Reason, Truth and History

REASON, TRUTH AND HISTORY
Downloaded from https://www.cambridge.org/core. University of Aberdeen, on 12 Oct 2020 at 14:22:52, subject to the Cambridge
Cambridge
Core terms of use, available Books Online © Cambridge
at https://www.cambridge.org/core/terms. University Press, 2010
https://doi.org/10.1017/CBO9780511625398
Cambridge
https://doi.org/10.1017/CBO9780511625398
REASON, TRUTH AND HISTORY
Hilary Putnam
CAMBRIDGE
UNIVERSITY PRESS
Cambridge
https://doi.org/10.1017/CBO9780511625398
PUBLISHED BY THE PRESS SYNDICATE OF THE UNIVERSITY OF CAMBRIDGE
The Pitt Building, Trumpington Street, Cambridge CB2 1RP, United Kingdom
CAMBRIDGE UNIVERSITY PRESS

The Edinburgh Building, Cambridge CB2 2RU, UK http: //www.cup.cam.ac.uk
40 West 20th Street, New York, NY 10011-4211, USA http: //www.cup.org
10 Stamford Road, Oakleigh, Melbourne 3166, Australia
© Cambridge University Press 1981
This book is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.
First published 1981

Reprinted 1982, 1985, 1986, 1987, 1989, 1990, 1991, 1992, 1993,
1994, 1995, 1997, 1998
Typeset in Sabon
A catalogue record for this book is available from the British Library
Library of Congress Cataloguing-in-Publication Data is available
ISBN 0-521-23035-7 hardback

ISBN 0-521-29776-1 paperback
Transferred to digital printing 2004
Cambridge
https://doi.org/10.1017/CBO9780511625398
FOR RUTH ANNA
Cambridge
https://doi.org/10.1017/CBO9780511625398
Cambridge
https://doi.org/10.1017/CBO9780511625398
Contents
Preface IX
1 Brains in a vat 1
2 A problem about reference 22
3 Two philosophical perspectives 49
4 Mind and body 75
5 Two conceptions of rationality 103
6 Fact and value 127
7 Reason and history 150
8 The impact of science on modern
conceptions of rationality 174
9 Values, facts and cognition 201
Appendix 217
Index 219
Cambridge
https://doi.org/10.1017/CBO9780511625398
Note
A reader who is unused to technical philosophy, or who wishes

to gain an overview of the argument of this book, might well
start by reading Chapter 5 to the end of the book, and only then
return to Chapters 1 to 4.
Cambridge
https://doi.org/10.1017/CBO9780511625398.001
Preface
In the present work, the aim which I have in mind is to break the
strangle hold which a number of dichotomies appear to have on
the thinking of both philosophers and laymen. Chief among
these is the dichotomy between objective and subjective views of
truth and reason. The phenomenon I am thinking of is this: once
such a dichotomy as the dichotomy between 'objective' and 'sub-
jective' has become accepted, accepted not as a mere pair of cat-
egories but as a characterization of types of views and styles of
thought, thinkers begin to view the terms of the dichotomy
almost as ideological labels. Many, perhaps most, philosophers
hold some version of the 'copy' theory of truth today, the con-
ception according to which a statement is true just in case it
'corresponds to the (mind independent) facts'; and the philoso-
phers in this faction see the only alternative as the denial of the
objectivity of truth and a capitulation to the idea that all schemes
of thought and all points of view are hopelessly subjective. Inev-
itably a bold minority (Kuhn, in some of his moods at least;
Feyerabend, and such distinguished continental philosophers as
Foucault) range themselves under the opposite label. They agree
that the alternative to a naive copy conception of truth is to see
systems of thought, ideologies, even (in the case of Kuhn and
Feyerabend) scientific theories, as subjective, and they proceed
to put forward a relativist and subjective view with vigor.
That philosophical dispute assumes somewhat the character
of ideological dispute is not, of itself, necessarily bad: new ideas,
even in the most exact sciences, are frequently both espoused
and attacked with partisan vigor. Even in politics, polarization
Cambridge
https://doi.org/10.1017/CBO9780511625398.002
x Preface
and ideological fervor are sometimes necessary to bring moral

seriousness to an issue. But in time, both in philosophy and pol-
itics, new ideas become old ideas; what was once challenging,
becomes predictable and boring; and what once served to focus
attention where it should be focussed, later keeps discussion
from considering new alternatives. This has now happened in
the debate between the correspondence views of truth and sub-
jectivist views. In the first three chapters of this book I shall try
to explain a conception of truth which unites objective and sub-
jective components. This view, in spirit at least, goes back to
ideas of Immanuel Kant; and it holds that we can reject a naive
'copy' conception of truth without having to hold that it's all a
matter of the Zeitgeist, or a matter of 'gestalt switches', or all a
matter of ideology.
The view which I shall defend holds, to put it very roughly,
that there is an extremely close connection between the notions
of truth and rationality; that, to put it even more crudely, the
only criterion for what is a fact is what it is rational to accept.
(I mean this quite literally and across the board; thus if it can be
rational to accept that a picture is beautiful, then it can be a fact
that the picture is beautiful.) There can be value facts on this
conception. But the relation between rational acceptability and
truth is a relation between two distinct notions. A statement can
be rationally acceptable at a time but not true; and this realist
intuition will be preserved in my account.
I do not believe, however, that rationality is defined by a set of
unchanging 'canons' or 'principles'; methodological principles
are connected with our view of the world, including our view of
ourselves as part of the world, and change with time. Thus I
agree with the subjectivist philosophers that there is no fixed,
ahistorical organon which defines what it is to be rational; but I
don't conclude from the fact that our conceptions of reason
evolve in history, that reason itself can be (or evolve into) any-
thing, nor do I end up in some fancy mixture of cultural relativ-
ism and 'structuralism' like the French philosophers. The dichot-
omy: either ahistorical unchanging canons of rationality or
cultural relativism is a dichotomy that I regard as outdated.
Another feature of the view is that rationality is not restricted
to laboratory science, nor different in a fundamental way in
laboratory science and outside of it. The conception that it is
seems to me a hangover from positivism; from the idea that the
Cambridge
https://doi.org/10.1017/CBO9780511625398.002
Preface xi
scientific world is in some way constructed out of 'sense data'

and the idea that terms in the laboratory sciences are 'operation-
ally defined'. I shall not devote much space to criticizing opera-
tionalist and positivist views of science; these have been thor-
oughly criticized in the last twenty-odd years. But the empiricist
idea that 'sense data' constitute some sort of objective 'ground
floor' for at least a part of our knowledge will be reexamined in
the light of what we have to say about truth and rationality (in
Chapter 3).
In short, I shall advance a view in which the mind does not
simply 'copy' a world which admits of description by One True
Theory. But my view is not a view in which the mind makes up
the world, either (or makes it up subject to constraints imposed
by 'methodological canons' and mind-independent 'sense-data').
If one must use metaphorical language, then let the metaphor be
this: the mind and the world jointly make up the mind and the
world. (Or, to make the metaphor even more Hegelian, the Uni-
verse makes up the Universe — with minds — collectively — play-
ing a special role in the making up.)
A final feature of my account of rationality is this: I shall try
to show that our notion of rationality is, at bottom, just one part
of our conception of human flourishing, our idea of the good.
Truth is deeply dependent on what have been recently called
'values' (Chapter 6). And what we said above about rationality
and history also applies to value and history; there is no given,
ahistorical, set of 'moral principles' which define once and for
all what human flourishing consists in; but that doesn't mean
that it's all merely cultural and relative. Since the current state in
the theory of truth — the current dichotomy between copy theo-
ries of truth and subjective accounts of truth — is at least partly
responsible, in my view, for the notorious 'fact/value' dichot-
omy, it is only by going to a very deep level and correcting our
accounts of truth and rationality themselves that we can get
beyond the fact/value dichotomy. (A dichotomy which, as it is
conventionally understood, virtually commits one to some sort
of relativism.) The current views of truth are alienated views;
they cause one to lose one part or another of one's self and the
world, to see the world as simply consisting of elementary par-
ticles swerving in the void (the 'physicalist' view, which sees the
scientific description as converging to the One True Theory), or
to see the world as simply consisting of 'actual and possible
Cambridge
https://doi.org/10.1017/CBO9780511625398.002
xii Preface
sense-data' (the older empiricist view), or to deny that there is a

world at all, as opposed to a bunch of stories that we make up
for various (mainly unconscious) reasons. And my purpose in
this work is to sketch the leading ideas of a non-alienated view.
My Herbert Spencer Lecture, 'Philosophers and Human
Understanding' (given at Oxford University, 1979) overlaps the
present text, having stemmed from work in progress, as does the
paper ' "Si Dieu est mort, alors tout est permis". . . (reflexions
sur la philosophic du langage)', Critique, 1980.
A research grant from the National Science Foundation * sup-
ported research connected with this book during the years
1978-80. I gratefully acknowledge this support.
Thomas Kuhn and Ruth Anna Putnam have studied drafts of
this book and given me able criticism and wise advice. I have
been helped also by advice and criticism from many friends,
including Ned Block, David Helman, and Justin Leiber, and the
students in my various lectures and seminars at Harvard. Several
chapters were read as lectures in Lima in the spring of 1980 (a
trip made possible by a grant from the Fulbright Commission),
and Chapter 2 was actually finished during my Lima stay. I ben-
efited in this period from discussions with Leopoldo Chiappo,
Alberto Cordero Lecca, Henriques Fernandez, Francisco Miro
Quesada, and Jorge Secada. The entire book (in an earlier ver-
sion) was read as lectures at the University of Frankfurt in the
summer of 1980, and I am grateful to my colleagues there (espe-
cially Wilhelm Essler and Rainer Trapp), to my very stimulating
group of students, and my other friends in Germany (especially
Dieter Henrich, Manon Fassbinder, and Wolfgang Stegmuller)
for encouragement and stimulating discussions.
All of my colleagues in the Harvard Philosophy Department
deserve to be singled out for individual thanks. In recent years
Nelson Goodman and I have detected a convergence in our
views, and while the first draft of the present book was written
before I had the opportunity to see his Ways of Worldmaking,
reading it and discussing these issues with him has been of great
value at a number of stages.
I am also grateful to Jeremy Mynott for encouragement and
advice in his capacity as editor.
* To study 'The Appraisal of Scientific Theories: Historical versus Formal
Methodological Approaches'; Agreement No. SOC78-04276.
Cambridge
https://doi.org/10.1017/CBO9780511625398.002
Brains in a vat
An ant is crawling on a patch of sand. As it crawls, it traces a

line in the sand. By pure chance the line that it traces curves and
recrosses itself in such a way that it ends up looking like a rec-
ognizable caricature of Winston Churchill. Has the ant traced a
picture of Winston Churchill, a picture that depicts Churchill?
Most people would say, on a little reflection, that it has not.
The ant, after all, has never seen Churchill, or even a picture of
Churchill, and it had no intention of depicting Churchill. It sim-
ply traced a line (and even that was unintentional), a line that
we can 'see as' a picture of Churchill.
We can express this by saying that the line is not 'in itself a
representation1 of anything rather than anything else. Similarity
(of a certain very complicated sort) to the features of Winston
Churchill is not sufficient to make something represent or refer
to Churchill. Nor is it necessary: in our community the printed
shape 'Winston Churchill', the spoken words 'Winston Chur-
chill', and many other things are used to represent Churchill
(though not pictorially), while not having the sort of similarity
1
In this book the terms 'representation' and 'reference' always refer to a
relation between a word (or other sort of sign, symbol, or
representation) and something that actually exists (i.e. not just an 'object
of thought'). There is a sense of 'refer' in which I can 'refer' to what does
not exist; this is not the sense in which 'refer' is used here. An older
word for what I call 'representation' or 'reference' is denotation.
Secondly, I follow the custom of modern logicians and use 'exist' to
mean 'exist in the past, present, or future'. Thus Winston Churchill
'exists', and we can 'refer to' or 'represent' Winston Churchill, even
though he is no longer alive.
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
2 Brains in a vat
to Churchill that a picture - even a line drawing- has. If simi-

larity is not necessary or sufficient to make something represent
something else, how can anything be necessary or sufficient for
this purpose? How on earth can one thing represent (or 'stand
for', etc.) a different thing?
The answer may seem easy. Suppose the ant had seen Winston
Churchill, and suppose that it had the intelligence and skill to
draw a picture of him. Suppose it produced the caricature inten-
tionally. Then the line would have represented Churchill.
On the other hand, suppose the line had the shape WINSTON
CHURCHILL. And suppose this was just accident (ignoring the
improbability involved). Then the 'printed shape' WINSTON
CHURCHILL would not have represented Churchill, although
that printed shape does represent Churchill when it occurs in
almost any book today.
So it may seem that what is necessary for representation, or
what is mainly necessary for representation, is intention.
But to have the intention that anything, even private lan-
guage (even the words 'Winston Churchill' spoken in my mind
and not out loud), should represent Churchill, I must have been
able to think about Churchill in the first place. If lines in the
sand, noises, etc., cannot 'in themselves' represent anything, then
how is it that thought forms can 'in themselves' represent any-
thing? Or can they? How can thought reach out and 'grasp'
what is external?
Some philosophers have, in the past, leaped from this sort of
consideration to what they take to be a proof that the mind is
essentially non-physical in nature. The argument is simple; what
we said about the ant's curve applies to any physical object. No
physical object can, in itself, refer to one thing rather than to
another; nevertheless, thoughts in the mind obviously do suc-
ceed in referring to one thing rather than another. So thoughts
(and hence the mind) are of an essentially different nature than
physical objects. Thoughts have the characteristic of intention-
ality - they can refer to something else; nothing physical has
'intentionality', save as that intentionality is derivative from
some employment of that physical thing by a mind. Or so it is
claimed. This is too quick; just postulating mysterious powers of
mind solves nothing. But the problem is very real. How is inten-
tionality, reference, possible?
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat
Magical theories of reference

We saw that the ant's 'picture' has no necessary connection with
Winston Churchill. The mere fact that the 'picture' bears a
'resemblance' to Churchill does not make it into a real picture,
nor does it make it a representation of Churchill. Unless the ant
is an intelligent ant (which it isn't) and knows about Churchill
(which it doesn't), the curve it traced is not a picture or even a
representation of anything. Some primitive people believe that
some representations (in particular, names) have a necessary
connection with their bearers; that to know the 'true name' of
someone or something gives one power over it. This power
comes from the magical connection between the name and the
bearer of the name; once one realizes that a name only has a
contextual, contingent, conventional connection with its bearer,
it is hard to see why knowledge of the name should have any
mystical significance.
What is important to realize is that what goes for physical
pictures also goes for mental images, and for mental representa-
tions in general; mental representations no more have a neces-
sary connection with what they represent than physical represen-
tations do. The contrary supposition is a survival of magical
thinking.
Perhaps the point is easiest to grasp in the case of mental
images, (Perhaps the first philosopher to grasp the enormous sig-
nificance of this point, even if he was not the first to actually
make it, was Wittgenstein.) Suppose there is a planet somewhere
on which human beings have evolved (or been deposited by alien
spacemen, or what have you). Suppose these humans, although
otherwise like us, have never seen trees. Suppose they have never
imagined trees (perhaps vegetable life exists on their planet only
in the form of molds). Suppose one day a picture of a tree is
accidentally dropped on their planet by a spaceship which passes
on without having other contact with them. Imagine them puz-
zling over the picture. What in the world is this? All sorts of
speculations occur to them: a building, a canopy, even an animal
of some kind. But suppose they never come close to the truth.
For us the picture is a representation of a tree. For these
humans the picture only represents a strange object, nature and
function unknown. Suppose one of them has a mental image
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
4 Brains in a vat
which is exactly like one of my mental images of a tree as a result

of having seen the picture. His mental image is not a represen-
tation of a tree. It is only a representation of the strange object
(whatever it is) that the mysterious picture represents.
Still, someone might argue that the mental image is in fact a
representation of a tree, if only because the picture which caused
this mental image was itself a representation of a tree to begin
with. There is a causal chain from actual trees to the mental
image even if it is a very strange one.
But even this causal chain can be imagined absent. Suppose
the 'picture of the tree' that the spaceship dropped was not really
a picture of a tree, but the accidental result of some spilled
paints. Even if it looked exactly like a picture of a tree, it was, in
truth, no more a picture of a tree than the ant's 'caricature' of
Churchill was a picture of Churchill. We can even imagine that
the spaceship which dropped the 'picture' came from a planet
which knew nothing of trees. Then the humans would still have
mental images qualitatively identical with my image of a tree,
but they would not be images which represented a tree any more
than anything else.
The same thing is true of words. A discourse on paper might
seem to be a perfect description of trees, but if it was produced
by monkeys randomly hitting keys on a typewriter for millions
of years, then the words do not refer to anything. If there were
a person who memorized those words and said them in his mind
without understanding them, then they would not refer to any-
thing when thought in the mind, either.
Imagine the person who is saying those words in his mind has
been hypnotized. Suppose the words are in Japanese, and the per-
son has been told that he understands Japanese. Suppose that as
he thinks those words he has a 'feeling of understanding'.
(Although if someone broke into his train of thought and asked
him what the words he was thinking meant, he would discover
he couldn't say.) Perhaps the illusion would be so perfect that
the person could even fool a Japanese telepath! But if he
couldn't use the words in the right contexts, answer questions
about what he 'thought', etc., then he didn't understand them.
By combining these science fiction stories I have been telling,
we can contrive a case in which someone thinks words which are
in fact a description of trees in some language and simultane-
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 5
ously has appropriate mental images, but neither understands

the words nor knows what a tree is. We can even imagine that
the mental images were caused by paint-spills (although the per-
son has been hypnotized to think that they are images of some-
thing appropriate to his thought - only, if he were asked, he
wouldn't be able to say of what). And we can imagine that the
language the person is thinking in is one neither the hypnotist
nor the person hypnotized has ever heard of - perhaps it is just
coincidence that these 'nonsense sentences', as the hypnotist sup-
poses them to be, are a description of trees in Japanese. In short,
everything passing before the person's mind might be qualita-
tively identical with what was passing through the mind of a
Japanese speaker who was really thinking about trees — but
none of it would refer to trees.
All of this is really impossible, of course, in the way that it is
really impossible that monkeys should by chance type out a copy
of Hamlet, That is to say that the probabilities against it are so
high as to mean it will never really happen (we think). But is is
not logically impossible, or even physically impossible. It could
happen (compatibly with physical law and, perhaps, compatibly
with actual conditions in the universe, if there are lots of intelli-
gent beings on other planets). And if it did happen, it would be
a striking demonstration of an important conceptual truth; that
even a large and complex system of representations, both verbal
and visual, still does not have an intrinsic, built-in, magical con-
nection with what it represents — a connection independent of
how it was caused and what the dispositions of the speaker or
thinker are. And this is true whether the system of representa-
tions (words and images, in the case of the example) is physically
realized — the words are written or spoken, and the pictures are
physical pictures - or only realized in the mind. Thought words
and mental pictures do not intrinsically represent what they are
about.
The case of the brains in a vat

