K' is consistent
in our sense then paraphrasing English sentences into an
artificial language could never be one of the means by
which to show that 'analytic in English' is inconsistent
in our sense. On this way of looking at the situation we
would always have to blame the paraphrasing rather than
conclude that our analyticity intuitions are inconsistent.
What I mean can be seen by considering the following kind
of case. We have some English sentence 8 which, let us
suppose, we believe to be synthetic. We then paraphrase
it into an L0 sentence S0 which we believe to be not of
kind K. So far so good. But then later on we come across
equally good evidence that it is of kind K. (Think, by
analogy, of a situation where one day someone gives an
argument for the claim that Russell's set (i.e. the set of
all sets that are not members of themselves) is not a
member of itself, and so believes that it is not a member
of itself. But he has a short memory and the next day he
gives an equally good argument for the claim that Russell's

set is a member of itself, and so believes that it is a

145

member of itself. Upon being reminded of the argument of
the previous day he concludes that his notions of set and
set membership are inconsistent. (Although he has a short
memory he is not irrational.))

Suppose that after careful checking we are just forced
to the conclusion that 'sentence of kind K' is inconsistent.
What then are we to say about the English sentence S? Accor-
ding to the suggestion that I want to reject, one of the
things we should never conclude is that our analyticity
intuitions are inconsistent. According to this suggestion
we should conclude, instead, that S0 was not really the best
paraphrase of S. But if the only reason for such a conclusion
is that we don't take this route then we would have to con-
clude that our analyticity intuitions are inconsistent then
my answer is that this is a poor reason. It is poor because
it just begs the question in favor of the claim that our
analyticity intuitions are consistent. (Notice, incidentally,
that if we did conclude that 'analytic in English' is incon-
sistent then rational reconstruction, as opposed to clarifi-
cation, would be in order.)

Let us turn now to some issues connected with the
interpretation of the artificial language. There is one
point which needs to be mentioned, especially for the sake
of those who feel uncomfortable with the notion of an arti-
ficial language. I haVe, for the sake of generality,
considered the case of an artificial language none of whose

sentences are sentences of English. A language, artificial

146

or otherwise, has to have an interpretation in order to be
a language. A language can have an interpretation without
one knowing what that interpretation is. (An example in my
case is Spanish, 3 language whose interpretation I am
ignorant of.) But, of course, for a language to be useful
one must know its interpretation. As mentioned earlier,
what exactly one is doing when one gives an interpretation
of an artificial language can be described in more than one
way. But the minimum requirement for an interpretation of
a language is that it should give us the truth conditions
for each of its sentences.10 Typically, this is done by
giving these truth conditions in terms of a metalanguage we
really understand. For English speakers this metalanguage
will be English. But if an artificial language is to have
any phiIOSOphical use we cannot permit the metalanguage to
contain just any sentence of plain old ordinary English.
Rather it must contain only those English sentences whose
syntactical structure and truth conditions are themselves
unproblematic. Nothing would be gained if we paraphrased
the sentence 'The present king of France is bald' into the
LO sentence 'Fa' whose truth conditions were given by saying
that 'a' denotes the present king of France, 'F' denotes
the class of bald things, and 'Fa' is true just in case the
present king of France is bald, or by paraphrasing 'The
absolute is lazy' into this same sentence whose truth
conditions were given by saying that 'a' denotes the absolute,
'F' denotes the class of lazy things, and 'Fa' is true just

in case the absolute is lazy.

147

Whether the sentences of an artificial language are
sentences of (some portion of) English or whether they are
sentences composed of artificially constructed symbols, the
end result of paraphrasing a sentence of English into an
artificial language can be described as either a direct or
indirect paraphrasing of an English sentence into some
restricted and unproblematic part of English.

We need not think of the relation between a sentence S
and its best paraphrase S' as being one of synonymy. In our
paraphrasing condition we have already mentioned one res-
traint on the relation between S and S'. A further aspect
of this relation is well expressed by Quine.

Its relation to S is just that the particular
business that the speaker was on that occasion
trying to get on with, with help of 8 among other
things, can be managed well enough to suit him

by using 8' instead of S. We can even let him
modify his purposes under the shift, if he pleases.