Here is a science fiction possibility discussed by philosophers:
imagine that a human being (you can imagine this to be yourself)
has been subjected to an operation by an evil scientist. The per-
son's brain (your brain) has been removed from the body and
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
6 Brains in a vat
placed in a vat of nutrients which keeps the brain alive. The

nerve endings have been connected to a super-scientific com-
puter which causes the person whose brain it is to have the illu-
sion that everything is perfectly normal. There seem to be peo-
ple, objects, the sky, etc; but really all the person (you) is
experiencing is the result of electronic impulses travelling from
the computer to the nerve endings. The computer is so clever
that if the person tries to raise his hand, the feedback from the
computer will cause him to 'see' and 'feel' the hand being raised.
Moreover, by varying the program, the evil scientist can cause
the victim to 'experience' (or hallucinate) any situation or envi-
ronment the evil scientist wishes. He can also obliterate the
memory of the brain operation, so that the victim will seem to
himself to have always been in this environment. It can even
seem to the victim that he is sitting and reading these very words
about the amusing but quite absurd supposition that there is an
evil scientist who removes people's brains from their bodies and
places them in a vat of nutrients which keep the brains alive. The
nerve endings are supposed to be connected to a super-scientific
computer which causes the person whose brain it is to have the
illusion that. . .
When this sort of possibility is mentioned in a lecture on the
Theory of Knowledge, the purpose, of course, is to raise the clas-
sical problem of scepticism with respect to the external world in
a modern way. {How do you know you aren't in this predica-
ment?) But this predicament is also a useful device for raising
issues about the mind/world relationship.
Instead of having just one brain in a vat, we could imagine
that all human beings (perhaps all sentient beings) are brains in
a vat (or nervous systems in a vat in case some beings with just
a minimal nervous system already count as 'sentient'). Of course,
the evil scientist would have to be outside — or would he? Per-
haps there is no evil scientist, perhaps (though this is absurd) the
universe just happens to consist of automatic machinery tending
a vat full of brains and nervous systems.
This time let us suppose that the automatic machinery is pro-
grammed to give us all a collective hallucination, rather than a
number of separate unrelated hallucinations. Thus, when I seem
to myself to be talking to you, you seem to yourself to be hearing
my words. Of course, it is not the case that my words actually
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 7
reach your ears — for you don't have (real) ears, nor do I have a
real mouth and tongue. Rather, when I produce my words, what
happens is that the efferent impulses travel from my brain to the
computer, which both causes me to 'hear' my own voice uttering
those words and 'feel' my tongue moving, etc., and causes you
to 'hear' my words, 'see' me speaking, etc. In this case, we are,
in a sense, actually in communication. I am not mistaken about
your real existence (only about the existence of your body and
the 'external world', apart from brains). From a certain point of
view, it doesn't even matter that 'the whole world' is a collective
hallucination; for you do, after all, really hear my words when
I speak to you, even if the mechanism isn't what we suppose it
to be. (Of course, if we were two lovers making love, rather than
just two people carrying on a conversation, then the suggestion
that it was just two brains in a vat might be disturbing.)
I want now to ask a question which will seem very silly and
obvious (at least to some people, including some very sophisti-
cated philosophers), but which will take us to real philosophical
depths rather quickly. Suppose this whole story were actually
true. Could we, if we were brains in a vat in this way, say or
think that we were?
I am going to argue that the answer is 'No, we couldn't.' In
fact, I am going to argue that the supposition that we are
actually brains in a vat, although it violates no physical law, and
is perfectly consistent with everything we have experienced, can-
not possibly be true. It cannot possibly be true, because it is, in
a certain way, self-refuting.
The argument I am going to present is an unusual one, and it
took me several years to convince myself that it is really right.
But it is a correct argument. What makes it seem so strange is
that it is connected with some of the very deepest issues in phi-
losophy. (It first occurred to me when I was thinking about a
theorem in modern logic, the 'Skolem-Lowenheim Theorem',
and I suddenly saw a connection between this theorem and some
arguments in Wittgenstein's Philosophical Investigations.)
A 'self-refuting supposition' is one whose truth implies its own
falsity. For example, consider the thesis that all general state-
ments are false. This is a general statement. So if it is true, then
it must be false. Hence, it is false. Sometimes a thesis is called
'self-refuting' if it is the supposition that the thesis is entertained
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
8 Brains in a vat
or enunciated that implies its falsity. For example, 'I do not exist'
is self-refuting if thought by me (for any 'me9). So one can be
certain that one oneself exists, if one thinks about it (as Des-
cartes argued).
What I shall show is that the supposition that we are brains in
a vat has just this property. If we can consider whether it is true
or false, then it is not true (I shall show). Hence it is not true.
Before I give the argument, let us consider why it seems so
strange that such an argument can be given (at least to philoso-
phers who subscribe to a 'copy' conception of truth). We con-
ceded that it is compatible with physical law that there should
be a world in which all sentient beings are brains in a vat. As
philosophers say, there is a 'possible world' in which all sentient
beings are brains in a vat. (This 'possible world' talk makes it
sound as if there is a place where any absurd supposition is true,
which is why it can be very misleading in philosophy.) The
humans in that possible world have exactly the same experiences
that we do. They think the same thoughts we do (at least, the
same words, images, thought-forms, etc., go through their
minds). Yet, I am claiming that there is an argument we can give
that shows we are not brains in a vat. How can there be? And
why couldn't the people in the possible world who really are
brains in a vat give it too?
The answer is going to be (basically) this: although the people
in that possible world can think and 'say' any words we can
think and say, they cannot (I claim) refer to what we can refer
to. In particular, they cannot think or say that they are brains in
a vat {even by thinking 'we are brains in a vat').
Turing's test
Suppose someone succeeds in inventing a computer which can
actually carry on an intelligent conversation with one (on as
many subjects as an intelligent person might). How can one
decide if the computer is 'conscious'?
The British logician Alan Turing proposed the following test:2
let someone carry on a conversation with the computer and a
conversation with a person whom he does not know. If he can-
2
A. M. Turing, 'Computing Machinery and Intelligence', Mind (1950),
reprinted in A. R. Anderson (ed.), Minds and Machines.
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 9
not tell which is the computer and which is the human being,
then (assume the test to be repeated a sufficient number of times
with different interlocutors) the computer is conscious. In short,
a computing machine is conscious if it can pass the 'Turing Test'.
(The conversations are not to be carried on face to face, of
course, since the interlocutor is not to know the visual appear-
ance of either of his two conversational partners. Nor is voice to
be used, since the mechanical voice might simply sound different
from a human voice. Imagine, rather, that the conversations are
all carried on via electric typewriter. The interlocutor types in
his statements, questions, etc., and the two partners — the
machine and the person — respond via the electric keyboard.
Also, the machine may lie — asked 'Are you a machine', it might
reply, 'No, I'm an assistant in the lab here.')
The idea that this test is really a definitive test of consciousness
has been criticized by a number of authors (who are by no means
hostile in principle to the idea that a machine might be con-
scious). But this is not our topic at this time. I wish to use the
general idea of the Turing test, the general idea of a dialogic test
of competence, for a different purpose, the purpose of exploring
the notion of reference.
Imagine a situation in which the problem is not to determine
if the partner is really a person or a machine, but is rather to
determine if the partner uses the words to refer as we do. The
obvious test is, again, to carry on a conversation, and, if no
problems arise, if the partner 'passes' in the sense of being indis-
tinguishable from someone who is certified in advance to be
speaking the same language, referring to the usual sorts of
objects, etc., to conclude that the partner does refer to objects as
we do. When the purpose of the Turing test is as just described,
that is, to determine the existence of (shared) reference, I shall
refer to the test as the Turing Test for Reference. And, just as
philosophers have discussed the question whether the original
Turing test is a definitive test for consciousness, i.e. the question
of whether a machine which 'passes' the test not just once but
regularly is necessarily conscious, so, in the same way, I wish to
discuss the question of whether the Turing Test for Reference
just suggested is a definitive test for shared reference.
The answer will turn out to be 'No'. The Turing Test for Ref-
erence is not definitive. It is certainly an excellent test in practice;
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
10 Brains in a vat
but it is not logically impossible (though it is certainly highly

improbable) that someone could pass the Turing Test for Refer-
ence and not be referring to anything. It follows from this, as we
shall see, that we can extend our observation that words (and
whole texts and discourses) do not have a necessary connection
to their referents. Even if we consider not words by themselves
but rules deciding what words may appropriately be produced
in certain contexts - even if we consider, in computer jargon,
programs for using words — unless those programs themselves
refer to something extra-linguistic there is still no determinate
reference that those words possess. This will be a crucial step in
the process of reaching the conclusion that the Brain-in-a-Vat
Worlders cannot refer to anything external at all (and hence can-
not say that they are Brain-in-a-Vat Worlders).
Suppose, for example, that I am in the Turing situation (play-
ing the 'Imitation Game', in Turing's terminology) and my part-
ner is actually a machine. Suppose this machine is able to win
the game ('passes' the test). Imagine the machine to be pro-
grammed to produce beautiful responses in English to state-
ments, questions, remarks, etc. in English, but that it has no
sense organs (other than the hookup to my electric typewriter),
and no motor organs (other than the electric typewriter). (As far
as I can make out, Turing does not assume that the possession
of either sense organs or motor organs is necessary for con-
sciousness or intelligence.) Assume that not only does the
machine lack electronic eyes and ears, etc., but that there are no
provisions in the machine's program, the program for playing
the Imitation Game, for incorporating inputs from such sense
organs, or for controlling a body. What should we say about
such a machine?
To me, it seems evident that we cannot and should not attrib-
ute reference to such a device. It is true that the machine can
discourse beautifully about, say, the scenery in New England.
But it could not recognize an apple tree or an apple, a mountain
or a cow, a field or a steeple, if it were in front of one.
What we have is a device for producing sentences in response
to sentences. But none of these sentences is at all connected to
the real world. If one coupled two of these machines and let
them play the Imitation Game with each other, then they would
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 11
go on 'fooling' each other forever, even if the rest of the world

disappeared! There is no more reason to regard the machine's
talk of apples as referring to real world apples than there is to
regard the ant's 'drawing' as referring to Winston Churchill.
What produces the illusion of reference, meaning, intelligence,
etc., here is the fact that there is a convention of representation
which we have under which the machine's discourse refers to
apples, steeples, New England, etc. Similarly, there is the illusion
that the ant has caricatured Churchill, for the same reason. But
we are able to perceive, handle, deal with apples and fields. Our
talk of apples and fields is intimately connected with our non-
verbal transactions with apples and fields. There are 'language
entry rules' which take us from experiences of apples to such
utterances as 'I see an apple', and 'language exit rules' which
take us from decisions expressed in linguistic form ('I am going
to buy some apples') to actions other than speaking. Lacking
either language entry rules or language exit rules, there is no
reason to regard the conversation of the machine (or of the two
machines, in the case we envisaged of two machines playing the
Imitation Game with each other) as more than syntactic play.
Syntactic play that resembles intelligent discourse, to be sure;
but only as (and no more than) the ant's curve resembles a biting
caricature.
In the case of the ant, we could have argued that the ant would
have drawn the same curve even if Winston Churchill had never
existed. In the case of the machine, we cannot quite make the
parallel argument; if apples, trees, steeples and fields had not
existed, then, presumably, the programmers would not have
produced that same program. Although the machine does not
perceive apples, fields, or steeples, its creator-designers did.
There is some causal connection between the machine and the
real world apples, etc., via the perceptual experience and knowl-
edge of the creator—designers. But such a weak connection can
hardly suffice for reference. Not only is it logically possible,
though fantastically improbable, that the same machine could
have existed even if apples, fields, and steeples had not existed;
more important, the machine is utterly insensitive to the contin-
ued existence of apples, fields, steeples, etc. Even if all these
things ceased to exist, the machine would still discourse just as
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
12 Brains in a vat
happily in the same way. That is why the machine cannot be

regarded as referring at all.
The point that is relevant for our discussion is that there is
nothing in Turing's Test to rule out a machine which is pro-
grammed to do nothing but play the Imitation Game, and that a
machine which can do nothing but play the Imitation Game is
clearly not referring any more than a record player is.
Brains in a vat (again)

Let us compare the hypothetical 'brains in a vat' with the
machines just described. There are obviously important differ-
ences. The brains in a vat do not have sense organs, but they do
have provision for sense organs; that is, there are afferent nerve
endings, there are inputs from these afferent nerve endings, and
these inputs figure in the 'program' of the brains in the vat just
as they do in the program of our brains. The brains in a vat are
brains; moreover, they are functioning brains, and they function
by the same rules as brains do in the actual world. For these
reasons, it would seem absurd to deny consciousness or intelli-
gence to them. But the fact that they are conscious and intelligent
does not mean that their words refer to what our words refer.
The question we are interested in is this: do their verbalizations
containing, say, the word 'tree' actually refer to trees? More gen-
erally: can they refer to external objects at all? (As opposed to,
for example, objects in the image produced by the automatic
machinery.)
To fix our ideas, let us specify that the automatic machinery is
supposed to have come into existence by some kind of cosmic
chance or coincidence (or, perhaps, to have always existed). In
this hypothetical world, the automatic machinery itself is sup-
posed to have no intelligent creator-designers. In fact, as we said
at the beginning of this chapter, we may imagine that all sentient
beings (however minimal their sentience) are inside the vat.
This assumption does not help. For there is no connection
between the word 'tree' as used by these brains and actual trees.
They would still use the word 'tree' just as they do, think just the
thoughts they do, have just the images they have, even if there
were no actual trees. Their images, words, etc., are qualitatively
identical with images, words, etc., which do represent trees in
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 13
our world; but we have already seen (the ant again!) that quali-
tative similarity to something which represents an object (Win-
ston Churchill or a tree) does not make a thing a representation
all by itself. In short, the brains in a vat are not thinking about
real trees when they think 'there is a tree in front of me' because
there is nothing by virtue of which their thought 'tree' represents
actual trees.
If this seems hasty, reflect on the following: we have seen that
the words do not necessarily refer to trees even if they are
arranged in a sequence which is identical with a discourse which
(were it to occur in one of our minds) would unquestionably be
about trees in the actual world. Nor does the 'program', in the
sense of the rules, practices, dispositions of the brains to verbal
behavior, necessarily refer to trees or bring about reference to
trees through the connections it establishes between words and
words, or linguistic cues and linguistic responses. If these brains
think about, refer to, represent trees (real trees, outside the vat),
then it must be because of the way the 'program' connects the
system of language to non-verbal input and outputs. There are
indeed such non-verbal inputs and outputs in the Brain-in-a-Vat
world (those efferent and afferent nerve endings again!), but we
also saw that the 'sense-data' produced by the automatic
machinery do not represent trees (or anything external) even
when they resemble our tree-images exactly. Just as a splash of
paint might resemble a tree picture without being a tree picture,
so, we saw, a 'sense datum' might be qualitatively identical with
an 'image of a tree' without being an image of a tree. How can
the fact that, in the case of the brains in a vat, the language is
connected by the program with sensory inputs which do not
intrinsically or extrinsically represent trees (or anything exter-
nal) possibly bring it about that the whole system of representa-
tions, the language-in-use, does refer to or represent trees or any-
thing external?
The answer is that it cannot. The whole system of sense-data,
motor signals to the efferent endings, and verbally or concep-
tually mediated thought connected by 'language entry rules' to
the sense-data (or whatever) as inputs and by 'language exit
rules' to the motor signals as outputs, has no more connection to
trees than the ant's curve has to Winston Churchill. Once we see
that the qualitative similarity (amounting, if you like, to quali-
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
14 Brains in a vat
tative identity) between the thoughts of the brains in a vat and

the thoughts of someone in the actual world by no means implies
sameness of reference, it is not hard to see that there is no basis
at all for regarding the brain in a vat as referring to external
things.
The premisses of the argument

I have now given the argument promised to show that the brains
in a vat cannot think or say that they are brains in a vat. It
remains only to make it explicit and to examine its structure.
By what was just said, when the brain in a vat (in the world
where every sentient being is and always was a brain in a vat)
thinks 'There is a tree in front of me', his thought does not refer
to actual trees. On some theories that we shall discuss it might
refer to trees in the image, or to the electronic impulses that
cause tree experiences, or to the features of the program that are
responsible for those electronic impulses. These theories are not
ruled out by what was just said, for there is a close causal con-
nection between the use of the word 'tree' in vat-English and the
presence of trees in the image, the presence of electronic impulses
of a certain kind, and the presence of certain features in the
machine's program. On these theories the brain is right, not
wrong in thinking 'There is a tree in front of me.' Given what
'tree' refers to in vat-English and what 'in front of refers
to, assuming one of these theories is correct, then the truth-
conditions for 'There is a tree in front of me' when it occurs in
vat-English are simply that a tree in the image be 'in front of the
'me' in question - in the image - or, perhaps, that the kind of
electronic impulse that normally produces this experience be
coming from the automatic machinery, or, perhaps, that the fea-
ture of the machinery that is supposed to produce the 'tree in
front of one' experience be operating. And these truth-
conditions are certainly fulfilled.
By the same argument, 'vat' refers to vats in the image in vat-
English, or something related (electronic impulses or program
features), but certainly not to real vats, since the use of 'vat' in
vat-English has no causal connection to real vats (apart from the
connection that the brains in a vat wouldn't be able to use the
word 'vat', if it were not for the presence of one particular vat —
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 15
the vat they are in; but this connection obtains between the use
of every word in vat-English and that one particular vat; it is not
a special connection between the use of the particular word 'vat'
and vats). Similarly, 'nutrient fluid' refers to a liquid in the image
in vat-English, or something related (electronic impulses or pro-
gram features). It follows that if their 'possible world' is really
the actual one, and we are really the brains in a vat, then what
we now mean by 'we are brains in a vat' is that we are brains in
a vat in the image or something of that kind (if we mean any-
thing at all). But part of the hypothesis that we are brains in a
vat is that we aren't brains in a vat in the image (i.e. what we are
'hallucinating' isn't that we are brains in a vat). So, if we are
brains in a vat, then the sentence 'We are brains in a vat' says
something false (if it says anything). In short, if we are brains in
a vat, then 'We are brains in a vat' is false. So it is (necessarily)
false.
The supposition that such a possibility makes sense arises
from a combination of two errors: (1) taking physical possibility
too seriously; and (2) unconsciously operating with a magical
theory of reference, a theory on which certain mental represen-
tations necessarily refer to certain external things and kinds of
things.
There is a 'physically possible world' in which we are brains
in a vat - what does this mean except that there is a description
of such a state of affairs which is compatible with the laws of
physics? Just as there is a tendency in our culture (and has been
since the seventeenth century) to take physics as our metaphys-
ics, that is, to view the exact sciences as the long-sought descrip-
tion of the 'true and ultimate furniture of the universe', so there
is, as an immediate consequence, a tendency to take 'physical
possibility' as the very touchstone of what might really actually
be the case. Truth is physical truth; possibility physical possibil-
ity; and necessity physical necessity, on such a view. But we have
just seen, if only in the case of a very contrived example so far,
that this view is wrong. The existence of a 'physically possible
world' in which we are brains in a vat (and always were and will
be) does not mean that we might really, actually, possibly be
brains in a vat. What rules out this possibility is not physics but
philosophy.
Some philosophers, eager both to assert and minimize the
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
16 Brains in a vat
claims of their profession at the same time (the typical state of

mind of Anglo-American philosophy in the twentieth century),
would say: 'Sure. You have shown that some things that seem to
be physical possibilities are really conceptual impossibilities.
What's so surprising about that?'
Well, to be sure, my argument can be described as a 'concep-
tual' one. But to describe philosophical activity as the search for
'conceptual' truths makes it all sound like inquiry about the
meaning of words. And that is not at all what we have been
engaging in.
What we have been doing is considering the preconditions for
thinking about, representing, referring to, etc. We have investi-
gated these preconditions not by investigating the meaning of
these words and phrases (as a linguist might, for example) but
by reasoning a priori. Not in the old 'absolute' sense (since we
don't claim that magical theories of reference are a priori
wrong), but in the sense of inquiring into what is reasonably
possible assuming certain general premisses, or making certain
very broad theoretical assumptions. Such a procedure is neither
'empirical' nor quite 'a priori', but has elements of both ways of
investigating. In spite of the fallibility of my procedure, and its
dependence upon assumptions which might be described as
'empirical' (e.g. the assumption that the mind has no access to
external things or properties apart from that provided by the
senses), my procedure has a close relation to what Kant called a
'transcendental' investigation; for it is an investigation, I repeat,
of the preconditions of reference and hence of thought - precon-
ditions built in to the nature of our minds themselves, though
not (as Kant hoped) wholly independent of empirical assump-
tions.
One of the premisses of the argument is obvious: that magical
theories of reference are wrong, wrong for mental representa-
tions and not only for physical ones. The other premiss is that
one cannot refer to certain kinds of things, e.g. trees, if one has
no causal interaction at all with them,3 or with things in terms
3
If the Brains in a Vat will have causal connection with, say, trees in the
future, then perhaps they can now refer to trees by the description 'the
things I will refer to as "trees" at such-and-such a future time'. But we
are to imagine a case in which the Brains in a Vat never get out of the
vat, and hence never get into causal connection with trees, etc.
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 17
of which they can be described. But why should we accept these