Hence the importance of taking as the paradigmatic
situation that in which the original speaker does

his own paraphrasing, as laymen do in their routine
dodging of ambiguities. The speaker can be

advised in his paraphrasing, and on occasion he can
even be enjoined to accept a proposed paraphrase or
substitute another or hold his peace; but his choice
is the only one that binds him. A foggy appreciation
of this point is expressed in saying that there is no
dictating another's meaning; but the notion of there
being a fixed, explicable, and as yet unexplained
meaning in the speaker's mind is gratuitous. The
real point is simply that the speaker is the one to
judge whether the substitution of S' for S in the
present context will forward his present or evolving
program of activity to his satisfaction.

(Some of what Quine says here is pretty vague. But when
he talks of letting the speaker modify his purposes under

the shift, I would disagree if this includes letting him

148

change his purpose from clarification to rational recon-
struction.)

In this passage Quine rejects the notion of there
being a fixed, explicable, and as yet unexplained meaning
in the speaker's mind. Thisxdew of Quine's is familiar,
and.is part and parcel of hisrejectionof the notion of
meaning and of his thesis of the indeterminacy of translation.
Closely similar to the view that a sentence has a fixed
unique meaning is the view that there is a fixed unique ideal
language into which all problematic sentences are to be para-
phrased. Whether these two views are correct is not part
of our present concerns. But I mention them only to set the
stage for another comment which is part of our present
concerns. Even if there were a unique ideal language into
which all problematic sentences were to be paraphrased (call
this language L0) the analyticity of a sentence of English
at which 'analytic in English' was originally unclear would
still be relative to the definition of 'sentence of kind K'.
Even under the assumption of an ideal language there might
still be many different definitions of 'sentence of kind K',
all of which would satisfy conditions 1 and 2, but which
would differ in assignments of analyticity to those English
sentences at which 'analytic in English' was originally
unclear. But this is unobjectionable since, before the
introduction of an artificial language, those sentences at
which 'analytic in English' was unclear were neither analytic

nor synthetic in English. For these sentences the claim that

149

any given one of them is analytic is neither true nor false.
Consequently it doesn't matter if one definition of 'sentence
of kind K' has the result that one of these sentences is
analytic and another definition has the result that it is
synthetic. As long as these definitions of 'sentence of

kind K' satisfy conditions 1 and 2 these predicates will be
analyticity predicates and will be equally acceptable, provided
that there are not further reasons for preferring one over

the other.

Chapter.4
Summary

Although analyticity is a semantical notion, its primary
interest lies in its epistemological function. Calling a sen-
tence analytically true is supposed to account for our belief
in or knowledge of its truth. But, as we have seen, when we
get more specific about the vague notion of "accounting for"
we see that the appeal to analyticity doesn't seem to provide
such an account.

In the first place, we have seen in detail why analyticity
has no power to explain our belief in or knowledge of the truth
of a sentence. This lack of explanatory power is analogous to
the fact that appealing to the dormitive virtue of opium doesn't
explain why opium puts peOple to sleep.

In the second place, the claim that a sentence is analy-
tically true doesn't provide a good justification for the claim
that it is true, since the claim that a sentence is analytic is
never known with more certainty or believed to a greater degree
than the claim that it is true. This criticism remains even if
we take the general view that no sentence is, or can be, known
with complete certainty. The criticism is not that analyticity
has justificatory power only if the analyticity of a sentence
can be known with complete certainty. The fact that we may
change our minds about whether a sentence is analytic and thus
withdraw previous attributions of analyticity is not part of the

criticism. The criticism is rather that at any given time

150

151

we are no surer of the analyticity of a sentence than we
are of its truth. It is this fact which robs analyticity
of any justificatory power which it might otherwise have.
Since analyticity is primarily an epistemological notion, I
consider the fact that this notion has the power to neither
explain nor justify to be the most serious of the criticisms
of analyticity which have been considered in this work.
Another criticism of analyticity which we considered is
the claim that this notion is not clear enough to be useful.
When used in an unreflective way, as it normally is, the
word 'unclear' can cover a multitude of sins. But in
Section 4 of Chapter 1 I have tried to isolate an important
sense of 'unclear' and have tried to show that analyticity
is unclear in this sense. One of the presuppositions of
my analysisof clarity is that if a predicate F is unclear
at an object a, in the sense that we neither believe it to
be true that a is an F nor believe it to be false that a is
an F, and furthermore do not know what to do to gain evidence
for the claim that a is an F, then the claim that a is an
F is neither true nor false. Now there are many sentences
which rival 'All bachelors are unmarried' in the extent to
which we believe them to be true, but are such that 'analytic'
is not clear at them. Such sentences are neither analytic
nor synthetic. Examples of such sentences might include
the axiom of choice and 'I exist'. (Let it not be said that
such sentences are perhaps synthetic a priori, for if they

are neither analytic nor synthetic then they are not synthetic

152

a priori either.) In Section 3 of Chapter 1 we saw that

the notion of analyticity has no explanatory power even in
the case of those sentences which were clearly analytic.