premisses? Since these constitute the broad framework within
which I am arguing, it is time to examine them more closely.
The reasons for denying necessary connections between

representations and their referents
I mentioned earlier that some philosophers (most famously,
Brentano) have ascribed to the mind a power, 'intentionality',
which precisely enables it to refer. Evidently, I have rejected this
as no solution. But what gives me this right? Have I, perhaps,
been too hasty?
These philosophers did not claim that we can think about
external things or properties without using representations at all.
And the argument I gave above comparing visual sense data to
the ant's 'picture' (the argument via the science fiction story
about the 'picture' of a tree that came from a paint-splash and
that gave rise to sense data qualitatively similar to our 'visual
images of trees', but unaccompanied by any concept of a tree)
would be accepted as showing that images do not necessarily
refer. If there are mental representations that necessarily refer (to
external things) they must be of the nature of concepts and not
of the nature of images. But what are concepts*
When we introspect we do not perceive 'concepts' flowing
through our minds as such. Stop the stream of thought when or
where we will, what we catch are words, images, sensations,
feelings. When I speak my thoughts out loud I do not think them
twice. I hear my words as you do. To be sure it feels different to
me when I utter words that I believe and when I utter words I
do not believe (but sometimes, when I am nervous, or in front of
a hostile audience, it feels as if I am lying when I know I am
telling the truth); and it feels different when I utter words I
understand and when I utter words I do not understand. But I
can imagine without difficulty someone thinking just these
words (in the sense of saying them in his mind) and having just
the feeling of understanding, asserting, etc., that I do, and real-
izing a minute later (or on being awakened by a hypnotist) that
he did not understand what had just passed through his mind at
all, that he did not even understand the language these words are
in. I don't claim that this is very likely; I simply mean that there
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
18 Brains in a vat
is nothing at all unimaginable about this. And what this shows

is not that concepts are words (or images, sensations, etc.), but
that to attribute a 'concept' or a 'thought' to someone is quite
different from attributing any mental 'presentation', any intro-
spectible entity or event, to him. Concepts are not mental presen-
tations that intrinsically refer to external objects for the very
decisive reason that they are not mental presentations at all.
Concepts are signs used in a certain way; the signs may be public
or private, mental entities or physical entities, but even when the
signs are 'mental' and 'private', the sign itself apart from its use
is not the concept. And signs do not themselves intrinsically
refer.
We can see this by performing a very simple thought experi-
ment. Suppose you are like me and cannot tell an elm tree from
a beech tree. We still say that the reference of 'elm' in my speech
is the same as the reference of 'elm' in anyone else's, viz. elm
trees, and that the set of all beech trees is the extension of 'beech'
(i.e. the set of things the word 'beech' is truly predicated of) both
in your speech and my speech. Is it really credible that the differ-
ence between what 'elm' refers to and what 'beech' refers to is
brought about by a difference in our concepts* My concept of
an elm tree is exactly the same as my concept of a beech tree (I
blush to confess). (This shows that the determination of refer-
ence is social and not individual, by the way; you and I both
defer to experts who can tell elms from beeches.) If someone
heroically attempts to maintain that the difference between the
reference of 'elm' and the reference of 'beech' in my speech is
explained by a difference in my psychological state, then let him
imagine a Twin Earth where the words are switched. Twin Earth
is very much like Earth; in fact, apart from the fact that 'elm'
and 'beech' are interchanged, the reader can suppose Twin Earth
is exactly like Earth. Suppose I have a Doppelganger on Twin
Earth who is molecule for molecule identical with me (in the
sense in which two neckties can be 'identical'). If you are a dual-
ist, then suppose my Doppelganger thinks the same verbalized
thoughts I do, has the same sense data, the same dispositions,
etc. It is absurd to think his psychological state is one bit differ-
ent from mine: yet his word 'elm' represents beeches, and my
word 'elm' represents elms. (Similarly, if the 'water' on Twin
Earth is a different liquid - say, XYZ and not H2O - then 'water'
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 19
represents a different liquid when used on Twin Earth and when

used on Earth, etc.) Contrary to a doctrine that has been with us
since the seventeenth century, meanings just aren't in the head.
We have seen that possessing a concept is not a matter of pos-
sessing images (say, of trees — or even images, 'visual' or 'acous-
tic', of sentences, or whole discourses, for that matter) since one
could possess any system of images you please and not possess
the ability to use the sentences in situationally appropriate ways
(considering both linguistic factors - what has been said
before — and non-linguistic factors as determining 'situational
appropriateness'). A man may have all the images you please,
and still be completely at a loss when one says to him 'point to
a tree', even if a lot of trees are present. He may even have the
image of what he is supposed to do, and still not know what he
is supposed to do. For the image, if not accompanied by the
ability to act in a certain way, is just a picture, and acting in
accordance with a picture is itself an ability that one may or may
not have. (The man might picture himself pointing to a tree, but
just for the sake of contemplating something logically possible;
himself pointing to a tree after someone has produced the - to
him meaningless — sequence of sounds 'please point to a tree'.)
He would still not know that he was supposed to point to a tree,
and he would still not understand 'point to a tree'.
I have considered the ability to use certain sentences to be the
criterion for possessing a full-blown concept, but this could eas-
ily be liberalized. We could allow symbolism consisting of ele-
ments which are not words in a natural language, for example,
and we could allow such mental phenomena as images and other
types of internal events. What is essential is that these should
have the same complexity, ability to be combined with each
other, etc., as sentences in a natural language. For, although a
particular presentation — say, a blue flash — might serve a partic-
ular mathematician as the inner expression of the whole proof
of the Prime Number Theorem, still there would be no tempta-
tion to say this (and it would be false to say this) if that mathe-
matician could not unpack his 'blue flash' into separate steps and
logical connections. But, no matter what sort of inner phenom-
ena we allow as possible expressions of thought, arguments
exactly similar to the foregoing will show that it is not the phe-
nomena themselves that constitute understanding, but rather the
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
20 Brains in a vat
ability of the thinker to employ these phenomena, to produce

the right phenomena in the right circumstances.
The foregoing is a very abbreviated version of Wittgenstein's
argument in Philosophical Investigations. If it is correct, then the
attempt to understand thought by what is called 'phenomeno-
logicaP investigation is fundamentally misguided; for what the
phenomenologists fail to see is that what they are describing is
the inner expression of thought, but that the understanding of
that expression - one's understanding of one's own thoughts -
is not an occurrence but an ability. Our example of a man pre-
tending to think in Japanese (and deceiving a Japanese telepath)
already shows the futility of a phenomenological approach to
the problem of understanding. For even if there is some intros-
pectible quality which is present when and only when one really
understands (this seems false on introspection, in fact), still that
quality is only correlated with understanding, and it is still pos-
sible that the man fooling the Japanese telepath have that quality
too and still not understand a word of Japanese.
On the other hand, consider the perfectly possible man who
does not have any 'interior monologue' at all. He speaks per-
fectly good English, and if asked what his opinions are on a
given subject, he will give them at length. But he never thinks (in
words, images, etc.) when he is not speaking out loud; nor does
anything 'go through his head', except that (of course) he hears
his own voice speaking, and has the usual sense impressions
from his surroundings, plus a general 'feeling of understanding'.
(Perhaps he is in the habit of talking to himself.) When he types
a letter or goes to the store, etc., he is not having an internal
'stream of thought'; but his actions are intelligent and purpose-
ful, and if anyone walks up and asks him 'What are you doing?'
he will give perfectly coherent replies.
This man seems perfectly imaginable. No one would hesitate
to say that he was conscious, disliked rock and roll (if he fre-
quently expressed a strong aversion to rock and roll), etc., just
because he did not think conscious thoughts except when speak-
ing out loud.
What follows from all this is that (a) no set of mental events -
images or more 'abstract' mental happenings and qualities —
constitutes understanding; and (b) no set of mental events is
necessary for understanding. In particular, concepts cannot be
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
Brains in a vat 21
identical with mental objects of any kind. For, assuming that

by a mental object we mean something introspectible, we have
just seen that whatever it is, it may be absent in a man who
does understand the appropriate word (and hence has the full
blown concept), and present in a man who does not have the
concept at all.
Coming back now to our criticism of magical theories of ref-
erence (a topic which also concerned Wittgenstein), we see that,
on the one hand, those 'mental objects' we can introspectively
detect — words, images, feelings, etc. — do not intrinsically refer
any more than the ant's picture does (and for the same reasons),
while the attempts to postulate special mental objects, 'con-
cepts', which do have a necessary connection with their refer-
ents, and which only trained phenomenologists can detect, com-
mit a logical blunder; for concepts are (at least in part) abilities
and not occurrences. The doctrine that there are mental presen-
tations which necessarily refer to external things is not only bad
natural science; it is also bad phenomenology and conceptual
confusion.
Cambridge
https://doi.org/10.1017/CBO9780511625398.003
A problem about reference
Why is it surprising that the Brain in a Vat hypothesis turns out

to be incoherent? The reason is that we are inclined to think that
what goes on inside our heads must determine what we mean
and what our words refer to. But it is not hard to see that this is
wrong. Ordinary indexical words, such as I, this, here, now, are
a counterexample of a trivial sort. I may be in the same mental
state as Henry when I think 'I am late to work' (imagine, if you
like, that Henry and I are identical twins) and yet the token of
the word T that occurs in my thought refers to me and the token
of the word T that occurs in Henry's thought refers to Henry. I
may be in the same mental state1 when I think 'I am late to work'
on Tuesday and when I think 'I am late to work' on Wednesday;
but the time to which my tensed verb 'am' refers is different in
the two cases. The case of natural kind terms is a more subtle
example of the same point.
Suppose, to spell out the case mentioned in the previous chap-
ter, that there are English speakers on Twin Earth (by a kind of
1
At least I may be in the 'same mental state' in the sense that the
parameters involved in the psychological process that results in my
thinking the thought may have the same values. My global mental state
is, to be sure, different since on Tuesday I believe 'this is Tuesday'
and on Wednesday I don't; but a theory that says the meaning of the
words changes whenever my global mental state changes would
not allow any words to ever have the same meaning, and would thus
amount to an abandonment of the very notion of word meaning.
Moreover, we could construct a Twin Earth story in which I and my
Doppelganger are in the same global mental state, and the reference
of T and 'now' is still different (the calendar on Twin Earth is not
synchronized with ours).
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
A problem about reference 23
miraculous accident they just evolved resembling us and speak-

ing a language which is, apart from a difference I am about to
mention, identical with English as it was a couple of hundred
years ago). I will assume these people do not yet have a knowl-
edge of Daltonian or post-Daltonian chemistry. So, in particular,
they don't have available such notions as 'H 2 O'. Suppose, now,
that the rivers and lakes on Twin Earth are filled with a liquid
that superficially resembles water, but which is not H2O. Then
the word 'water' as used on Twin Earth refers not to water but
to this other liquid (say, XYZ). Yet there is no relevant differ-
ence in the mental state of Twin Earth speakers and speakers on
Earth (in, say, 1750) which could account for this difference in
reference. The reference is different because the stuff is differ-
ent.2 The mental state by itself, in isolation from the whole situ-
ation, does not fix the reference.
Some philosophers have objected to this example, however.
These philosophers suggest that one should say, if such a planet
is ever discovered, that 'There are two kinds of water', and not
that our word 'water' does not refer to the Twin Earth liquid. If
we ever find lakes and rivers full of a liquid other than H2O that
superficially resembles water, then we will have falsified the
statement that all water is H2O, according to these critics.
It is easy to modify the example so as to avoid this argument.
First of all, the liquid on Twin Earth need not be that similar to
water. Suppose it is actually a mixture of 20% grain alcohol and
80% water, but the body chemistry of the Twin Earth people is
such that they do not get intoxicated or even taste the difference
between such a mixture and H2O. Such a liquid would be differ-
ent from water in many ways; yet a typical speaker might be
unacquainted with these differences, and thus be in exactly the
same mental state as a typical speaker in 1750 on Earth. Of
course Twin Earth 'water' tastes different from Earth water to
us; but it does not taste different to them. And it behaves differ-
ently when you boil it; but must an English speaker have noticed
exactly when water boils and exactly what takes place in order
to associate a fairly standard conceptual content with the word
'water'?
2
See 'The Meaning of "Meaning" ' in my Mind, Language, and Reality
{Philosophical Papers, vol. 1), Cambridge University Press, 1975, for an
extended discussion of this point.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
24 A problem about reference
It may be objected that there might well be experts on Twin

Earth who do know things about 'water' (for instance, that it is
a mixture of two liquids) that we do not know about water (did
not believe about water in 1750, because they aren't true of
water), and hence that the collective mental state of Twin Earth
English speakers is different from the collective mental state of
Earth English speakers (in 1750). One might concede that the
reference of a person's term isn't fixed by his individual mental
state, but insist that the total mental state of all the members of
the language community fixes the reference of the term.
One difficulty with this is that it might not have been the case
that people on Earth or Twin Earth had developed that much
chemistry in 1750. If the term had the same meaning and refer-
ence on Earth prior to the development of chemistry that it does
today (in ordinary use), and if the term had the same meaning
and reference on Twin Earth prior to the development of the
corresponding knowledge that it does today, then we can go
back to this earlier time when the collective mental states of the
two communities were the same in all respects relevant to fixing
the extension of 'water', and argue that the extension was differ-
ent then (as it is now) and so the collective mental state does not
fix the extension. Should we then say the reference changed
when chemistry was developed? That the term used to refer to
both kinds of water (in spite of the difference in taste to us!), and
only refers to different kinds after chemistry is developed?
If we say that the reference of their terms or of our terms
changed when they or we developed chemistry (to the extent of
being able to distill liquids, tell that water plus alcohol is a mix-
ture, etc.) then we will have to say that almost every scientific
discovery changes the reference of our terms. We did not dis-
cover that water (in the pre-scientific sense) was H2O on such a
view; rather we stipulated it. To me this seems clearly wrong.
What we meant by water all along was whatever had the same
nature as the local stuff picked out by that term; and we discov-
ered that water in that sense was H2O; what the people on Twin
Earth meant by 'water' all along was the stuff in their environ-
ment picked out by that term, and their experts discovered that
'water' in that sense is a mixture of two liquids.
If we agree that 'water' does not change meaning (in either
language) when experts make such discoveries as 'water is H 2 O'
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
or 'water is a mixture of two liquids', or does not change its

ordinary meaning and reference (of course it may develop more
technical uses as a result of such discoveries), and that 'water' in
its ordinary Earth meaning and reference does not include mix-
tures of alcohol and water, then we must say that expert knowl-
edge is not what accounts for the difference in the meaning of
the word 'water' on Earth and on Twin Earth. Nor does it
account for the reference: for we could consider yet another
Twin Earth where water was a different mixture and the expert
knowledge was the same (rather scanty) expert knowledge as on
the first Twin Earth. Or, as just indicated, we could simply imag-
ine that experts on Earth and on Twin Earth did not yet exist.
The word 'water' would still refer to different stuff even if the
collective mental state in the two communities were the same.
What goes on inside people's heads does not fix the reference of
their terms. In a phrase due to Mill, 'the substance itself com-
pletes the job of fixing the extension of the term.
Once we see that mental state (in either the individualistic or
the collective sense) does not fix reference, then we should not
be surprised that the Brains in a Vat could not succeed in refer-
ring to external objects (even though they have the same mental
states we have), and hence could not say or think that they are
Brains in a Vat.
Intentions, extensions, and 'notional worlds'

In order to look at the problem of how the reference of our terms
is fixed, given that it is not fixed simply by our mental states, it
will be convenient to have available some technical terms. In
logic the set of things a term is true of is called the extension of
the term. Thus the extension of the term 'cat' is the set of cats. If
a term has more than one sense, then we pretend the word car-
ries invisible subscripts (so that there are really two words and
not one), e.g. 'rabbit/ - extension: the set of rabbits - 'rabbit2' -
extension: the set of cowards. (Strictly speaking, the extension
of terms in a natural language is always somewhat fuzzy: but we
shall pretend for simplicity that the borderline cases have been
somehow legislated.)
A word like T which refers to different people on different
occasions will have not an extension but an extension-function:
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
that is a function which determines an extension in each context

of use. In the case of the word T the extension-function is rather
simple; it is simply the function/^) whose value for any speaker
x is the set consisting of just x. The argument *xf which ranges
over the relevant parameter used to describe the context (in this
case, the speaker) is referred to in semantics as an index. Indices
are needed for times, for things demonstratively referred to, and
for yet other features of context in a full semantic treatment (but
we shall ignore the details).
The set of things which makes up the extension of 'cat' is dif-
ferent in different possible situations or 'possible worlds'. In a
possible world M in which there are no cats, the extension of
'cat' is the empty set. If my cat Elsa had had offspring, then the
extension of 'cat' would have at least one member it does not
have in the actual world. (We can express this by saying that in
each possible world M in which Elsa had offspring, the extension
of 'cat' includes members it does not include in the actual
world.)
We can indicate the way in which the extension of a term
varies with the possible world M in exactly the way in which we
indicate how the extension of the word T varies with the
speaker: by using a function. We assume a set of abstract objects
called 'possible worlds' to represent the various states of affairs
or possible world histories, and we associate with the term 'cat'
a function f{M) whose value on each possible world M is the set
of possible objects which are cats in the world M. This function,
following Montague and Carnap, I shall refer to as the inten-
sion3 of the word 'cat'. Similarly, the intension of the two-place
predicate 'touches' is the function f(M) whose value on any pos-
sible world M is the set of ordered pairs of possible objects which
touch each other in the world M; the intension of the three-place
predicate 'x is between y and z' is the function whose value in
any possible world is the set of ordered triples (x, y, z) such that
x is between y and z, and so on. The intension of a word like T,
whose extension in any world is context-dependent, will be a
more complicated function having as arguments both the possi-
ble world and the indices representing the context.
3
Montague, R., Formal Philosophy, Yale University, 1974. This use of
'intension' is not the traditional one which I discussed in 'The
Meaning of "Meaning" '.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
What the intension does is to specify how the extension

depends on the possible world. It thus represents what we are
interested in, the extension associated with a term, in a very
complete way, since it says what that extension would have been
in any possible world.
The reason 'intension' (in this sense) cannot be identified with
meaning is that any two terms which are logically equivalent
have the same extension in every possible world, and hence the
same intension, but a theory which cannot distinguish between
terms with the same meaning and terms which are only equiva-
lent in logic and mathematics is inadequate as a theory of mean-
ing. 'Cube' and 'regular polyhedron with six square faces' are
logically equivalent predicates. So the intension of these two
terms is the same, namely the function whose value in any pos-
sible world is the set of cubes in that world; but there is a differ-
ence in meaning which would be lost if we simply identified the
meaning with this function.
Let me emphasize that possible worlds, sets, and functions are
to be thought of as abstract extra-mental entities in this theory,
and not to be confused with representations or descriptions of
these entities.
Frege thought that the meaning (Sinn) of an expression was
an extra-mental entity or concept which could somehow be
'grasped' by the mind. Such a theory cannot help us with inten-
sions in our sense. In the first place, as just noted, there are dif-
ferences in meaning which are not captured by intension; so the
understanding of a term cannot consist only in associating it
with an intension. More important, if we assume that we have
no 'sixth sense' which enables us to directly perceive extra-men-
tal entities, or to do something analogous to perceiving them
('intuiting' them, perhaps), then 'grasping' an intension, or any
extra-mental entity, must be mediated by representations in
some way. (This also seems clear introspectively, to me at least.)
But the whole problem we are investigating is how representa-
tions can enable us to refer to what is outside the mind. To
assume the notion of 'grasping' an X which is external to the
mind would be to beg the whole question.
If I say of someone that he 'believes there is a glass of water
on the table', then I normally attribute to him the capacity to
refer to water. But, as we have seen, being able to refer to water
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
requires being directly or indirectly linked to actual water