But now, when we come to consider the unclarity of 'analytic'
we can see a new way in which this unclarity might show a
further weakness in the explanatory power of analyticity.
For if there are sentences which are neither analytic nor
synthetic but which rival 'All bachelors are unmarried' in
the extent to which we believe them to be true then we may
well begin to suspect that perhaps even in the case of 'All
bachelors are unmarried' we do not really explain our belief
in its truth by saying that it is analytic. We may well
begin to suspect that 'analytic' has no more to it than the
less exciting word 'obvious'.

In Section 1 of Chapter 2 we saw that the attempt to
"explain away" the difficulty in deciding whether a sentence
is analytic by blaming this difficulty on the obscurity or
unclarity of the words in the sentence ultimately presupposes
the notion of intensional obscurity, a notion which is just
as unclear and problematic as that of analyticity.

The unclarity of 'analytic' and its cognates, such as
'means the same as' pops up to plague us again when we
consider the question of whether there are any analytically
true sentences. There certainly sssp to be such sentences.
And one characteristic which is often said to be a defining
characteristic of analytically true sentences is that if

anyone disagrees with us as to the truth of such a sentence

153

then he must be using this sentence with a different meaning
than it normally has. If this property is taken to be a
"defining' property of analyticity then the question arises
of whether there areany sentences which have this property.
In Section 5 of Chapter 2 I have taken an arbitrarily chosen
sentence which seems to be analytic (i.e. the sentence
'Everything is self identical') and have considered various
things which we might conclude from the fact that someone
disagreed with us as to the truth of this sentence. These
various possible conclusions are:
1. The person in question means something different by

his utterance than what it is normally taken to mean.
2. He doesn't mean something different.
3. The question of whether he means something different

is meaningless.

The question of the clarity or unclarity of the expression
'means the same as' enters the picture when we consider
conclusion 3. If someone said that nothing is self identical
then perhaps we might want to conclude that he is using
these words to mean something different than what we normally
take them to mean. If, however, he says that certain strange
particles are not self identical although everything else is,
then we may very well conclude that it is meaningless to ask
whether the sense~of his words are the same as the sense of
our words. How are we to decide whether or not he is using
these words in the same or a different sense? Needless to

say, it won't do to pound the table and say that he must be

 

154

attaching a different meaning to 'identical with' than we
normally do. For this is a question presumably to be
decided by evidence. I can, of course, dictate my own
meaning; i.e. I can declare by fiat that I am using two
expressions synonymously. But in the ipispsubjective case
we are considering the question of whether one person attaches
the same meaning to an expression as another person is a
question for which the request for evidence is relevant,
provided of course that the question is not meaningless.

And if the request for evidence is considered irrelevant
then we would have to conclude that the question of inter-
subjective synonymy is meaningless; i.e. the claim that our
speaker is attaching a different meaning to 'identical with'
is neither true nor false.

If conclusion 2 is taken then it follows that the sen-
tence 'Everything is self identical' is not analytically
true.

Conclusion 1 can be further subdivided into 2 possi-
bilities:
la. The person's utterance can be translated into a sentence

we already understand.
lb. His utterance cannot be translated into any sentence we

already understand.

Under 1b. we would still be free to say that in our
language it is analytically true that everything is self
identical. But this would be small comfort if we had reason

to believe that the person's utterance was true because of

155

the fact that it was part of a theory which was successful
in explaining and predicting phenomena.

In order to show that 'Everything is self identical'
is analytically true, in any interesting sense, the friends
of analyticity would have to show that conclusion la will
always be the most reasonable conclusion to draw from the
fact that someone disagrees with the truth of this sentence.
However, I know of no good reasons for such a claim.

The question of whether there are any analytically true
sentences can also be approached by considering the question
of whether there are any such things as meanings. For if
there are no such things as meanings then there are no sen-
tences which are true in virtue of their meaning or in
virtue of sameness of meaning of some of their constituent
parts.

The claim that there are no such things as meanings in
the required sense can be supported by a pragmatic argument
to the effect that postulation of meanings is not required
in order to describe successful translation and communication.
For if a theoretical kind of entity is not required in the
service of our descriptive and explanatory purposes then this
is good evidence that there are no entities of that kind.

In Section 5 of Chapter 1 I gave arguments for the claim
that the most that is required to describe successful trans-
lation and communication is the notion of coextensiveness,
or sameness of reference, and that the further notion of

synonymy or sameness of meaning is not required.

156

The notion of the utility of analyticity also arose in
Section 2 of Chapter 2 when we gave arguments for the claim
that even if the languages that you and I spoke were ideal
in the sense that 'analytic' was clear at each of their
sentences this would not necessarily result in an increase
in communication and understanding between us or reduce the
possibility of merely verbal disagreements between us. For
the fact that the language I speak has the same analytically
true sentences as the language you speak does not at all
guarantee that any given expression means the same to you as
it does to me. The belief that it does provide such a guarantee
may arise from neglecting the distinction between intrasub-
jective and intersubjective synonymy.

One way to supposedly guarantee that there are analytically
true sentences is to declare by fiat that a sentence shall
be analytically true or that two expressions shall be synonymous.
This seems to work especially well on expressions which do not
have any meaning prior to the fiat, as in the case where we
introduce a new symbol for abbreviatory purposes. Even here
however, Quine has argued that language works in such a way
that a sentence declared analytically true by fiat at one
time will become incorporated into the language in such a way
that the artificiality of its analyticity and of its truth
will be lost during subsequent deve10pments, and at a later
time it may suit our purposes to consider this sentence as
being falsified by further experiences. The friends of

analyticity may well want to say that if this happens then

157

it shows that we have changed our language. But such a
response merely begs the question of the intelligibility
of the distinction between changing our language and
changing our beliefs without changing our language.

The weakness of fiat in instituting truth can be seen
in a particularly clear light when we consider the case
where we have a set of sentences each one of which is
declared true by fiat. In such a case there is no guarantee
that we won't be able to later on derive a contradiction
from this set of sentences. And if we were to derive such
a contradiction then not all of the sentences in the set
would be true, in spite of the fact that all of them were
originally declared true by fiat. Notice that this last
point does not presuppose that the non-logical expressions
of the sentences declared true by fiat have any independent
meaning, or to put the same opint differently, the derivation
of a contradiction need not depend on the meanings, if any,
of the non-logical expressions in the sentences.

Another point that we made about the notion of truth by
fiat or analyticity by fiat was that this notion would not
help us in justifying a claim that a sentence already in
use (such as 'All bachelors are unmarried') is analytic
since we will be less certain about a historical claim to
the effect that this sentence was once declared true by
fiat than we will be about the claim that this sentence is

true, or analytically true.

158

A question we considered in Section 3 of Chapter 2
was the question of whether claims of interlinguistic
synonymy were settleable by legitimate standards of
scientific procedure. We examined Carnap's criterion of
synonymy and found that it had two defects. First of all
it was subject to a kind of circularity. In order to apply
this criterion of synonymy we had to already know, at least
in part, what the criterion was supposed to provide us with.
It was seen that this circularity was only partial, although
seeing it as partial required us to suppose the dubious
claim that when we learn a foreign language in the way a
native might we are attaching the same "meanings" to the
foreign expressions as the native does.

Secondly, we saw that Carnap's criterion doesn't
provide one of the things which we could reasonably want
such a criterion to provide us with; i.e. increased clarity
of the predicate 'synonymous', i.e. more clarity than we had
before the introduction of the criterion.

One of the most prominent characteristics of many
definitions of 'analytic' is that they indirectly presuppose
the very notion to be defined, or otherswhich are equally
problematic (and in the same way). One of these definitions
was Grice and Strawson's understanding-believing definition
of analyticity. One definition which didn't seem to have this
question begging characteristic was Grice and Strawson's
definition of sentential synonymy in terms of sameness of
confirming and disconfirming experiences, a definition which

grows out of the verifiability theory of meaning. (We could

159

then define analyticity in terms of sentential synonymy.)
This definition was seen to be not precise enough, although
worthy of further development. However, there did not seem.
to be any hope that such a further development would provide
much material for answering the objection that the notion
of analyticity has neither explanatory nor justificatory
power.