(H2O); the statement 'John believes there is a glass of water in
front of him' is not just a statement about what goes on in John's
head, but is in part a statement about John's environment, and
John's relationship to that environment. If it turns out that John
is a Twin Earth person, then what John believes when he says
'there is a glass of water on the table' is that there is a glass
containing a liquid which in fact consists of water and grain
alcohol on the table.
Husserl introduced a device which is useful when we wish to
talk of what goes on in someone's head without any assumptions
about the existence or nature of actual things referred to by the
thoughts: the device of bracketing.* If we 'bracket' the belief that
we ascribe to John when we say 'John believes that there is a
glass of water on the table' then what we ascribe to John is sim-
ply the mental state of an actual or possible person who believes
that there is a glass of water on the table (in the full 'unbrack-
eted' ordinary sense). Thus, if John on Twin Earth cannot taste
the difference between water and water-cum-grain-alcohol, he
may be in the same mental state as an actual or possible speaker
of Earth English when he says 'there is a glass of water on the
table', notwithstanding the fact that what he refers to as water
would make a reasonable highball. We will say that he has the
bracketed belief that [there is a glass of water on the table]. In
effect, the device of bracketing subtracts entailments from the
ordinary belief locution (all the entailments that refer to the
external world, or to what is external to the thinker's mind).
Daniel Dennett has recently used the locution 'notional world'
in a way related to the way Husserl used bracketing.5 The total-
ity of a thinker's bracketed beliefs constitute the description of
the thinker's notional world, in Dennett's sense. Thus, people on
Twin Earth have roughly the same notional world (and even the
same notional water) that we do; it is just that they live on a
different real planet (and refer to different actual stuff as
'water'). And the Brains in a Vat of the previous chapter could
have had the same notional world we do down to the last detail,
4
Husserl, Ideas; General Introduction to Pure Phenomenology, Allen and
Unwin, 1969. (Originally appeared in 1913).
5
Dennett, D. 'Beyond Belief, in Thought and Object, Andrew Woodfield
(ed.), Oxford University Press, forthcoming.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
if you like; it is just that none of their terms had any external
world reference at all. The traditional theory of meaning
assumed that a thinker's notional world determines the inten-
sions of his terms (and these, together with the fact that a partic-
ular possible world M is the actual one, determine the extensions
of the terms and the truth-values of all the sentences). We have
seen that the traditional theory of meaning is wrong; and this is
why the literature today contains many different concepts (e.g.,
'intension' and 'notional world') and not a single unitary con-
cept of 'meaning'. 'Meaning' has fallen to pieces. But we are left
with the task of picking up the pieces. If intension and extension
are not directly fixed by notional world, then how are they
fixed?
The received view of interpretation

The most common view of how interpretations of our language
are fixed by us, collectively if not individually, is associated with
the notions of an operational constraint and a theoretical con-
straint. Operational constraints were originally conceived of
rather naively; we simply stipulate (conventionally, as it were)
that a certain sentence (say, 'Electricity is flowing through this
wire') is to be true if and only if a certain test result is observed
(the voltmeter needle being deflected, or, in a phenomenalistic
version, my having the visual impression of seeing the voltmeter
needle being deflected). This sort of crude operationalism no
longer has any defenders because it has been appreciated that (1)
the links between theory and experience are probabilistic and
cannot be correctly formalized as perfect correlations (even if
there is current flowing through the wire, there are always low
probability events or background conditions which could pre-
vent the voltmeter needle from being deflected); and (2) even
these probabilistic links are not simple semantic correlations, but
depend on empirical theory which is subject to revision. On a
naive operationist account every time a new way of testing
whether a substance is really gold is discovered, the meaning and
reference of 'gold' undergoes a change. (In fact, we shouldn't
speak of a new test for gold being discovered.) On an operation-
ist picture, theories are tested sentence by sentence (the stipu-
lated operational meanings of the individual sentences tell you
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
how to go about testing the theory); on the more recent picture,

theories 'meet the test of experience as a corporate body', as
Quine puts it.
It is possible, however, to relax the notion of an operational
constraint so as to overcome or bypass all of these objections.
Thus one can restrict the class of interpretations (assignments of
intensions to the predicates of one's language) that will be
accepted as admissible by constraints of the form: 'an admissible
interpretation is such that most of the time the sentence S is true
when the experiential condition £ is fulfilled' (respectively, 'such
that most of the time the sentence S is false when E is fulfilled').
Such constraints model the idea that there are probabilistic rela-
tions between truth or falsity of sentences in the language and
experience. And, secondly, one can take the view that these con-
straints are revisable as theory develops. Rather than thinking of
them as meaning stipulations, as crude operationism did, one
can think of them as tentative restrictions on the class of admis-
sible interpretations; and with Peirce (who wrote 50 years before
Bridgeman announced 'operationism'!) one can take the view
that the ideal set of operational constraints is itself something
that we successively approximate in the course of empirical
inquiry, and not something we just stipulate. In short, one can
take the view that it is the operational constraints that rational
inquirers would impose, if they observed and experimented and
reasoned as well as is possible, the constraints that they would
adopt in the state of 'reflective equilibrium', that singles out the
interpretation of our terms; the constraints we actually accept at
any given time have the status of a rational estimate or approx-
imation.
Such a view is compatible with Quine's insistence that the
theory-experience links are just as much subject to revision as
any other aspect of our corporate body of knowledge. And it
does not see each such revision as a 'meaning change': such revi-
sions can be and often are simply better efforts to specify what
it is we have already been talking about; that which earlier the-
ory-cum-operational-constraints captured only inadequately.
In addition to restricting the class of admissible interpretations
by means of operational constraints (or successive approxima-
tions to a Peircian ideal set of operational constraints), one can
also have constraints which refer to formal properties of the the-
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
ory. For example, 'an admissible interpretation is such that it

turns out to be true that different effects always have different
causes'. Kant held that such a 'theoretical constraint' was part of
rationality itself: we impose the principle of determinism on the
world rather than discovering it. In this form, the constraint is
certainly too strong: the price of preserving determinism might
be too great a complication of our system of knowledge as a
whole. But this sort of constraint can be relaxed just as opera-
tional constraints have been relaxed. (For example, one can
require that determinism be preserved whenever the 'cost' in
terms of complications in other parts of the theory is not too
great; in this form, the constraint seems to be one we accept.)
Theoretical constraints are often stated as constraints on the-
ory acceptance rather than as constraints on theory interpreta-
tion, but they can easily be reinterpreted to play the latter role.
Thus, if an author states the constraint of 'conservativism' or
'preservationism' as a constraint on theory acceptance ('do not
accept a theory which requires giving up a great many previously
accepted beliefs if an - otherwise equally 'simple' - theory is
available which preserves those beliefs and agrees with observa-
tion'), then we can reformulate the constraint as a constraint on
interpretation thus: 'an admissible interpretation is such that it
renders true sentences which have been accepted for a long time,
except where this would require undue complication in the the-
ory consisting of the set of sentences true under the interpreta-
tion, or too great a revision in the operational constraints'.
Again, it has been widely held that no inductive logic is possible
unless we impose some a priori ordering (called a 'simplicity
ordering' or 'plausibility ordering') on the hypotheses which
may be accepted given particular observational data (although
the ordering may itself be different in different experimental or
observational contexts); the constraint that 'the set of sentences
which is true under an admissible interpretation must not be
lower down in the simplicity ordering than any other set with
the same observational or experiential consequences would cor-
respond to the constraint in inductive logic that one is to accept
the most simple (or most 'plausible') of the hypotheses compat-
ible with one's observations.
Theoretical constraints of many other kinds have been pro-
posed in the literature of the philosophy of science. There are
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
constraints which, like 'simplicity', refer to properties of the set

of accepted sentences, and constraints which refer to the history
of the inquiry by which that set came to be accepted. But the
details need not concern us. The reasons for being attracted to
the idea that the admissible interpretations of our language (in
the sense of admissible intension-assignments to the terms of the
language) are fixed by operational and theoretical constraints
are obvious: whether or not one is having an experience of a
certain kind is something the mind is able to judge (philosophical
problems about 'experience' notwithstanding). So if a theory
implies or contains a sentence which is associated with an expe-
rience E by an operational constraint of some kind, probabilistic
or whatever, then the thinker can know if the theory works, or
if there is some awkwardness of fit, at least in this case, by seeing
whether or not he has the experience £. Since the constraints
that we use to test the theory alsofixthe extensions of the terms,
the thinker's estimate of the theory's 'working' is at the same
time an estimate of its truth. Since the speaker's knowledge of
these constraints is knowledge of the intensions of the terms,
grasping a correct semantics would tell one, for any proposed
theory of the world, what our world would have to be like for
the theory to be true.
Furthermore, if we idealize by supposing thinkers to have
what economists call 'perfect information' about each other,
each thinker knows the formal structure of the accepted theory
T and the past history of the research program to which it
belongs, the previous beliefs that it does or does not preserve,
etc. So each thinker is in a position to know if the theoretical
constraints are met or not. (If we do not wish to idealize by
assuming perfect information, then we can still say that the col-
lective body of thinkers is in a position to know this.)
In short, if the received view is correct, then we would have an
elegant account of how intensions and extensions are fixed (in
principle — of course the details are too complicated to fill in at
the present stage of methodological knowledge). But, unfortu-
nately, the received view does not work!
Why the received view doesn't work

The difficulty with the received view is that it tries to fix the
intensions and extensions of individual terms byfixingthe truth-
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
conditions for whole sentences. The idea, as we just saw, is that

the operational and theoretical constraints (the ones rational
inquirers would accept in some sort of ideal limit of inquiry)
determine which sentences in the language are true. Even if this
is right, however, such constraints cannot determine what our
terms refer to. For there is nothing in the notion of an opera-
tional or theoretical constraint to do this directly. And doing it
indirectly, by putting down constraints which pick out the set of
true sentences, and then hoping that by determining the truth-
values of whole sentences we can somehow fix what the terms
occurring in those sentences refer to, won't work.
That it won't work has been shown by Quine.6 I shall extend
previous 'indeterminacy' results in a very strong way. I shall
argue that even if we have constraints of whatever nature which
determine the truth-value of every sentence in a language in
every possible world, still the reference of individual terms
remains indeterminate. In fact, it is possible to interpret the
entire language in violently different ways, each of them com-
patible with the requirement that the truth-value of each sen-
tence in each possible world be the one specified. In short, not
only does the received view not work; no view which only fixes
the truth-values of whole sentences can fix reference, even if it
specifies truth-values for sentences in every possible world.
The detailed proof is technical, and I think it appropriate to
give it in an Appendix. What I shall give here is an illustration of
the method of the proof only, and not the detailed proof.
Consider the sentence
(1) A cat is on a mat. (Here and in the sequel 'is on' is
tenseless, i.e. it means 'is, was, or will be on'.)
Under the standard interpretation this is true in those possible
worlds in which there is at least one cat on at least one mat at
some time, past, present, or future. Moreover, 'cat' refers to cats
and 'mat' refers to mats. I shall show that sentence (1) can be
reinterpreted so that in the actual world 'cat' refers to cherries
and 'mat' refers to trees without affecting the truth-value of sen-
tence (1) in any possible world. ('Is on' will keep its original
interpretation.)
6
See his 'Ontological Relativity', in Ontological Relativity and Other
Essays, Columbia University Press, 1969.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
The idea is that sentence (1) will receive a new interpretation

in which what it will come to mean is:
(a) A car" is on a maf\
The definition of the property of being a cat* (respectively, a
mar") is given by cases, the three cases being:
(a) Some cat is on some mat, and some cherry is on some
tree.
(b) Some cat is on some mat, and no cherry is on any
tree.
(c) Neither of the foregoing.
Here is the definition of the two properties:
DEFINITION OF 'CAT*'
x is a cat* if and only if case (a) holds and x is a cherry;
or case (b) holds and x is a cat; or case (c) holds and x is
a cherry.
DEFINITION OF 'MAT*'
x is a mat* if and only if case (a) holds and x is a tree;
or case (b) holds and x is a mat; or case (c) holds and
x is a quark.
Now, in possible worlds falling under case (a), 'A cat is on a

mat' is true, and 'A cat* is on a mat*' is also true (because a
cherry is on a tree, and all cherries are cats* and all trees are
mats* in worlds of this kind). Since in the actual world some
cherry is on some tree, the actual world is a world of this kind,
and in the actual world 'cat*' refers to cherries and 'mat*' refers
to trees.
In possible worlds falling under case (b), 'A cat is on a mat' is
true, and 'A cat* is on a mat*' is also true (because in worlds
falling under case (b), 'cat' and 'cat*' are coextensive terms and
so are 'mat' and 'mat*'). (Note that although cats are cats* in
some worlds — the ones falling under case (b) — they are not
cats* in the actual world.)
In possible worlds falling under case (c), 'A cat is on a mat' is
false and 'A cat* is on a mat*' is also false (because a cherry
can't be on a quark).
Summarizing, we see that in every possible world a cat is on a
mat if and only if a cat* is on a mat*. Thus, reinterpreting the
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
word 'cat' by assigning to it the intension we just assigned to

'cat*' and simultaneously reinterpreting the word 'mat' by
assigning to it the intension we just assigned to 'mat*' would
only have the effect of making 'A cat is on a mat' mean what 'A
cat* is on a mat*' was defined to mean; and this would be per-
fectly compatible with the way truth-values are assigned to 'A
cat is on a mat' in every possible world.
In the Appendix, I show that a more complicated reinterpre-
tation of this kind can be carried out for all the sentences of a
whole language. It follows that there are always infinitely many
different interpretations of the predicates of a language which
assign the 'correct' truth-values to the sentences in all possible
worlds, no matter how these 'correct' truth-values are singled
out. Quine argued for a similar conclusion in Word and Object;
in Quine's example (as applied to English) 'There is a rabbit over
there' was interpreted to mean 'There is a rabbit-slice over there'
(where a 'rabbit-slice' is a three-dimensional spatial cross-
section of the whole four-dimensional space—time rabbit), or,
alternatively again, to mean, 'There is rabbithood being exem-
plified again.' (This last reinterpretation also reinterprets the
syntactic form of the sentence, or at least its logical grammar.)
Quine makes the point I just made, that truth-conditions for
whole sentences underdetermine reference. Since 'rabbit-slices',
'rabbithood', and 'undetached rabbit-parts' all have a close con-
nection to rabbits, one might come away from Word and Object
with the impression that all reinterpretations that leave a sen-
tence's truth-value unchanged are at least closely connected with
the standard interpretation (in the way that rabbit-parts and rab-
bithood are connected with rabbits). The argument spelled out
in the Appendix and illustrated in this chapter shows that the
truth conditions for 'A cat is on a mat' don't even exclude the
possibility that 'cat' refers to cherries.
'Intrinsic' and 'extrinsic'

Perhaps the first idea that comes to mind when one is confronted
by non-standard interpretations, such as the one that interprets
'cat' as cat* and 'mat' as mat* is to dismiss them as presenting
us with an unimportant paradox. But genuine paradoxes are
never unimportant; they always show something is wrong with
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
the way we have been thinking. Perhaps the second reaction is to

protest that cat* and mat* are 'queer' properties; surely our
terms correspond to 'sensible' properties (such as being a cat or
being a mat) and not to such 'funny' properties as these. One
might explicate the way in which cat* (or, rather, cathood* or
cat*hood) is a funny property by pointing out that one can
'build a machine' to 'inspect' things and 'tell' whether or not
they are cats (a human being is such a 'machine'), but one cannot
build a machine that will tell (in any world which resembles ours
in its laws and general conditions) whether or not something is
a cat*. If the machine (or a person) looks at something and sees
it is neither a cat nor a cherry, then they can tell it is not a cat*;
but if the thing is either a cat or a cherry then the device or the
person needs to be informed of the truth-values of 'A cat is on a
mat' and 'A cherry is on a tree' to decide if it is inspecting or
seeing a cat*, and these truth-values go beyond what it can learn
by just examining the object presented to it for inspection.
Unfortunately, one can reinterpret 'sees' (say, as sees*) so that
the two sentences (3) John (or whoever) sees a cat; and (4) John
sees* a cat*, will have the same truth-value in every possible
world (by the method given in the Appendix). So whenever a
person sees a cat, he is seeing* a cat*; the experience we typically
have when we see a cat is the experience we typically have when
we see* a cat*, and so on. Similarly, we can reinterpret 'inspects'
and 'tells' so that, when a machine inspects a cat, it is inspecting*
a cat*, and when it 'tells' something is a cat, it is telling* that it
is a cat*.
To use an illustration (suggested by Nozick), suppose half of
us (the females perhaps) use 'cat' to mean 'cat*', 'mat' to mean
'mat*', 'look' to mean 'look*', 'tells' to mean 'tells*', and so on.
Suppose the other half (the males) use 'cat' to denote cats, 'mat'
to denote mats, 'look' to denote looking, and so on. How could
we ever know?7 (If you ask a male what 'cat' refers to, he will
answer 'to cats, of course' and so will a female, whatever 'cat'
refers to.)
7
A female might answer that the supposition that she is referring to
cats* when she says 'cat' is incoherent (because within her language
whatever she refers to as a 'cat' is a cat). This answer is small comfort; it
does not exclude the possibility that what she calls a cat is what males
call a cat*, and vice versa; and this is Nozick's point.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
The point is that the fact that one can build a machine to
inspect things and tell if they are cats differentiates cats from
cats* if one can be sure 'inspect' and 'tell' refer to inspecting and
telling, and it is no easier to say how the reference of these words
is fixed than to say how the reference of 'cat' is fixed. One might
say that when I look at something and think that it is a cat, my
'mental representations', the visual images or tactile images, the
verbalized thought 'cat', and so on, refer to cathood and to var-
ious other physical or biological properties (being a certain
shape, being a certain color, belonging to a certain species) and
not to their counterparts; this may be true, but it just repeats
that the reference is fixed one way rather than the other. This is
what we want to explain and not the explanation sought.
'But,' one might protest, 'the definitions of'cat*" and "mat*"
given above refer to things other than the object in question
(cherries on trees and cats on mats), and thus signify extrinsic
properties of the objects that have these properties. In the actual
world, every cherry is a car"; but it would not be a cat*, even
though its intrinsic properties would be exactly the same, if no
cherry were on any tree. In contrast, whether or not something is
a cat depends only upon its intrinsic properties.' Is the distinc-
tion here referred to, the distinction between intrinsic and extrin-
sic properties, one that will enable us to characterize and rule
out 'queer' interpretations?
The trouble with this suggestion is a certain symmetry in the
relation of'cat' and 'mat' to 'cat*' and 'mat*'. Thus, suppose we
define 'cherry*' and 'tree*' so that in possible worlds falling
under case (a) cherries* are cats and trees* are mats; in possible
worlds falling under case (b) cherries* are cherries and trees* are
trees; and in possible worlds falling under case (c) cherries* are
cats and trees* are photons. Then we can define 'cat' and 'mat'
by means of the *-terms as follows: Cases:
(a*) Some cat* is on some mat*, and some cherry* is on
some tree*.
(b*) Some cat* is on some mat*, and no cherry* is on
any tree*.
(c*) Neither of the foregoing.
Strangely enough, these cases are just our old (a), (b), (c) under
a new description. Now we define:
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
DEFINITION OF 'CAT'
x is a cat if case (a*) holds and x is a cherry"'; or case
(b *) holds and x is a cat *; or case (c *) holds and x is a
cherry*. (Note that in all three cases cats come out being
cats.)
DEFINITION OF 'MAT'
x is a mat if and only if case (a*) holds and x is a tree*;
or case (b*) holds andx is a mat*; or case (c*) holds
andx is a quark*. (Supposing quark* to be defined so
that in cases of type (c *) quarks * are mats, in all three
cases mats come out being mats.)
The upshot is that viewed from the perspective of a language
which takes 'cat*', 'mat*', etc., as primitive properties, it is 'cat'
and 'mat' that refer to 'extrinsic' properties, properties whose
definitions mention objects other than x; while relative to 'nor-
mal' language, language which takes 'cat' and 'mat' to refer to
cathood and mathood (you know which properties I mean, dear
reader!), it is 'cat*' and 'mat*' that refer to 'extrinsic' properties.
Better put, being 'intrinsic' or 'extrinsic' are relative to a choice
of which properties one takes as basic; no property is intrinsic or
extrinsic in itself.
'Survival' and evolution