Another definition which didn't seem to presuppose
notions which are equally problematic as the notion being
defined was a definition of intralinguistic intrasubjective
sentential synonymy in terms of stimulus synonymy. We saw
that although the latter notion is a better reconstruction
of the former notion than Quine thinks it is, it still
falls short.

In Section 9 of Chapter 2 we examined the hope that
perhaps some day the notion of meaning could be made
legitimate by the process of correlating meanings with brain
states. However, the carrying out of such a program seemed
to rest on the dubious claim that we learn the use of our
words in a step-by-step one-by-one fashion. It was also
seen to rest on a confusion between intersubjective and intra-'
subjective synonymy as well as on the subtractive fallacy
which claims that if an expression is meaningful or signi-
ficant then there is a meaning which it has.

Since Carnap is the foremost friend of analyticity'and
since he has considered artificial languages to be a most

useful tool for the investigation of philosophical problems

160

we have, in Chapter 3, considered the way in which the use
of an artificial language can clarify the notion of analy-
ticity. We also gave a criterion of adequacy to be used in
judging whether a predicate, defined for a formal language,

is an analyticity predicate for that language.

10.
11.
12.
13.

FOOTNOTES

See Willard Van Orman Quine, Word and Object (Cambridge,
Mass., 1960), pp. 56, 57, 66, 67. Talk of "verbal links",
"verbal connections", and "semantic anchorings" seems no
no more enlightening than talk of synonymy. See also
Quine's reply to Stroud in Willard Van Orman Quine,
"Relies," in Words and Objections: Essa s on the Work
of W. V. Quine, eds. Donald DEVidson and Jahho_Hint1EEa
(Dodfehht, Holland, 1969), pp. 316-319. Here Quine
points out the utility of a translator being able to
deal with sentences wholesale by abstracting shared
skeletal forms. While this utility is undeniable, it
doesn't seem to provide a basis for an explanation of
the analytic appearance of apparently analytic sentences.

 

 

Gilbert Harman, "Quine on Meaning and Existence, I,"
Review pi Metaphysics, ii (1967-1968), pp. 125, 127.

 

Ibid., p. 127.

Quine, Word and Object, p. 38. Hereafter referred to
as "WO".

 

See Rudolf Carnap, An Introduction to the Philosophy of
Science, ed. Martin—Gardner (New Yo?k,19667, pp. 260:—
267. Arthur Pap, Semantics and Necessarijruth (New
Haven, 1958), p. 423. RudolfICarnap, "Meaning and
Synonymy in Natural Language," in Meaning and Necessity
(Chicago, 1967), p. 233.

 

 

 

 

Willard Van Orman Quine, "Two Dogmas of Empiricism",
in From s Logical Point pi View (New York, 1963), p. 36.

 

Ibid.

Willard Van Orman Quine, "Carnap and Logical Truth," in
The Philosophy of Rudolf Carnap, ed. Paul Arthur Schilpp,
The Library of hiving Philosophers, XI (London, 1963),

p. 390.

 

 

Philip Kitcher has taken this view (in private conversa-
tion) of what Quine is objecting to in the notion of
analyticity. I owe this interpretation of Quine to
Kitcher.

Harman, p. 129.

Quine, "Carnap and Logical Truth", p. 390.

See Carnap's reply to Quine in Schilpp, p. 917.

Ibid., p. 922.

161

162

 

 

 

l4. Harman, pp. 125, 127.

15. Quine, "Two Dogmas of Empiricism", p. 32.

16. H. P. Grice and P. F. Strawson, "In Defense of a Dogma",
in Analyiicity, eds. James E. Harris, Jr. and Richard H.
Severens (Chicago, 1970), p. 69.

17. Ibid., p. 58.

18. Harman, p. 138.

19. Ibid.

20. Quine, ”Two Dogmas of Empiricism", p. 26.

21. W0, p. 207.

22. Quine, ”Carnap and Logical Truth", p. 403.

23. W0, p. 206.

24. Quine, "Carnap and Logical Truth", p. 402.

Chapter 2

1. This kind of move is made by Carnap in "Meaning and
Synonymy in Natural Languages”, p. 240.

2. Grice and Strawson, p. 69.

3. Ibid.

4. This distinction is similar to Carnap's distinction
between extensional and intensional vagueness in
”Meaning and Synonymy in Natural Languages." I say
'similar to' rather than 'identical with' because it
is not sufficiently clear to me, from what Carnap says,
whether his notion of extensional vagueness is such
that a predicate F is extensionally vague for some
person if an actual object about which the person is
undecided as to whether it is an F must also be an
object which the person believes to be actual. At
any rate, my definition of extensional obscurity doesn't
make this belief requirement with respect to the
existence of the object concerned.

5. See WO, pp. 194, 195 for an explanation of this notion.

6. Cf. Willard Van Orman Quine, "Notes on the Theory of

Reference," in From p Logical Point pi View (New York,
1963), p. 138.

 

 

10.
11.
12.
13.
14.
15.
16.
17.

18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.

163
As is done by Carnap in "Meaning and Synonymy in
Natural Languages”, p. 240.

Rudolf Carnap, "Meaning Postulates", in Meaning and
Necessiiy (Chicago, 1967), p. 225.

 

 

Gilbert Ryle, "Discussion: Meaning and Necessity",
Philosophy, 24 (1949), pp. 69-76.

 

Carnap, "Meaning and Synonymy in Natural Languages".
Ibid., p. 238.
Grice and Strawson, p. 67.

Ibid.

 

Ibid., p. 66.

W0, p. 66.

W0, p. 206.

Quine in Schilpp, p. 387. See also WO, pp. 58, 59.
Quine makes his comments about pro-logical peoples
with reference to truth functional tautologies. But
it seems to me that '(x)(x=x)' is no less a logical
truth than a statement of the form 'p or not p'.
Moreover, one could construct a story (albeit more
complicated) similar to the one about the strange
particle theory using some truth functional tautology
instead of '(x)(x=x)'.

Quine, ”Carnap and Logical Truth", pp. 394, 395.
Quine, ”Two Dogmas of Empiricism", p. 26.

Quine, "Carnap and Logical Truth", pp. 394, 395.
Quine, "Two Dogmas of Empiricism", p. 38.

Ibid., p. 41.

Ibid., p. 40.

Grice and Strawson, pp. 71, 72.

W0, p. 64.

' Ibid., p. 62
‘Ibid., p. 63.

Willard Van Orman Quine, "On the Reasons for Indeter-
minacy of Translation", Journal pi Philosophy, 67 (1970),
pp. 178-183.

 

164

 

 

 

 

 

29. W0, p. 54.

30. Ibid., p. 63.

31. Ibid.

32. See Quine's comment on p. 206 of WO against what he calls
the subtractive fallacy involved in the claim that if an
expression is meaningful then there must be a meaning
which it has.

Chapter 3

1. See Quine, Section VIII, in "Carnap and Logical Truth"
and Section 4 of "Two Dogmas of Empiricism."

2. The Encyclopedia pi Philosophy, ed. Paul Edwards,

Vols. I and II (New York, 19677, p. 168.

3. E. W. Beth, "Carnap's Views on the Advantages of
Constructed Systems Over Natural Languages in Philosophy
of Science", in The Philoso h of Rudolf Carna , ed.
Paul Arthur Schilpp, The Lihrary—of E1V1ng PhlIosophers,
XI (London, 1963), footnote 56, p. 492.

4. Herbert G. Bohnert, "Carnap's Theory of Definition and
Analyticity”, in The Philosophy pi Rudolf Carnap, ed.
Paul Arthur Schilpp, The Library of Living Philosophers,
XI (London, 1963), p. 421.

5. Quine, "Notes on the Theory of Reference", p. 138.

6. Bohnert, pp. 416, 147.

7. Quine, "Two Dogmas of Empiricism", p. 33.

8. Thus, I agree with Hanna in upholding the distinction

between clarification and rational reconstruction. See
Joseph Hanna, "An Explication of 'Explication'," in
Philosophy of Science, 35 (1968), pp. 28-44. On page

 

41 of his ahhicle Hanna lists four possible attitudes
towards what in his paper is called the correspondence
condition. Hanna opts for view (c). My own view seems
to be closer to view (a). However, I find it difficult
to compare his views with mine. This is due to several
factors. First of all, to the extent that the notion
of a definitive intension is clear to me, it seems that
'analytic' has no definitive intension. Secondly, it
is not sufficiently clear to me whether Hanna's cor-
respondence condition requires the explicandum and
explicatum to merely be believed to be coextensive or

 

10.