The suggestion is popular nowadays that the evolutionary pro-
cess itself has somehow produced a correspondence between our
words and mental representations and external things; people
say that we would not have survived if there had not been such
a correspondence, and that this correspondence is, at least in a
primitive way, the relation of reference.
But what do 'correspondence' and 'reference' have to do with
survival? For that matter, what does truth have to do with sur-
vival?
Here opinions differ. Some philosophers believe that we
would not survive if (sufficiently many of) our beliefs were not
true. Other philosophers claim that even our best established sci-
entific beliefs aren't true, or at least that we have no reason to
think they are. Thomas Kuhn has suggested that our beliefs only
'refer' to objects within those beliefs (somewhat in the way in
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
which 'Hamlet' only refers to a person in a play); the success of

science is explained by trial-and-error, not by any correspon-
dence between its objects and real things, Kuhn says. Bas van
Fraassen, in a new book, argues that a successful theory need
not be true but only 'observationally adequate', i.e. correctly
predict observation. He too explains the success (or 'observa-
tional adequacy') of science as the product of trial-and-error.
If these philosophers are right, then the whole idea of using
evolution to justify belief in an objective relation of reference is
undercut. Evolution, on such instrumentalist views, only estab-
lishes a correspondence between some terms (the observation
terms) and 'permanent possibilities of sensation'. Such a corre-
spondence is not reference, unless we are willing to abandon the
idea that external things (the observable ones) are more than
constructs out of sensations.
I believe that the other philosophers are right, however (the
ones who say we would not survive if sufficiently many of our
beliefs were not true).
The reason I believe this is that trial-and-error does not
explain why our theories are 'observationally adequate'; that
can only be explained by referring to characteristics of the envi-
ronment-human interaction which explain why trial-and-error
is successful. (Trial-and-error does not succeed in all enterprises,
after all!) To posit that the interaction produces in our minds
false theories which just happen to have successful predictions
as consequences is to posit a totally inexplicable series of coin-
cidences. But how does the fact that our beliefs are (approxi-
mately) true explain our survival?
Some of our beliefs are intimately connected with action. If I
believe the sentence 'I will get something I value very much if I
push that button' (assume I understand this sentence in a normal
way, or at least associate the normal 'bracketed' or 'notional'
belief with it), then I will reach out my hand and push the but-
ton. Call beliefs of the form 'If I do x, I will get . . .', where the
blank describes a goal the agent has, directive beliefs. If too
many of our directive beliefs are false, we will perform too many
unsuccessful actions; so truth of (sufficiently many of) our direc-
tive beliefs is necessary for survival.
Now, our directive beliefs are themselves derived from many
other beliefs: beliefs about the characteristics and causal powers
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
of external things, and beliefs about our own characteristics and

powers. If these beliefs were mainly false, would it not be a mere
coincidence if they nonetheless led to true prediction of experi-
ence and to true directive beliefs? So, since (sufficiently many of)
our directive beliefs are true, and the best explanation of this fact
is that many of our other beliefs (the ones constituting our 'the-
ory of the everyday world') are at least approximately true, we
are justified in believing that our theory of the everyday world is
at least approximately true, and that we would not have sur-
vived if this were not the case.
Imagine, now, that some of us are actually referring to the
things that are assigned to our terms by the non-standard inter-
pretation / (described in the Appendix). This interpretation
agrees with the standard interpretation on terms referring to our
notional world, our sensations, our volitions, etc. So 'I seem to
myself to push the button', when understood in the 'bracketed
sense' (as meaning that I have a certain subjective experience of
voluntarily pushing a button) has not just the same truth condi-
tions but the same interpretation under/ and under the 'normal'
interpretation /, and so does 'I seem to myself to get the satisfac-
tion I expected.'
Now, if sufficiently many of our directive beliefs are true
under the non-standard interpretation /, then we will certainly
be successful, and we will certainly survive (since if we weren't
alive we wouldn't be attaining these goals) and have offspring
(since if they weren't alive they wouldn't be attaining these
goals). In short,/-truth of (sufficiently many) directive beliefs is
as good for 'evolutionary success' as /-truth. In fact it is /-truth,
since the truth conditions for every sentence (not just directive
beliefs) are the same under / and under /. My directive beliefs
are not only associated with the same subjective experience un-
der the interpretation / and under the interpretation/; they have
the same truth conditions. From the point of view (or non-point
of view) of 'evolution', all that is necessary is that sufficiently
many of my beliefs be true under any interpretation that con-
nects those beliefs with the relevant actions. Evolution may pro-
duce in me a tendency to have true beliefs (of certain kinds); but
this only means that evolution affects linguistically mediated or
conceptually mediated survival via its tendency to produce in us
representation systems whose sentences or sentence-analogues
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
have certain truth conditions (and certain action conditions, or

'language exit rules'). But the truth-conditions for whole sen-
tences were just shown not to determine the reference of sen-
tence parts (nor does adding the 'language exit rules' help, for
these are preserved under/). It follows that it is simply a mistake
to think that evolution determines a unique correspondence (or
even a reasonably narrow range of correspondences) between
referring expressions and external objects.
Intentions: pure and impure

We have seen that nature does not single out any one correspon-
dence between our terms and external things. Nature gets us to
process words and thought signs in such a way that sufficiently
many of our directive beliefs will be true, and so that sufficiently
many of our actions will contribute to our 'inclusive genetic fit-
ness'; but this leaves reference largely indeterminate. W. V.
Quine has urged that that is what reference in fact is — indeter-
minate! It is just an illusion, he thinks, that the terms in our
language have determinate well-defined counterparts. As he puts
it,
For, consider again our standard regimented nota-
tion, with a lexicon of interpreted predicates and some
fixed range of values for the variables of quantifica-
tion. The sentences of this language that are true remain
true under countless reinterpretations of the predicates
and revisions of the range of values of the variables.
Indeed any range of the same size can be made to serve by
a suitable reinterpretation of the predicates. If the range
of values is infinite, any infinite range can be made to
serve; this is the Skolem-Lowenheim theorem. The true
sentences stay true under all such changes.
Perhaps then our primary concern belongs with the
truth of sentences and with their truth conditions, rather
than with the reference of terms.
In the next chapter I will explore the alternative here sug-
gested, of giving up the idea that has so far been the premiss of
the entire discussion: that words stand in some sort of one—one
relation to (discourse-independent) things and sets of things. It
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
may seem, however, that there is a much simpler way out: why
not just say that it is our intentions, implicit or explicit, that fix
the reference of our terms?
At the beginning of the discussion in the previous chapter, I
rejected this as not constituting an informative answer on the
ground that having intentions (of the relevant kind) presupposes
the ability to refer. It may be good at this stage to expand upon
this brief remark.
The problem is that the notions 'intention' and 'mental state'
have a certain ambiguity. Let us call a mental state a pure mental
state if its presence or absence depends only on what goes on
'inside' the speaker. Thus whether or not I have a pain depends
only on what goes on 'inside' me, but whether or not I know
that snow is white depends not only on whether or not some-
thing goes on 'inside' me (believing or being confident that snow
is white), but also on whether or not snow is white, and thus is
something 'outside' my body and mind. Thus pain is a pure men-
tal state but knowledge is an impure mental state. There is a
(pure) mental state component to knowledge, but there is also a
component which is not mental in any sense: this is the compo-
nent that corresponds to the condition that what a man believes
is not knowledge unless the belief is true. I am not in the 'state'
of knowing that snow is white if I am not in a suitable pure
mental state; but being in a suitable pure mental state is never
sufficient for knowing that snow is white; the world has to coop-
erate as well.
What about belief? We have defined bracketed belief
('notional world') so that having a bracketed belief that [there is
water on the table] or having a notional world which includes
there being water on the table is a pure mental state. But, in
accordance with what was said before, believing that there is
water on the table (without any 'bracketing') presupposes that
one's word 'water' actually refers to water, and this depends on
the actual nature of certain 'paradigms', one's direct or indirect
causal relations to those paradigms, and so on. When I have the
belief that there is water on the table, my Doppleganger on Twin
Earth has the same bracketed belief but not the same belief
because his word 'water' refers to water-with-grain alcohol and
not to water. In short, believing that there is water on the table
is an impure mental state. (Brains in a Vat could not be in this
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
state, although they could be in the corresponding 'bracketed'

state.)
What goes for belief goes for intention as well. Pure mental
states of intending - e.g. intending that the term 'water' refer to
water in one's notional world — do not fix real world reference
at all. Impure mental states of intending — e.g. intending that the
term 'water' refer to actual water - presuppose the ability to
refer to (real) water.
Some philosophers have suggested that belief can be defined
in terms of the state I called 'bracketed belief and reference,
thus:
John believes that snow is white = John believes that
[snow is white]
(i.e. snow is white in John's notional world)
and the words 'snow' and 'white9 in John's thought {or
whatever words he uses to express this belief) refer
to snow and to the property white, respectively.
Without accepting this as a correct and complete analysis of
what it is to believe that snow is white, we can accept this
account as making a point which is certainly correct: that believ-
ing presupposes the ability to refer. And in exactly the same way,
intending presupposes the ability to refer! Intentions are not
mental events that cause words to refer: intentions (in the ordi-
nary 'impure' sense) have reference as an integral component.
To explain reference in terms of (impure) intention would be
circular. And the problem of how pure mental states of intend-
ing, believing, etc., can (in the proper causal setting) constitute
or cause reference is just what we have found so puzzling.
The origin of the puzzle

At first blush, nothing seems more obvious than that our words
and mental representations refer. When I think or say 'the cat
just went out', the thought is usually about our cat Mitty; the
word 'cat' in the sentence I think or say refers to a set of entities
of which Mitty is a member. Yet we have just seen that the
nature of this relation of 'aboutness' or reference is puzzling.
The distinction between real world and notional world (and
the correlative distinction between beliefs and bracketed beliefs,
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
or intentions and bracketed intentions) itself explains part of the

puzzle. The reason that it is surprising and troubling to discover
that there are unintended 'admissible interpretations' of our lan-
guage (where by an admissible interpretation I mean simply an
interpretation that satisfies the appropriate operational and the-
oretical constraints) is, in part, that no such 'indeterminacy' rises
in the 'notional world' of the speaker. In my notional world, cats
and cats51' are quite distinct (in fact, in my notional world cats*
are cherries). 'There is a cat on a mat' and 'there is a car" on a
mat*' may be logically equivalent, but they contain terms with
quite different notional referents; thus it seems strange indeed
that there should be any confusion between the real world refer-
ents of the one belief and the real world referents of the other.
But if the number of cats happens to be equal to the number
of cherries, then it follows from theorems in the theory of models
(as Quine remarks in the passage quoted above) that there is a
reinterpretation of the entire language that leaves all sentences
unchanged in truth value while permuting the extensions of 'cat'
and 'cherry'. By the techniques just mentioned, such reinterpre-
tations can be constructed so as to preserve all operational and
theoretical constraints (and by the techniques we illustrated with
the 'cat/cat*' example, they can be extended so as to provide
'intensions', or functions which determine an extension in each
possible world, and not just extensions in the actual world). This
does not contradict the statements just made about our 'notional
world', or subjective belief system, for the following reason: the
fact that in our belief system or 'notional world' no cat is a
cherry means that in each admissible interpretation of that belief
system (each assignment of external world referents to the terms,
images, and other representations we employ in thought) the
referents of 'cat' and the referents of 'cherry' must be disjoint
sets. But the disjointness of these sets is comparable with the
(remarkable) fact that what is the set of 'cats' in one admissible
interpretation may be the set of 'cherries' in a different (but
equally admissible) interpretation. From the fact that notional
cats are wholly different from notional cherries it only follows
that real cats are wholly different from real cherries if the num-
ber of admissible interpretations is exactly one. If there is more
than one admissible interpretation of the whole language (as
there will be if the admissible interpretations are singled out only
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
by operational and theoretical constraints), then two terms

which refer to disjoint sets in each admissible interpretation can
have the same potential referents when the totality of all admis-
sible interpretations is considered. From the fact that notional
cats are as different as can be from notional cherries it does not
follow that there are determinate disjoint sets of cats-in-
themselves and cherries-in-themselves.
What makes this so distressing is that operational plus theo-
retical constraints are the natural way in which to allow the
actual empirical context to determine the admissible interpreta-
tion (or interpretations) of one's representational system. Such
constraints can to some extent determine which sentences in
one's language are true and which false; it is the slack between
truth-conditions and reference that remains.
Quine, as we remarked, would be willing to put up with the
slack and simply acknowledge that reference is indeterminate. A
young philosopher, Hartry Field,8 has recently suggested a dif-
ferent view. In Field's view reference is a 'physicalistic relation',
i.e. a complex causal relation between words or mental represen-
tations and objects or sets of objects. It is up to empirical science
to discover what that physicalistic relation is, Field suggests.
There is, however, a problem with this suggestion too. Sup-
pose there is a possible naturalistic or physicalistic definition of
reference, as Field contends. Suppose
(1) x refers to y if and only if x bears R to y
is true, where R is a relation definable in natural science vocab-
ulary without using any semantical notions (i.e. without using
'refers' or any other words which would make the definition
immediately circular). If (1) is true and empirically verifiable,
then (1) is a sentence which is itself true even on the theory that
reference is fixed as far as (and only as far as) is determined by
operational plus theoretical constraints. (1) is a sentence which
would be part of our 'reflective equilibrium' or 'ideal limit' the-
ory of the world.
If reference is only determined by operational and theoretical
constraints, however, then the reference of (x bears R to y' is
8
Field, H., 'Tarski's Theory of Truth', The Journal of Philosophy, vol. 69.
Field's view is discussed in my Meaning and the Moral Sciences,
Routledge and Kegan Paul, 1978.
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
itself indeterminate, and so knowing that (1) is true will not help.
Each admissible model of our object language will correspond
to a model of our meta-language in which (1) holds; the inter-
pretation of 'x bears R to y' will fix the interpretation of (x refers
to y\ But this will only be a relation in each admissible model; it
will not serve to cut down the number of admissible models at
all.
This is, of course, not at all what Field intends. What Field is
claiming is that {a) there is a determinate unique relation
between words and things or sets of things; and (b) this relation
is the one to be used as the reference relation in assigning a truth
value to (1) itself. But this is not necessarily expressed by just
saying (1), as we have just seen; and it is a puzzle how we could
learn to express what Field wants to say.
Putting this last puzzle aside, let us consider the view that (1),
understood as Field wants us to understand it (as describing the
determinate, unique relation between words and their referents),
is true. If (1) is true, so understood, what makes it true? Given
that there are many 'correspondences' between words and
things, even many that satisfy our constraints, what singles out
one particular correspondence R? Not the empirical correctness
of (1); for that is a matter of our operational and theoretical
constraints. Not, as we have seen, our intentions (rather R enters
into determining what our intentions signify). It seems as if the
fact that R is reference must be a metaphysically unexplainable
fact, a kind of primitive, surd, metaphysical truth.
This kind of primitive, surd, metaphysical truth, if such there
be, must not be confused with the sort of 'metaphysically neces-
sary' truth recently introduced by Saul Kripke.9
Kripke's point, which is closely related to points made above
about the reference of natural kind terms (terms for animal, veg-
etable and mineral species, for example), was that given that, as
a matter of fact,
(2) Water is H2O
(i.e. given that (2) is true in the actual world), and given that
(Kripke points out) speakers intend that the term 'water' shall
9
See his Naming and Necessity, Harvard University Press, 1980.
(Originally given as lectures in 1970.)
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
refer to just those things that have the same lawful behavior and
the same ultimate composition as various standard samples of
actual water (i.e. speakers have such intentions even when talk-
ing about hypothetical cases or 'possible worlds'), it follows that
(2) must also be true in every possible world; for to describe a
hypothetical liquid which is not H2O but which has some simi-
larities to water is only to describe a hypothetical liquid which
resembles water, and not to describe a possible world in which
water isn't H2O. It is 'metaphysically necessary' (true in all pos-
sible worlds) that water is H2O; but this 'metaphysical necessity'
is explained by mundane chemistry and mundane facts about
speakers' intentions to refer.
If there is a determinate physicalistic relation R (whether it be
definable in the language of natural science in finitely many
words or not) which just is reference (independently of how or
whether we describe that relation), this fact cannot itself be the
consequence of our intentions to refer; rather, as we have repeat-
edly noted, it enters into determining what our very intentions
to refer signify. Kripke's view, that 'water is H 2 O' is true in all
possible worlds, could be right even if reference in the actual
world is fixed only by operational and theoretical constraints;
the view presupposes the notion of reference, it does not tell us
whether reference is determinate or what reference is.
To me, believing that some correspondence intrinsically just is
reference (not as a result of our operational and theoretical con-
straints, or our intentions, but as an ultimate metaphysical fact)
amounts to a magical theory of reference. Reference itself
becomes what Locke called a 'substantial form' (an entity which
intrinsically belongs with a certain name) on such a view. Even
if one is willing to contemplate such unexplainable metaphysical
facts, the epistemological problems that accompany such a meta-
physical view seem insuperable. For, assuming a world of mind-
independent, discourse-independent entities (this is the presup-
position of the view we are discussing), there are, as we have
seen, many different 'correspondences' which represent possible
or candidate reference relations (infinitely many, in fact, if there
are infinitely many things in the universe). Even requiring that
(1) be true under whichever notion of truth corresponds to the
metaphysically singled-out 'real' relation of reference does not
exclude any of these candidates, if (1) is itself empirically accept-
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
able (acceptable given our operational and theoretical con-

straints), as we have seen. But then there are infinitely many dif-
ferent possible 'surd metaphysical truths' of the form (R is the
real (metaphysically singled-out) relation of reference'. If the
holder of the view allows that it is conceivable that his view is
not quite right, and that reference may be metaphysically singled
out without being totally determinate (the metaphysically sin-
gled-out R may allow for a plurality of admissible interpreta-
tions) then it is even conceivable that the operational-/?/ws-
theoretical-constraints-view is metaphysically correct after all!
For why could it not be a surd metaphysical fact that reference
is the relation: x refers to y in at least one admissible model M.
Note that all these infinitely many metaphysical theories are
compatible with the same sentences being true, the same 'theory
of the world', and the same optimal methodology for discover-
ing what is true!
Cambridge
https://doi.org/10.1017/CBO9780511625398.004
Two philosophical perspectives
The problems we have been discussing naturally give rise to two