11.

165

to actually pp coextensive. (The remarks in section 8
of his paper, as well as his example, on page 37, about
objects which are known to be fish suggest the latter
(since knowledge impl1es truth).). Thirdly, in the

case of 'analytic in English' I find it difficult to
make a distinction between a situation in which this
explicandum has been given a "preliminary clarification"
and a situation where it has not. Has this predicate,
at the present state of philosophical research, already
been given a "preliminary clarification"?

For an exposition of the "strong analogies" view, and
of the related idea that 'x is an analyticity predicate
for L' is a recipe term, see Bohnert, pp. 421, 422.

See Donald Davidson, "Truth and Meaning”, Synthese, i1
(1967), pp. 304-323 for a development of this idea.

 

WO, p. 160.

BIBLIOGRAPHY

BIBLIOGRAPHY

(This list includes only such works as happen to have been

alluded to, by title or otherwise, in the course of the book.)

Beth, E. W., "Carnap's View on the Advantages of Constructed
Systems Over Natural Languages in Philosophy of Science",
in The Philoso h of RudOlf carna , ed. Paul Arthur
SchiIhp. The E1hr§?y of E1V1ng Philosophers, XI
(London: Cambridge University Press, 1963), pp. 469-502.

Bohnert, Herbert G., "Carnap's Theory of Definition and
Analyticity", in The Philosophy of Rudolf Carnap, ed.
Paul Arthur Schilpp. The Library—of Living Philosophers,
XI (London: Cambridge University Press, 1963), pp.
407-430.

 

 

Carnap, Rudolf, hp Introduction to the Philosophy pi Science,
ed. Martin Gardner (New YorkT' Basic Books, 1966I.

 

 

, "Meaning and Synonymy in Natural Languages", in

Meaning and Necessity (Chicago: University of Chicago
Press, 19675, pp 33-247.

, "Meaning Postulates", in Meaning and Necessity
IChicago: University of Chicago Press, 1967), pp. 222-229.

 

 

Davidson, Donald, "Truth and Meaning", Synthese, i1, (1967),
pp. 304-323.

 

The Encyclopedia of Philoso h , ed. Paul Edwards, Vols. I
and II (New Yark: ColI1er-Macmillan, 1967).‘

Grice, H. P. and Strawson, P. F., "In Defense of a Dogma",
in Analyticiiy, eds. James F. Harris, Jr., and Richard
H. Severens (Chicago: Quadrangle, 1970), pp. 54-74.

 

Hanna, Joseph F., "An Explication of 'Explication'",
Philosophy pi Science, is (1968), pp. 28-44.

 

Harman, Gilbert, "Quine on Meaning and Existence, 1",
Review pi Metaphysics, ii (1967-68), pp. 124-151.

 

Pap, Arthur, Semantics and Necessary Truth (New Haven:
Yale UniversityIPress, 19587.

 

Quine, Willard Van Orman, Word and Object (Cambridge, Mass.:
M.I.T. Press, 1960).

 

166

167

Quine, Willard Van Orman, "Two Dogmas of Empiricism", in
From p Logical Point pi View (New York: Harper 6 Row,
1963), pp. 20-46.

 

, "Notes on the Theory of Reference", in From p
Logical Point pi View (New York: Harper 8 Row, 1963),
pp. 130-I38.

 

 

, "Carnap and Logical Truth", in The Philoso h of
Rudolf Carnap, ed. Paul Arthur SchiIhh. e L1 rah? of
Living PhilOSOphers, XI (London: Cambridge University
Press, 1963), pp. 385-406.

 

, "Replies", in Words and Objections: Essays on the
Work of W. X; Quine, eds. Donald Davidson and Jakko
H1ntikka—(Dordrec t, Holland: Reidel, 1969), pp. 316-319.

 

 

 

, "On the Reasons for Indeterminacy of Translation",
_____30urnal 2i Philoseehx. 91 (1970). pp. 178-183.

 

Ryle, Gilbert, "Discussion: Meaning and Necessity",
Philosophy, 24 (1949), pp. 69-76.

 

 

"‘IIIIIIIIIIIIIIIIII“