philosophical points of view (or two philosophical tempera-
ments, as I called them in the Introduction). It is with these
points of view, and with their consequences for just about every
issue in philosophy that I shall be concerned: the question of
'Brains in a Vat' would not be of interest, except as a sort of
logical paradox, if it were not for the sharp way in which it
brings out the difference between these philosophical perspec-
tives.
One of these perspectives is the perspective of metaphysical
realism. On this perspective, the world consists of some fixed
totality of mind-independent objects. There is exactly one true
and complete description of 'the way the world is'. Truth
involves some sort of correspondence relation between words or
thought-signs and external things and sets of things. I shall call
this perspective the externalist perspective, because its favorite
point of view is a God's Eye point of view.
The perspective I shall defend has no unambiguous name. It
is a late arrival in the history of philosophy, and even today it
keeps being confused with other points of view of a quite differ-
ent sort. I shall refer to it as the internalist perspective, because
it is characteristic of this view to hold that what objects does the
world consist off is a question that it only makes sense to ask
within a theory or description. Many 'internalist' philosophers,
though not all, hold further that there is more than one 'true'
theory or description of the world. 'Truth', in an internalist
view, is some sort of (idealized) rational acceptability - some
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
50 Two philosophical perspectives
sort of ideal coherence of our beliefs with each other and with
our experiences as those experiences are themselves represented
in our belief system — and not correspondence with mind-inde-
pendent or discourse-independent 'states of affairs'. There is no
God's Eye point of view that we can know or usefully imagine;
there are only the various points of view of actual persons
reflecting various interests and purposes that their descriptions
and theories subserve. ('Coherence theory of truth'; 'Non-real-
ism'; 'Verificationism'; 'Pluralism'; 'Pragmatism'; are all terms
that have been applied to the internalist perspective; but every
one of these terms has connotations that are unacceptable
because of their other historic applications.)
Internalist philosophers dismiss the 'Brain in a Vat' hypothe-
sis. For us, the 'Brain in a Vat World' is only a story, a mere
linguistic construction, and not a possible world at all. The idea
that this story might be true in some universe, some Parallel
Reality, assumes a God's Eye point of view from the start, as is
easily seen. For from whose point of view is the story being told?
Evidently not from the point of view of any of the sentient crea-
tures in the world. Nor from the point of view of any observer
in another world who interacts with this world; for a 'world' by
definition includes everything that interacts in any way with the
things it contains. If you, for example, were the one observer
who was not a Brain in a Vat, spying on the Brains in a Vat, then
the world would not be one in which all sentient beings were
Brains in a Vat. So the supposition that there could be a world
in which all sentient beings are Brains in a Vat presupposes from
the outset a God's Eye view of truth, or, more accurately, a No
Eye view of truth — truth as independent of observers altogether.
For the externalist philosopher, on the other hand, the
hypothesis that we are all Brains in a Vat cannot be dismissed so
simply. For the truth of a theory does not consist in its fitting the
world as the world presents itself to some observer or observers
(truth is not 'relational' in this sense), but in its corresponding to
the world as it is in itself. And the problem that I posed for the
externalist philosopher is that the very relation of correspon-
dence on which truth and reference depend (on his view) cannot
logically be available to him if he is a Brain in a Vat. So, if we
are Brains in a Vat, we cannot think that we are, except in the
bracketed sense [we are Brains in a Vat]; and this bracketed
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
Two philosophical perspectives 51
thought does not have reference conditions that would make it

true. So it is not possible after all that we are Brains in a Vat.
Suppose we assume a 'magical theory of reference'. For exam-
ple, we might assume that some occult rays — call them 'noetic
rays' 1 -connect words and thought-signs to their referents.
Then there is no problem. The Brain in a Vat can think the
words, 'I am a brain in a vat', and when he does the word 'vat'
corresponds (with the aid of the noetic rays) to real external vats
and the word 'in' corresponds (with the aid of the noetic rays)
to the relation of real spatial containment. But such a view is
obviously untenable. No present day philosopher would
espouse such a view. It is because the modern realist wishes to
have a correspondence theory of truth without believing in
'noetic rays' (or, believing in Self-Identifying Objects2 - objects
that intrinsically correspond to one word or thought-sign rather
than another) that the Brain in a Vat case is a puzzler for him.
As we have seen, the problem is this: there are these objects
out there. Here is the mind/brain, carrying on its
thinking/computing. How do the thinker's symbols (or those of
his mind/brain) get into a unique correspondence with objects
and sets of objects out there?
The reply popular among externalists today is that while
indeed no sign necessarily corresponds to one set of things rather
than another, contextual connections between signs and external
things (in particular, causal connections) will enable one to
explicate the nature of reference. But this doesn't work. For
example, the dominant cause of my beliefs about electrons is
probably various textbooks. But the occurrences of the word
'electron' I produce, though having in this sense a strong connec-
tion to textbooks, do not refer to textbooks. The objects which
are the dominant cause of my beliefs containing a certain sign
may not be the referents of that sign.
The externalist will now reply that the word 'electron' is not
connected to textbooks by a causal chain of the appropriate
type. (But how can we have intentions which determine which
causal chains are 'of the appropriate type' unless we are already
able to refer?)
1
'Noetic rays' was suggested to me by Zemach.
2
The term 'Self Identifying Object' is from Substance and Sameness by
David Wiggins (Blackwell, 1980).
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
For an internalist like myself, the situation is quite different.

In an internalist view also, signs do not intrinsically correspond
to objects, independently of how those signs are employed and
by whom. But a sign that is actually employed in a particular
way by a particular community of users can correspond to par-
ticular objects within the conceptual scheme of those users.
'Objects' do not exist independently of conceptual schemes. We
cut up the world into objects when we introduce one or another
scheme of description. Since the objects and the signs are alike
internal to the scheme of description, it is possible to say what
matches what.
Indeed, it is trivial to say what any word refers to within the
language the word belongs to, by using the word itself. What
does 'rabbit' refer to? Why, to rabbits, of course! What does
'extraterrestrial' refer to? To extraterrestrials (if there are any).
Of course the externalist agrees that the extension of 'rabbit'
is the set of rabbits and the extension of 'extraterrestrial' is the
set of extraterrestrials. But he does not regard such statements
as telling us what reference is. For him finding out what refer-
ence is, i.e. what the nature of the 'correspondence' between
words and things is, is a pressing problem. (How pressing, we
saw in the previous chapter.) For me there is little to say about
what reference is within a conceptual system other than these
tautologies. The idea that causal connection is necessary is
refuted by the fact that 'extraterrestrial' certainly refers to extra-
terrestrials whether we have ever causally interacted with any
extraterrestrials or not!
The externalist philosopher would reply, however, that we
can refer to extraterrestrials even though we have never inter-
acted with any (as far as we know) because we have interacted
with terrestrials and we have experienced instances of the rela-
tion 'not from the same planet as' and instances of the property
'intelligent being'. And we can define an extraterrestrial as an
intelligent being that is not from the same planet as terrestrials.
Also, 'not from the same planet as' can be analyzed in terms of
'not from the same place as' and 'planet' (which can be further
analyzed). Thus the externalist gives up the requirement that we
have some 'real' connection (e.g. causal connection) with every-
thing we are able to refer to, and requires only that the basic
terms refer to kinds of things (and relations) that we have some
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
real connection to. Using the basic terms in complex conbina-

tions we can then, he says, build up descriptive expressions
which refer to kinds of things we have no real connection to, and
that may not even exist (e.g. extraterrestrials).
In fact, already with a simple word like 'horse' or 'rabbit' he
might have observed that the extension includes many things we
have not causally interacted with (e.g. future horses and rabbits,
or horses and rabbits that never interacted with any human
being). When we use the word 'horse' we refer not only to the
horses we have a real connection to, but also to all other things
of the same kind.
At this point, however, we must observe that 'of the same
kind' makes no sense apart from a categorial system which says
what properties do and what properties do not count as similar-
ities. In some ways, after all, anything is 'of the same kind' as
anything else. This whole complicated story about how we refer
to some things by virtue of the fact that they are connected with
us by 'causal chains of the appropriate kind', and to yet other
things by virtue of the fact that they are 'of the same kind' as
things connected with us by causal chains of the appropriate
kind, and to still other things 'by description', is not so much
false as otiose. What makes horses with which I have not inter-
acted 'of the same kind' as horses with which I have interacted
is that fact that the former as well as the latter are horses. The
metaphysical realist formulation of the problem once again
makes it seem as if there are to begin with all these objects in
themselves, and then I get some kind of a lassoo over a few of
these objects (the horses with which I have a 'real' connection,
via a 'causal chain of the appropriate kind'), and then I have the
problem of getting my word ('horse') to cover not only the ones
I have 'lassooed' but also the ones I can't lassoo, because they
are too far away in space and time, or whatever. And the 'solu-
tion' to this pseudo-problem, as I consider it to be — the meta-
physical realist 'solution' — is to say that the word automatically
covers not just the objects I lassooed, but also the objects which
are of the same kind — of the same kind in themselves. But then
the world is, after all, being claimed to contain Self-Identifying
Objects, for this is just what it means to say that the world, and
not thinkers, sorts things into kinds.
In a sense, I would say, the world does consist of 'Self-Identi-
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
fying Objects' — but not a sense available to an externalist. If, as

I maintain, 'objects' themselves are as much made as discovered,
as much products of our conceptual invention as of the 'objec-
tive' factor in experience, the factor independent of our will,
then of course objects intrinsically belong under certain labels;
because those labels are the tools we used to construct a version
of the world with such objects in the first place. But this kind of
'Self-Identifying Object' is not mind-independent; and the exter-
nalist wants to think of the world as consisting of objects that
are at one and the same time mind-independent and Self-Identi-
fying. This is what one cannot do.
Internalism and relativism

Internalism is not a facile relativism that says, 'Anything goes'.
Denying that it makes sense to ask whether our concepts 'match'
something totally uncontaminated by conceptualization is one
thing; but to hold that every conceptual system is therefore just
as good as every other would be something else. If anyone really
believed that, and if they were foolish enough to pick a concep-
tual system that told them they could fly and to act upon it by
jumping out of a window, they would, if they were lucky enough
to survive, see the weakness of the latter view at once. Internal-
ism does not deny that there are experiential inputs to knowl-
edge; knowledge is not a story with no constraints except inter-
nal coherence; but it does deny that there are any inputs which
are not themselves to some extent shaped by our concepts, by
the vocabulary we use to report and describe them, or any inputs
which admit of only one description, independent of all concep-
tual choices. Even our description of our own sensations, so dear
as a starting point for knowledge to generations of epistemolo-
gists, is heavily affected (as are the sensations themselves, for
that matter) by a host of conceptual choices. The very inputs
upon which our knowledge is based are conceptually contami-
nated; but contaminated inputs are better than none. If contam-
inated inputs are all we have, still all we have has proved to be
quite a bit.
What makes a statement, or a whole system of statements - a
theory or conceptual scheme - rationally acceptable is, in large
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
part, its coherence and fit; coherence of 'theoretical' or less

experiential beliefs with one another and with more experiential
beliefs, and also coherence of experiential beliefs with theoretical
beliefs. Our conceptions of coherence and acceptability are, on
the view I shall develop, deeply interwoven with our psychology.
They depend upon our biology and our culture; they are by no
means 'value free'. But they are our conceptions, and they are
conceptions of something real. They define a kind of objectivity,
objectivity for us, even if it is not the metaphysical objectivity of
the God's Eye view. Objectivity and rationality humanly speak-
ing are what we have; they are better than nothing.
To reject the idea that there is a coherent 'external' perspec-
tive, a theory which is simply true 'in itself, apart from all pos-
sible observers, is not to identify truth with rational acceptabil-
ity. Truth cannot simply be rational acceptability for one
fundamental reason; truth is supposed to be a property of a
statement that cannot be lost, whereas justification can be lost.
The statement 'The earth is flat' was, very likely, rationally
acceptable 3,000 years ago; but it is not rationally acceptable
today. Yet it would be wrong to say that 'the earth is flat' was
true 3,000 years ago; for that would mean that the earth has
changed its shape. In fact, rational acceptability is both tensed
and relative to a person. In addition, rational acceptability is a
matter of degree; truth is sometimes spoken of as a matter of
degree (e.g., we sometimes say, 'the earth is a sphere' is approx-
imately true); but the 'degree' here is the accuracy of the state-
ment, and not its degree of acceptability or justification.
What this shows, in my opinion, is not that the externalist
view is right after all, but that truth is an idealization of rational
acceptability. We speak as if there were such things as epistemi-
cally ideal conditions, and we call a statement 'true' if it would
be justified under such conditions. 'Epistemically ideal condi-
tions', of course, are like 'frictionless planes': we cannot really
attain epistemically ideal conditions, or even be absolutely cer-
tain that we have come sufficiently close to them. But frictionless
planes cannot really be attained either, and yet talk of friction-
less planes has 'cash value' because we can approximate them
to a very high degree of approximation.
Perhaps it will seem that explaining truth in terms of justifi-
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
cation under ideal conditions is explaining a clear notion in

terms of a vague one. But 'true' is not so clear when we move
away from such stock examples as 'Snow is white.' And in any
case, I am not trying to give a formal definition of truth, but an
informal elucidation of the notion.
The simile of frictionless planes aside, the two key ideas of the
idealization theory of truth are (1) that truth is independent of
justification here and now, but not independent of all justifica-
tion. To claim a statement is true is to claim it could be justified.
(2) truth is expected to be stable or 'convergent'; if both a state-
ment and its negation could be 'justified', even if conditions were
as ideal as one could hope to make them, there is no sense in
thinking of the statement as having a truth-value.
The 'similitude' theory

The theory that truth is correspondence is certainly the natural
one. Before Kant it is perhaps impossible to find any philosopher
who did not have a correspondence theory of truth.
Michael Dummett has recently3 drawn a distinction between
non-realist (i.e. what I am calling 'internalist') views and reduc-
tionist views in order to point out that reductionists can be
metaphysical realists, i.e. subscribers to the correspondence the-
ory of truth. Reductionism, with respect to a class of assertions
(e.g. assertions about mental events) is the view that assertions
in that class are 'made true' by facts which are outside of that
class. For example, facts about behavior are what 'make true'
assertions about mental events, according to one kind of reduc-
tionism. For another example, the view of Bishop Berkeley that
all there 'really is' is minds and their sensations is reductionist,
for it holds that sentences about tables and chairs and other
ordinary 'material objects' are actually made true by facts about
sensations.
If a view is reductionist with respect to assertions of one kind,
but only to insist on the correspondence theory of truth for sen-
3
Dummett's views are set out in 'What is a theory of Meaning I, IP in
Truth and Other Enigmas (Harvard, 1980). His forthcoming
(eventually) William James Lectures (given at Harvard in 1976) develop
them in much more detail.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
tences of the reducing class, then that view is metaphysical realist

at base. A truly non-realist view is non-realist all the way down.
The error is often made of regarding reductionist philosophers
as non-realists, but Dummett is surely right; their disagreement
with other philosophers is over what there really is, and not over
the conception of truth. If we avoid this error, then the claim I
just made, that it is impossible to find a philosopher before Kant
who was not a metaphysical realist, at least about what he took
to be basic or unreducible assertions, will seem much more
plausible.
The oldest form of the correspondence theory of truth, and
one which endured for approximately 2,000 years, is one that
ancient and medieval philosophers attributed to Aristotle. That
Aristotle actually held it I am not sure; but it is suggested by his
language. I shall call it the similitude theory of reference; for it
holds that the relation between the representations in our minds
and the external objects that they refer to is literally a similarity.
The theory, like modern theories, employed the idea of a men-
tal representation. This presentation, the mind's image of the
external thing, was called a phantasm by Aristotle. The relation
between the phantasm and the external object by virtue of which
the phantasm represents the external object to the mind is
(according to Aristotle) that the phantasm shares a form with
the external object. Since the phantasm and the external object
are similar (share the form), the mind, in having available the
phantasm, also has directly available the very form of the exter-
nal object.4
Aristotle himself says that the phantasm does not share with
the object such properties as redness (i.e. the redness in our
minds is not literally the same property as the redness of the
object), which can be perceived by one sense, but does share such
properties as length or shape which can be perceived by more
than one sense (which are 'common sensibles' as opposed to 'sin-
gle sensibles').
In the seventeenth century the similitude theory began to be
restricted, much as it had been by Aristotle. Thus Locke and
Descartes held that in the case of a 'secondary' quality, such as
a color or a texture, it would be absurd to suppose that the prop-
4
See De Anima, Book III, Ch. 7 and 8.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
erty of the mental image is literally the same property as the

property of the physical thing. Locke was a Corpuscularian, that
is, an advocate of the atomic theory of matter, and like a modern
physicist he conceived that what answers to the sensuous pre-
sented redness of my image of a red piece of cloth is not a simple
property of the cloth, but a very complex dispositional property
or 'Power', namely the Power to give rise to sensations of this
particular kind (sensations which exhibit 'subjective red', in the
language of psychophysics). This power in turn has an explana-
tion, which we did not know in Locke's day, in the particular
micro-structure of the piece of cloth which leads it to selectively
absorb and reflect light of different wave-lengths. (This sort of
explanation was already given by Newton.) If we say that having
such a microstructure is 'being red' in the case of a piece of cloth,
then clearly whatever the nature of subjective red may be, the
event in my mind (or even my brain) that takes place when I
have a sensation of subjective red does not involve anything in
my mind (or brain) 'being red'. The properties of a physical thing
which make it an instance of physical red and the properties of
a mental event which make it an instance of subjective red are
quite different. A red piece of cloth and a red after-image are not
literally similar. They do not share a Form.
For those properties (shape, motion, position) which his Cor-
puscularian philosophy led him to regard as basic and irreduci-
ble, Locke was willing to keep the similitude theory of reference,
however. (Actually, some Locke scholars today dispute this; but
Locke does say that there is a 'similitude' between the idea and
the object in the case of the primary qualities and that there is
'no similitude' between the idea of red or warmth and the red-
ness or warmth in the object.5 And the reading of Locke I am
describing was the universal one among his contemporaries and
among eighteenth century readers as well.)
Berkeley's tour de force

Berkeley discovered a very unwelcome consequence of the
similitude theory of reference: it implies that nothing exists
except mental entities ('spirits and their ideas', i.e. minds and
5
See An Essay Concerning Human Understanding, Book II, Ch. VIII.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
their sensations). It is generally unappreciated that the premiss

from which Berkeley worked - the similitude theory - was not
something he merely learned from Locke (or read into Locke)
but was the accepted theory of reference before his time and,
indeed, for a hundred years afterwards; but we have just
remarked how venerable this theory actually was.
Berkeley's argument is very simple. The usual philosophical
argument against the similitude theory in the case of secondary
qualities is correct (the argument from the relativity of percep-
tion), but it goes just as well in the case of primary qualities. The
length, shape, motion of an object are all perceived differently
by different perceivers and by the same perceiver on different
occasions. To ask whether a table is the same length as my image
of it or the same length as your image of it is to ask an absurd
question. If the table is three feet long, and I have a good clear
view of it, do I have a three foot long mental image? To ask the
question is to see its senselessness. Mental images do not have a
physical length. They cannot be compared with the standard
measuring rod in Paris. Physical length and subjective length
must be as different as physical redness and subjective redness.
To state Berkeley's conclusion another way, Nothing can be
similar to a sensation or image except another sensation or
image. Given this, and given the (still unquestioned) assumption
that the mechanism of reference is similitude between our 'ideas'
(i.e. our images or 'phantasms') and what they represent, it at
once follows that no 'idea' (mental image) can represent or refer
to anything but another image or sensation. Only phenomenal
objects can be thought about, conceived, referred to. And if you
can't think of something, you can't think it exists. Unless we
treat talk of material objects as highly derived talk about regu-
larities in our sensations, it is totally unintelligible.
The tendency, in his own time and later, to see Berkeley as
almost insanely perverse, almost scandalous, if brilliant, was due
to the unacceptability of his conclusion that matter does not
really exist (except as a construction from sensations), and not
to anything peculiar about his premisses. But the fact that one
could derive such an unacceptable conclusion from the simili-
tude theory produced a crisis in philosophy. Philosophers who
did not wish to follow Berkeley in Subjective Idealism had to
come up with a different account of reference.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
Kant's account of knowledge and truth

I want to say that, although Kant never quite says that this is
what he is doing, Kant is best read as proposing for the first time
what I have called the 'internalist' or 'internal realist' view of
truth.
To begin with, it is clear that Kant regarded Berkeley's Subjec-
tive Idealism as quite unacceptable (this much he explicitly says),
and also regarded causal realism — the view that we directly per-
ceive only sensations, and infer material objects via some kind of
problematical inference, as equally unacceptable. A view on
which it is only a very dubious hypothesis that there is a table in
front of me as I write these pages is a 'scandal', Kant says.
Secondly, I take it that Kant saw clearly how Berkeley's argu-
ment works: he saw that it depends on the similitude theory of
reference, and that rejecting Berkeley's argument requires reject-
ing that theory. Here I am attributing a view to Kant that Kant
does not express in these words (indeed, talk of 'reference' as the
relation between mental signs and what they stand for is very
recent, although the problem of the relation between mental
signs and what they stand for is very ancient). But we shall see
that what Kant did say has precisely the effect of giving up the
similitude theory of reference.
Let me suggest a way of reading Kant that may be helpful,
although it is only a first approximation to a right interpretation.
Think of Kant as accepting Berkeley's point that the argument
from the relativity of perception applies as much to the so-called
'primary' qualities as to the secondary ones, but making a differ-
ent response than Berkeley made. Berkeley's response, recall,
was to scrap the distinction between primary qualities and sec-
ondary qualities and fall back on just what Locke would have
called 'simple' qualities of sensation as the basic entities we can
refer to. Locke's own treatment of secondary qualities, recall,
was to say that (as properties of the physical object) we can only
conceive of them as Powers, as properties — nature unspecified —
which enable the object to affect us in a certain way. Saying that
something is red, or warm, or furry, is saying that it is so-and-so
in relation to us, not how it is from a God's Eye point of view.
I suggest that (as a first approximation) the way to read Kant
is as saying that what Locke said about secondary qualities is
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
true of all qualities — the simple ones, the primary ones, the sec-
ondary ones alike (indeed, there is little point of distinguishing
them).6
If all properties are secondary, what follows? It follows that
everything we say about an object is of the form: it is such as to
affect us in such-and-such a way. Nothing at all we say about
any object describes the object as it is 'in itself, independently of
its effect on us, on beings with our rational natures and our
biological constitutions. It also follows that we cannot assume
any similarity ('similitude', in Locke's English) between our idea
of an object and whatever mind-independent reality may be ulti-
mately responsible for our experience of that object. Our ideas
of objects are not copies of mind-independent things.
This is very much the way Kant describes the situation. He
does not doubt that there is some mind-independent reality; for
him this is virtually a postulate of reason. He refers to the ele-
ments of this mind-independent reality in various terms: thing-
in-itself (Ding an sich); the noumenal objects or noumena; col-
lectively, the noumenal world. But we can form no real concep-
tion of these noumenal things; even the notion of a noumenal
world is a kind of limit of thought (Grenz-Begriff) rather than a
clear concept. Today the notion of a noumenal world is per-
ceived to be an unnecessary metaphysical element in Kant's
thought. (But perhaps Kant is right: perhaps we can't help think-
6
Kant gives a summary of his own view in precisely this way in the
Prolegomena:
Long before Locke's time, but assuredly since him, it has been
generally assumed and granted without detriment to the actual
existence of external things that many of their predicates may be
said to belong, not to the things in themselves, but to their
appearances, and to have no proper existence outside our repre-
sentation. Heat, color, and taste, for instance, are of this kind.
Now, if I go farther and, for weighty reasons, rank as mere
appearances the remaining qualities of bodies also, which are
called primary - such as extension, place, and, in general, space,
with all that which belongs to it (impenetrability or materiality,
shape, etc.) - no one in the least can adduce the reason of its
being inadmissible. As little as the man who admits colors not to
be properties of the object in itself, but only as modifications of
the sense of sight, should on that account be called an idealist, so
little can my thesis be named idealistic merely because I find that
more, nay, all the properties which constitute the intuition of a
body belong merely to its appearance.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
ing that there is somehow a mind-independent 'ground' for our

experience even if attempts to talk about it lead at once to non-
sense.)
At the same time, talk of ordinary 'empirical' objects is not
talk of things-in-themselves but only talk of things-for-us.
The really subtle point is that Kant regards all of these points
as applying to sensations ('objects of internal sense') as well as
to external objects. This may seem strange: what is the problem
about whether or not an idea corresponds to a sensation? But
Kant is on to something profound.
Suppose I have a sensation E. Suppose I describe E; say, by
asserting (E is a sensation of red/ If 'red' just means like this,
then the whole assertion just means (E is like this' (attending to
£), i.e. E is like E — and no judgment has really been made. As
Wittgenstein puts it, one is reduced to virtually a grunt. On the
other hand, if 'red' is a true classifier, if I am claiming that this
sensation E belongs in the same class as sensations I call 'red' at
other times, then my judgment goes beyond what is immediately
given, beyond the 'bare thatness', and involves an implicit refer-
ence to other sensations, which I am not having at the present
instant, and to time (which, according to Kant, is not something
noumenal but rather a form in which we arrange the 'things-for-
us').7 Whether the sensations I have at different times that I clas-
sify as sensations of red are all 'really' (noumenally) similar is a
question that makes no sense; if they appear to be similar (e.g. if
I remember the previous sensations as similar to this one, and
anticipate that future sensations which I will so classify will in
their turn seem to be similar to this one, as this one is then remem-
bered) then they are similar-for-me.
Kant says again and again, and in different words, that the
objects of inner sense are not transcendentally real (noumenal)
that they are 'transcendentally ideal' (things-for-us), and that
they are no more and no less directly knowable than so-called
Here I am being deliberately anachronistic and describing Kant's view by

means of an example taken from Wittgenstein's Philosophical
Investigations. But Wittgenstein's example has deeply Kantian roots:
Hegel, writing shortly after Kant, and aware of Kant's doctrine, made
precisely the point that any judgment, even of sense impression, has to
go beyond what is 'given' to be a judgment at all.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
'external' objects. The sensations I call 'red' can no more be

directly compared with noumenal objects to see if they have the
same noumenal property than the objects I call 'pieces of gold'
can be directly compared with noumenal objects to see if they
have the same noumenal property.
The reason that 'All properties are secondary' is only a first
approximation to Kant's view is this: 'All properties are secon-
dary' (i.e. all properties are Powers) suggests that saying of a
chair that it is made of pine, or whatever, is attributing a Power
(the disposition to appear to be made of pine to us) to a nou-
menal object; saying of the chair that it is brown is attributing a
different Power to that same noumenal object; and so on. On
such a view there would be one noumenal object corresponding
to each object in what Kant calls 'the representation', i.e. one
noumenal object corresponding to each thing-for-us. But Kant
explicitly denies this. This is the point at which he all but says
that he is giving up the correspondence theory of truth.
Kant does not, indeed, say he is giving up the correspondence
theory of truth. On the contrary, he says that truth is the 'cor-
respondence of a judgment to its object'. But this is what Kant
called a 'nominal definition of truth'. On my view, identifying
this with what the metaphysical realist means by 'the correspon-
dence theory of truth' would be a grave error. To say whether
Kant held what a metaphysical realist means by 'the correspon-
dence theory of truth' we have to see whether he had a realist
conception of what he called 'the object' of an empirical judg-
ment.
On Kant's view, any judgment about external or internal
objects (physical things or mental entities) says that the nou-
menal world as a whole is such that this is the description that a
rational being (one with our rational nature) given the informa-
tion available to a being with our sense organs (a being with our
sensible nature) would construct. In that sense, the judgment
ascribes a Power. But the Power is ascribed to the whole nou-
menal world; you must not think that because there are chairs
and horses and sensations in our representation, that there are
correspondingly noumenal chairs and noumenal horses and nou-
menal sensations. There is not even a one-to-one correspondence
between things-for-us and things in themselves. Kant not only
gives up any notion of similitude between our ideas and the
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
things in themselves; he even gives up any notion of an abstract

isomorphism. And this means that there is no correspondence
theory of truth in his philosophy.
What then is a true judgment? Kant does believe that we have
objective knowledge: we know laws of mathematics, laws of
geometry, laws of physics, and many statements about individ-
ual objects — empirical objects, things for us. The use of the term
'knowledge' and the use of the term 'objective' amount to the
assertion that there still is a notion of truth. But what is truth if
it is not correspondence to the way things are in themselves?
As I have said, the only answer that one can extract from
Kant's writing is this: a piece of knowledge (i.e. a 'true state-
ment') is a statement that a rational being would accept on suf-
ficient experience of the kind that it is actually possible for
beings with our nature to have. 'Truth' in any other sense is
inaccessible to us and inconceivable by us. Truth is ultimate
goodness of fit.
The empiricist alternative

So far as our argument has gone, it is still possible for a philos-
opher to avoid giving up the correspondence theory of truth
and the similitude theory of reference by restricting them to sen-
sations and images. And many philosophers continued to believe
even after Kant that similitude is the mechanism by which we
are able to have ideas that refer to our own (and, although this
was more controversial, other people's) sensations, and that this
is the primary case of reference from an epistemological point of
view.
To see why this doesn't work, recall that the heart of Berke-
ley'^s argument was the contention that nothing can resemble an
'idea' (sensation or image) except another 'idea', i.e. there can be
no resemblance between the mental and the physical. Our ideas
can resemble other mental entities, but they cannot resemble
'matter', according to Berkeley.
At this point, we must stop and realize that this is in an impor-
tant way false. In fact, everything is similar to everything else in
infinitely many respects. For example, my sensation of a type-
writer at this instant and the quarter in my pocket are both sim-
ilar in the respect that some of their properties (the sensation's
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
occurring right now and the quarter's being in my pocket right

now) are effects of my past actions; if I had not sat down to type,
I would not be having the sensation; and the quarter would/not
be in my pocket if I had not put it there. Both the sensation and
the quarter exist in the twentieth century. Both the sensation and
the quarter have been described in English. And so on and so on.
The number of similarities one can find between any two objects
is limited only by ingenuity and time.
In a particular context, 'similarity' may have a more restricted
meaning, of course. But to just ask 'are A and B similar?' when
we have not specified, explicitly or implicity, what kind of simi-
larity is at issue, is to ask an empty question.
From this simple fact it already follows that the idea that
similitude is the private mechanism of reference must lead to an
infinite regress. Suppose, to use an example due to Wittgenstein,
someone is trying to invent a 'private language', a language
which refers to his own sensations as they are directly given to
him. He focusses his attention on a sensation X and introduces
a sign E which he intends to apply to exactly those entities which
are qualitatively identical with X. In effect, he intends that E
should apply to all and only those entities which are similar to
X.
If this is all he intends — if he does not specify the respect in
which something has to be similar to X to fall under the classi-
fication E — then his intention is empty, as we just saw. For
everything is similar to X in some respect.
If, on the other hand, he specifies the respect; if he thinks the
thought that a sensation is E if and only if it is similar to X in
respect R; then, since he is able to think this thought, he is
already able to refer to the sensations for which he is trying to
introduce a term E, and to the relevant property of those sensa-
tions! But how did he get to be able to do this? (If we answer,
'By focussing his attention on two other sensations, Z, W, and
thinking the thought that two sensations are similar in respect R
if and only if they are similar to Z, W, then we are involved in
a regress to infinity.)
The difficulty with the similitude theory of reference is the
same as the difficulty with the 'causal chain of the appropriate
kind' theory that we mentioned earlier. If I just say, 'The word
"horse" refers to objects which have the property whose occur-
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
rence causes me on certain occasions to produce the utterance

"there is a horse in front of me" ', then one difficulty is that there
are too many such properties. For example, let H-A (for 'Horse
Appearance') be that property of total perceptual situations
which elicits the response 'there is a horse in front of me' from a
competent normal speaker of English. Then the property H-A is
present when I say 'there is a horse in front of me' (even when I
am experiencing an illusion), but 'horse' does not refer to situa-
tions with that property, but rather to certain animals. The pres-
ence of an animal with the property of belonging to a particular
natural kind and the presence of a perceptual situation with the
property H-A are both connected to my utterance 'There is a
horse in front of me' by causal chains. In fact, the occurrence of
horses in the Stone Age is connected with my utterance 'There is
a horse in front of me' by a causal chain. Just as there are too
many similarities for reference to be merely a matter of similari-
ties, so there are too many causal chains for reference to be
merely a matter of causal chains.
On the other hand, if I say 'the word "horse" refers to objects
which have a property which is connected with my production
of the utterance "There is a horse in front of me" on certain
occasions by a causal chain of the appropriate type\ then I have
the problem that, if I am able to specify what is the appropriate
type of causal chain, I must already be able to refer to the kinds
of things and properties that make up that kind of causal chain.
But how did I get to be able to do this?
The conclusion is not that there are no terms which have the
logic ascribed by the similitude theory, any more than the con-
clusion is that there are no terms which refer to things which are
connected to us by particular kinds of causal chains. The conclu-
sion is simply that neither similitude nor causal connection can
be the only, or the fundamental, mechanism of reference.
Wittgenstein on 'following a rule'

Consider the example I mentioned in passing, of the man who
attempts to specify the respect R (the respect in which sensations
must be similar to X if they are to be correctly classified as E) by
saying or thinking that two things are similar in the respect R
just in case they are similar in just the way Z, W are similar. This
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
fails, of course, because any two things Z, W are themselves sim-

ilar in more than one way (in fact, in infinitely many ways).
Trying to specifiy a similarity relation by giving finitely many
examples is like trying to specify a function on the natural num-
bers by giving its first 1,000 (or 1,000,000) values: there are
always infinitely many functions which agree with any given
table on any finite set of values, but which diverge on values not
listed in the table.
This is connected with another point that Wittgenstein makes
in Philosophical Investigations and that was mentioned at the
end of Chapter 1. Whatever introspectible signs or 'presenta-
tions' I may be able to call up in connection with a concept can-
not specify or constitute the content of the concept. Wittgenstein
makes this point in a famous section which concerns 'following
a rule' — say, the rule 'add one'. Even if two species in two pos-
sible worlds (I state the argument in most un-Wittgensteinian
terminology!) have the same mental signs in connection with the
verbal formula 'add one', it is still possible that their practice
might diverge; and it is the practice that fixes the interpretation:
signs do not interpret themselves, as we saw. Even if someone
pictures the relation 'A is the successor of JET (i.e. A =B + 1) just
as we do and has agreed with us on some large finite set of cases
(e.g. that 2 is the successor of 1, 3 is the successor of 2, . . .,
999,978 is the successor of 999,977), still he may have a diver-
gent interpretation of 'successor' which will only reveal itself in
some future cases. (Even if he agrees with us in his 'theory' - i.e.
what he says about 'successor of; he may have a divergent inter-
pretation of the whole theory, as the Skolem—Lowenheim
Theorem shows.)
This has immediate relevance to philosophy of mathematics,
as well as to philosophy of language. First of all, there is the
question of finitism: human practice, actual and potential,
extends only finitely far. Even if we say we can, we cannot 'go
on counting forever'. If there are possible divergent extensions
of our practice, then there are possible divergent interpretations
of even the natural number sequence — our practice, or our men-
tal representations, etc., do not single out a unique 'standard
model' of the natural number sequence. We are tempted to think
they do because we easily shift from 'we could go on counting'
to 'an ideal machine could go on counting' (or, 'an ideal mind
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
could go on counting'); but talk of ideal machines (or minds) is

very different from talk oi actual machines and persons. Talk of
what an ideal machine could do is talk within mathematics, it
cannot fix the interpretation of mathematics.
In the same way, Wittgenstein holds that talk of 'similarity'
and 'the same sensation' or 'the same experience' is talk within
psychological theory; it cannot fix the interpretation of psycho-
logical theory. That, the interpretation of psychological theory
and terminology, is fixed by our actual practice, our actual stan-
dards of correctness and incorrectness.
In Ways of Worldmaking8 Nelson Goodman makes a closely
related point: it is futile to try to have a notion of what the per-
ceptual facts 'really are' independently of how we conceptualize
them, of the descriptions that we give of them and that seem
right to us. Thus, after discussing a finding by the psychologist
Kolers that a disproportionate number of engineers and physi-
cians are unable to see apparent motion at all, that is 'motion'
produced by lights which successively flash at different positions,
Goodman comments (p. 92):
Yet if an observer reports that he sees two distinct flashes,
even at distances and intervals so short that most
observers see one moving spot, perhaps he means that he
sees the two as we might say we see a swarm of molecules
when we look at a chair, or as we do when we say we
see a round table top even when we look at it from
an oblique angle. Since an observer can become adept at
distinguishing apparent from real motion, he may take
the appearance of motion as a sign that there are two
flashes, as we take the oval appearance of the table top as
a sign that it is round; and in both cases the signs may
be or become so transparent that we look through them
to physical events and objects. When the observer visually
determines that what is before him is what we agree is
before him, we can hardly charge him with an error
in visual perception. Shall we say, rather, that he mis-
understands the instruction, which is presumably just
to tell what he sees? Then how, without prejudicing
the outcome, can we so reframe the instruction as to
8
Published by Hackett, 1978.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
prevent such a 'misunderstanding'? Asking him to make

no use of prior experience and to avoid all concept-
ualization will obviously leave him speechless; for to
talk at all he must use words.
Grasp of 'Forms' and empirical association

A Platonist or Neo-Platonist of an antique vintage would have
dealt with this issue in a much simpler way. Such a philosopher
would have said that when we attend to a particular sensation
we also perceive a Universal or a Form, i.e. the mind has the
ability to grasp properties in themselves, and not just to attend
to instances of those properties. Such a philosopher would say it
is the Nominalism of Wittgenstein and Goodman, their refusal
to have any truck with Forms and with the direct grasp of
Forms, that makes it seem to them that there is any problem
with the similitude theory.
While just positing a mysterious power of 'grasping Forms' is
hardly a solution, it might seem that an analogue of this power
is available to us. Properties of things do enter into causal expla-
nations; when I have a sensation and it elicits the response 'this
is a sensation of red', my response is partly caused by the fact
that the sensation had a property. True, some philosophers are
so nominalistic that they would deny the existence of such enti-
ties as 'properties' altogether; but science itself does not hesitate
to talk freely of properties. Can we not say that, when Wittgen-
stein's privateer (the man who wanted to invent a private lan-
guage) attended to X and said *E3 then what caused the response
'E' was a causal interaction involving a certain property, and
that property (whatever it was) is the relevant 'similarity' that
other sensations must have to X to be correctly classified as E?
The observation that talk of 'properties' is perfectly scientifi-
cally legitimate is correct; but this does not help rehabilitate Pla-
tonism. We interact with properties only by interacting with
their instances; and these instances always are instances oimany
properties at the same time. There is no such thing as just inter-
acting with a property 'in itself. Talk of the properties causally
associated with a sensation cannot do the work that the notion
of the (unique) Form of the sensation did in Platonistic philoso-
phy.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
To spell this out: when I have a sensation of blue, I have a

sensation of blue, and I also have a sensation with the complex
property of being such as to be classed by me at that instance
under that particular verbal label. Merely attending to this sen-
sation does not constitute 'grasping' one of these properties. To
pick out the property associated in just one of these ways with
my sensation or with the verbal label is our old friend, the prob-
lem of the Causal Chain of the Appropriate Type again.
To see this, observe, first of all, that when my total perceptual
experience elicits the response 'I am having the sensation of
blue', I am not always right. I myself have had the experience of
referring to 'the man in the blue sweater' two or three times
before someone pointed out that the sweater was green. I don't
mean the sweater looked blue; I realized that I had been misde-
scribing the sweater the instant the other person spoke. (I don't
often have occasion to say 'I am having the sensation of blue',
but if I did, then in such a case I would probably have said it two
or three times until someone — wondering, perhaps, why I would
have the sensation of blue when I was looking at something that
was obviously green - queried me, whereupon I would have
taken back my previous phenomenal report.) This already shows
that the property of eliciting the report 'I am having the sensa-
tion of blue', or whatever, is not the same property as the prop-
erty of being a sensation of blue, or a sensation of whatever the
relevant quality might be.
Philosophers often refer to such a case as a 'slip of the tongue'.
This seems to me to be an unfortunate terminology. The word
'green' might have been on my lips, and I might have found
myself, frustratingly, saying 'blue'. That would have been a slip
of the tongue. But in the case I described I didn't even notice I
was misdescribing until someone questioned my report (and
might never have noticed otherwise).
Another explanation which is suggested is that when I said
'blue' I meant green. By now it should be clear that when we say
things we don't go around 'meaning' things in the sense of hold-
ing meanings in mind. To say I 'meant' green is just to say that
I instantly accepted the correction (and felt funny when I realized
the way I had been speaking). This is just to repeat what hap-
pened, not to explain it.
Whatever the explanation may be (perhaps some slip-up in the
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
verbal processing unit of my brain), the point is that, just as the

property A-H described a few pages back will elicit the report
There is a horse in front of me' even on occasions when no horse
is present in the environment, so there is a complex property of
my total mind-set which will elicit 'I am having a sensation of
blue', when I am not having the sensation of blue (or, anyway,
would deny that I was if I were queried). No mechanism of
empirical association is perfect. If we decide to stipulate that I
am having a sensation of blue whenever I am having a sensation
which elicits that report (or which elicits that report and is such
that the report does not seem 'wrong' to me on second thought),
then on folk psychological theory, and perhaps on scientific psy-
chological theory as well, there could be occasions when it will
be true that I am having a sensation of blue by this criterion
although, for one of a variety of reasons, the quality of the sen-
sation is not blue. Moreover, as Wittgenstein puts it, on such a
criterion, whatever seems right to me is going to be right - i.e.
the distinction between making a report of my sensation that
really is correct and making a report that seems to me to be
correct will have been abandoned. Perhaps we should abandon
or at least qualify it; perhaps, as Goodman seems to be suggest-
ing, the question of whether one is 'really' having the kind of
sensation one thinks one is makes no sense, apart from special
cases, such as the case in which one would take the report back
if queried; but to abandon this distinction is not a possible move
for a metaphysical realist, for the sharp distinction between what
really is the case and what one judges to be the case is precisely
what constitutes metaphysical realism.
Could one always be wrong about the quality of one's past

sensations?
Another way to bring out what is involved is to consider the
question: 'Could one always be wrong about one's past sensa-
tions?' On the similitude theory, the answer is clearly 'yes'. For
according to that theory, my previous sensations either are or
aren't similar to the sensations I now describe by the various
verbal labels 'sensation of red', 'pain', etc., and whether they are
or aren't is a totally different question from whether I then class-
ified them under those same verbal labels. Perhaps the world is
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
such that what we call a 'sensation of red' at an even numbered

minute from the beginning of the Christian Era is actually simi-
lar in quality to what we call a 'sensation of green' at an odd-
numbered minute, but our memory always deceives us in such a
way that we never notice. Then the sensation I classified under
the verbal label 'sensation of red' one minute ago would not be
similar to the sensation I now classify under that same label.
There is something very odd about this alleged possibility,
however. For one thing, the sense in which 'I would never notice'
is very strong: if I treat my 'sensations of red' at different times
as reliable signs of the various correlated physical occurrences
(such as fire, the signal to stop, etc.) then I will be successful in
all my actions. The 'wrong' similarity class (the class that lumps
together the sensations I call sensations of red, in spite of the fact
that they are not 'really' all of the same 'quality') would be the
one that I had better use in connection with my problem-solving
activities. But then is it really the wrong similarity class?
If we don't suppose that the notion of similarity is self-
interpreting, then this case could be redescribed as a case in
which the relation called 'similarity' by the external observer
who is telling us about the case simply differs from the relation
called 'similarity' by us. If we take this view, then the hypothesis
that we are 'really' wrong about our past sensations collapses:
from an internalist point of view there is no intelligible notion of
sensations at different times being 'similar' apart from our stan-
dards of rational acceptability.
The correspondence theory of truth again

By now the reader may be convinced that the similitude theory
of reference is thoroughly dead. But why should we conclude
that the correspondence theory of truth must be given up? Even
if the notion of a 'similarity' between our concepts and what
they refer to doesn't work, couldn't there be some kind of an
abstract isomorphism, or, if not literally an isomorphism, some
kind of abstract mapping of concepts onto things in the (mind-
independent) world? Couldn't truth be defined in terms of such
an isomorphism or mapping?
The trouble with this suggestion is not that correspondences
between words or concepts and other entities don't exist, but
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
that too many correspondences exist. To pick out just one cor-
respondence between words or mental signs and mind-
independent things we would have already to have referential
access to the mind-independent things. You can't single out a
correspondence between two things by just squeezing one of
them hard (or doing anything else to just one of them); you can-
not single out a correspondence between our concepts and the
supposed noumenal objects without access to the noumenal
objects.
One way to see this is the following. Sometimes incompatible
theories can actually be intertranslatable. For example, if New-
tonian physics were true, then every single physical event could
be described in two ways: in terms of particles acting at a dis-
tance, across empty space (which is how Newton described grav-
itation as acting), or in terms of particles acting on fields which
act on other fields (or other parts of the samefield),which finally
act 'locally' on other particles. For example, the Maxwell equa-
tions, which describe the behavior of the electro-magnetic field,
are mathematically equivalent to a theory in which there are
only action-at-a-distance forces between particles, attracting and
repelling according to the inverse square law, travelling not
instantaneously but rather at the speed of light ('retarded poten-
tials'). The Maxwell field theory and the retarded potential the-
ory are incompatible from a metaphysical point of view, since
either there are or there aren't causal agencies (the 'fields') which
mediate the action of separated particles on each other (a realist
would say). But the two theories are mathematically intertrans-
latable. So if there is a 'correspondence' to the noumenal things
which makes one of them true, then one can define another cor-
respondence which makes the other theory true. If all it takes to
make a theory true is abstract correspondence (never mind
which), then incompatible theories can be true.
To an internalist this is not objectionable: why should there
not sometimes be equally coherent but incompatible conceptual
schemes which fit our experiential beliefs equally well? If truth
is not (unique) correspondence then the possibility of a certain
pluralism is opened up. But the motive of the metaphysical real-
ist is to save the notion of the God's Eye Point of View, i.e. the
One True Theory.
Not only may there be correspondence between objects and
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
(what we take to be) incompatible theories (i.e. the same objects

can be what logicians call a 'model' for incompatible theories),
but even if we fix the theory and fix the objects there are (if the
number of objects is infinite) infinitely many different ways in
which the same objects can be used to make a model for a given
theory. This simply states in mathematical language the intuitive
fact that to single out a correspondence between two domains
one needs some independent access to both domains.
What we have is the demise of a theory that lasted for over
two thousand years. That it persisted so long and in so many
forms in spite of the internal contradictions and obscurities
which were present from the beginning testifies to the natural-
ness and the strength of the desire for a God's Eye View. Kant,
who first taught us that this desire is unfulfiUable, thought that it
was nonetheless built into our rational nature itself (he suggested
sublimating this 'totalizing' impulse in the project of trying to
realize 'the highest good in the world' by reconciling the moral
and empirical orders in a perfected system of social institutions
and individual relationships). The continued presence of this
natural but unfulfiUable impulse is, perhaps, a deep cause of the
false monisms and false dualisms which proliferate in our cul-
ture; be this as it may, we are left without the God's Eye View.
Cambridge
https://doi.org/10.1017/CBO9780511625398.005
Mind and body
Parallelism, interactionism, identity

In the seventeenth century the great philosophers Descartes, Spi-
noza, and Leibniz all realized that there was a serious problem
about the relation of mind to material body. To some extent, the
relation was already a problem for Plato, of course, and for all
of the philosophers that came after; but it became much more of
a problem with the rise of modern physics. In the seventeenth
century, people became aware that the physical world is strik-
ingly causally closed. The way in which it is causally closed is
best expressed in terms of Newtonian physics: no body moves
except as the result of the action of some force. Forces can be
completely described by numbers: three numbers suffice to
determine the direction, and one number suffices to describe the
magnitude of any force. The acceleration produced by a force
has exactly the same direction as the force, and the magnitude of
the acceleration can be deduced from the mass of the body and
the magnitude of the force according to Newton's First Law,
F = ma. When more than one force acts on a body, the resultant
force can be computed by the parallelogram law.
It is important to recognize how very different such a physics,
stressing number and precise algorithms for computation as it
does, is from the essentially qualitative thinking of the middle
ages. In medieval thought almost anything could exert an 'influ-
ence' on anything else. (Our word 'influenza' is a survival of this
medieval way of thinking. Evil spirits were thought to exert an
influence — questa influenza, in Italian — on the air which in turn
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
76 Mind and body
influenced the sufferers of the illness.) In such a way of thinking,

it is not so surprising that mind can 'influence' body.
In the time of the philosophers I mentioned, the mathematical
way of thinking was beginning to appear and to push aside this
older way of thinking. The new way of thinking did not fully
develop until Newton, but in special cases Descartes already had
the parallelogram of forces, and in still more primitive cases
Leonardo da Vinci already had it. These thinkers saw that phys-
ics could be done in something like the way it is done now. They
saw that what physics deals with is force and motion, and they
rejected the qualitative style of explanation. Rather, they con-
ceived that the mechanical world had a logic of its own, had a
'program', as we would say, and that it followed that program
unless something disturbed it.
It seemed to these thinkers that mental events could do one of
two things. (1) They could parallel physical events, e.g. events in
the brain. The model is a pair of synchronized clocks: the body
is a clock which has been wound up and which runs its happy or
unhappy way until death, and likewise the entire physical world
runs its happy or unhappy way from creation to the Last Judg-
ment (or to gravitational collapse, in a modern version). And the
mental events run their happy or unhappy way, and somehow,
perhaps by divine providence, it has been arranged so that brain
event B will always occur just when sensation S is occurring. (2)
They could interact with physical events. The mental events
might actually be causing brain events, and vice versa.
Descartes' rather notorious form of the interactionist view, the
suggestion that the mind can influence matter when the matter
is very, very ethereal (and that, in fact, it in some way pushes the
matter in the pineal gland), was less the crazy speculation that it
might seem to be, and more a hangover from a set of medieval
doctrines.1 In the earlier way of thinking, the mind was thought
of as acting on the 'spirit' which in turn acted on 'matter', and
spirit was not thought of as totally immaterial. 'Spirit' was just
the in-between sort of stuff that the medieval philosophers' ten-
dency to introduce in-betweens between any two adjacent terms
in the series of kinds of being naturally led them to postulate. It
See The Discarded Image, by C. S. Lewis, especially Chapter VII, sec. F,

for a description of the medieval view (Cambridge, 1964).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 11
was like a gas with just a little bit of push. As soon as 'spirit' is
dropped out, and the mind is really thought of as totally imma-
terial, then the push of the mind on even very ethereal matter in
the pineal gland appears very strange. One can't quite visualize
that.
The most naive version of the interactionist view conceives of
the mind as a sort of ghost, capable of inhabiting different bodies
(but without change in the way it thinks, feels, remembers, and
exhibits personality, judging from the spate of popular books
about reincarnation and 'remembering previous lives') or even
capable of existing without a body (and continuing to think, feel,
remember, and exhibit personality). This version, which
amounts to little more than superstition, is vulnerable to the
objection that there is enormous evidence (some of which was
already known in the seventeenth century) that the functions of
thought, feeling, and memory involve the brain in an essential
way. Indeed, on such a version it is not clear why we should
have complicated brains at all. If all that is needed is a 'steering
wheel', that could be a lot smaller than the human brain.
To avoid such scientific objections, sophisticated interaction-
ists such as Descartes maintained that the mind and the brain are
an essential unity. In some way it is the mind—brain unity that
thinks, feels, remembers, and exhibits personality. This means
that what we ordinarily call the mind is not the mind at all, but
the mind-brain unity. What this doctrine means, what it means
to say that something can consist of two substances as different
as mind and matter are supposed to be and still be an essential
unity, is, however, very obscure.
The parallelist alternative is also very strange. What makes
the mental event accompany the brain event? One daring
seventeenth-century philosopher suggested that mental events
might actually be identical with brain events and other physical
events, and that was Spinoza. The suggestion in a contemporary
form is that the event of my being in pain on a particular occa-
sion might be the same event as the event of my brain being in
some state B on that occasion. (I will also express this view by
saying that, on such a view, the properties of having that partic-
ular sort of pain and being in brain state B are identical. I prefer
to talk in this way because I think we have more of a logical
theory of properties at the present time than we do of events, but
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
78 Mind and body
I think the idea can be couched in either way. The idea, in this
terminology, is that the property of the person, that the person
is experiencing sensation Q, could be the same property as the
property of being in brain state B.) In this form the suggestion
was put forward by Diderot, for example, in the eighteenth cen-
tury, and became 'mainstream' in the 1940s and 1950s. Materi-
alism and the identity theory began to be taken seriously for the
first time, and the suggestion began to be advanced that some-
thing like Spinoza's view (or Spinoza's view minus its elaborate
theological and metaphysical embellishments) is right: we are
really dealing with one world, and the fact that we do not know
until we do a great deal of science that the states of having pains,
hearing sounds, experiencing visual sensations, and so on, are in
reality brain states doesn't mean that they can't be.
The first contemporary form of this identity theory was
advanced by several writers, one of the best known being the
Australian philosopher J. J. C. Smart. At first the suggestion was
that a sensation, say, a particular sensation of blue, is identical
with a certain neuro-physiological state. A variant on this, sug-
gested first by myself, I believe, is a view called functionalism.2
On the functionalist view there is indeed an identity here, but
Smart was looking at the wrong sort of brain property to figure
as the other term in the identity. According to the functionalist,
the brain has properties which are in a sense not physical.
Now, what do I mean by saying that the brain has non-
physical properties? I mean properties which are definable in
terms that do not mention the brain's physics or chemistry. If it
seems strange that a system which is physical should have prop-
erties which are not physical, consider a computing machine. A
computing machine has many physical properties. It has a cer-
tain weight, for example; it has a certain number of circuit chips,
or whatever. It has economic properties, such as having a certain
price; and it also has functional properties, such as having a cer-
tain program. Now this last kind of property is non-physical in
the sense that it can be realized by a system quite apart from
what its, as it were, metaphysical or ontological composition
2
N. Block's Readings in Philosophy of Psychology (Harvard 1980)
contains an excellent collection of articles on Functionalism. My own
papers are reprinted as Chapters 14 through 22 of my Mind, Language
and Reality, Philosophical Papers, Vol. 2 (Cambridge, 1975).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 79
might be. A disembodied spirit might exhibit a certain program,

a brain might exhibit a certain program, a machine might exhibit
a certain program and the functional organization of these three,
the disembodied spirit, the brain, the machine, could be exactly
the same even though their matter, their stuff, is totally different.
Psychological properties exhibit the same characteristic; the
same psychological property (e.g. being angry) can be a property
of members of thousands of different species which may have
quite different physics and chemistry (some of these species
might be extraterrestrial; and perhaps robots will someday
exhibit anger). The suggestion of the functionalist is that the
most plausible 'monistic' theory in the twentieth century, the
most plausible theory that avoids treating Mind and Matter as
two separate sorts of substance or two separate realms of prop-
erties, is that psychological properties are identical with func-
tional properties.
Today I am still inclined to think that that theory is right; or
at least that it is the right naturalistic description of the
mind/body relation. There are other, 'mentalistic', descriptions
of this relation which are also correct, but not reducible to the
world-picture we call 'Nature' (indeed the notions of 'rational-
ity', 'truth', and 'reference' belong to such a 'mentalistic' ver-
sion). I shall say something about this later (Chapter 6). This fact
does not dismay me: for, as Nelson Goodman has emphasized,
one of the attractive features of non-realism is that it allows the
possibility of alternative right versions of the world. I am, how-
ever, attracted to the idea that one right version is a naturalistic
version, in which thought-forms, images, sensations, etc. are
functionally characterized physical occurrences; and what I wish
to discuss here is a difficulty with the functionalist theory that
occurred to me some years ago: that is that the theory has diffi-
culty with the qualitative character of sensations. When one
thinks of relatively abstract pure psychological states, e.g. what
we called a 'bracketed' belief, i.e. a thought considered only in
its 'notional' content, or of such diffuse emotional states as being
jealous or being angry, then the identification of these with func-
tional states of the whole system seems very plausible; but when
one thinks of having a presented quality, e.g. experiencing a par-
ticular shade of blue, the identification is implausible.
An example I have used for many years in lectures is a variant
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
80 Mind and body
of the famous example of the 'inverted spectrum'. The inverted

spectrum example (which appears in the writings of Locke3)
involves a chap who walks about seeing things so that blue looks
red to him and red looks blue to him (or so that his subjective
colors resemble the colors on a color negative rather than the
colors on a color positive). One's first reaction on hearing of
such a case might be to say, 'Poor chap, people must pity him.'
But how would anyone ever know? When he sees anything blue,
it looks red to him, but he's been taught to call that color blue
ever since he was an infant, so that if one asked him what color
the object is he would say 'blue'. So no one would ever know.
My variation was the following: imagine your spectrum
becomes inverted at a particular time in your life and you
remember what it was like before that. There is no epistemolog-
ical problem about 'verification'. You wake up one morning and
the sky looks red, and your red sweater appears to have turned
blue, and all the faces are an awful color, as on a color negative.
Oh my God! Now, perhaps you could learn to change your way
of talking, and to call things that look red to you 'blue', and
perhaps you could get good enough so that if someone asked
you what color someone's sweater was you would give the 'nor-
mal' answer. But at night, let us imagine that you would moan,
'Oh, I wish the colors looked the way they did when I was a
child. The colors just don't look the way they used to.'
In this case, it seems that one even knows what must have
happened. Some 'wires' must have gotten 'crossed' in the brain.
The inputs from blue light, that used to go to one mechanism in
the brain, now go to another, and the inputs from red light go to
the first. In other words, something has switched around the
realizations, the physical states. The physical state that formerly
played the functional role of signalling the presence of 'objective'
blue in the environment now signals the presence of 'objective'
red in the environment.
Now suppose we adopted the following 'functionalist' theory
of subjective color: 'a sensation is a sensation of blue (i.e. has
the qualitative character that I now describe in that way) just in
3
An Essay Concerning Human Understanding, Book II, Chapter 32 (sec.
14).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 81
case the sensation (or the corresponding physical event in the

brain) has the role of signalling the presence of objective blue in
the environment'. This theory captures one sense of the phrase
'sensation of blue', but not the desired 'qualitative' sense. If this
functional role were identical with the qualitative character, then
one couldn't say that the quality of the sensation has changed.
(If this is not clear, then imagine that after the spectrum inver-
sion, and after learning to compensate for it linguistically, you
experience an attack of amnesia which wipes out all memory of
what colors used to look like. In this case it would seem as if the
sensation you are now calling a 'sensation of blue' could have
almost exactly the functional role that the sensation you used to
call the 'sensation of blue' used to have, while having a totally
different character.) But the quality has changed. The quality
doesn't seem to be a functional state in this case.
It seems to me that the most plausible move for a functionalist
to make if such cases are really possible is to say, 'Yes, but the
"qualitative character" is just the physical realization.' And to
say that for this special kind of psychological property, for qual-
ities, the older form of the identity theory was the right one. If
the reader is fairly materialistically inclined, he or she probably
thinks that the property of having the sensation is a brain prop-
erty. Readers who are not materialistically inclined probably
think that the property of having the sensation is correlated with
a brain state. Probably most people hold one of these two views:
the view that sensation-states are correlated with brain-states, or
the view that sensation-states are identical with brain-states. As
so often happens, the question becomes debated over and over
in the same way. The way it is always discussed is, 'given that B
is correlated with Q, is B actually identical with Q ?' We know that
this sensation-state parallels this brain-state, is it or is it not the
case that 'the sensation-state is identical with the brain-state?'
The more the discussion goes on in that way, the more the
concept of correlation comes to seem unproblematical. Correla-
tion isn't (much) discussed because everyone knows that there is
at least a correlation. Identity is discussed because that is what
is problematical. But I am going to try to show you that even
correlation is problematical, not in the sense that there is evi-
dence of non-correlation, but in the epistemological sense that //
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
82 Mind and body
there is a correlation, one can never know which it is. The prob-
lem will not depend on assuming materialism, but it will depend
upon the fact that we think that there is at least a correlation.
Identity theory and the a priori

What made the revival of interest in the identity theory and other
'monist' theories possible, if not initially (not with Smart and
some of the early identity theorists), at least starting around
1960, was the change in the epistemological climate. Identity
theory was not taken seriously prior to the 1960s for the reason
that philosophers 'knew' it was false. And they thought they
knew it was false not on the basis of empirical evidence (for what
sort of empirical evidence could show that a sensation-state is
not a brain-state?), buttf priori. One thinks about it and one just
sees a priori that a sensation-state couldn't be a brain-state, or
perhaps one sees that it is meaningless to say that a sensation-
state is a brain-state in the way in which it is meaningless to say
that the number three is blue. Prior to 1950 or 1960 people
thought they just knew, or many people thought they knew, that
sensation-states can't be physical. Other people thought they
knew those people were wrong. But argument was impossible.
The majority would say, 'Look, we can't prove to you that it is
impossible for a sensation-state to be a neurophysiological state,
we can't prove to you that every number has a successor, we
can't prove to you that the number three is not blue, but these
are things we just know; these are truths of reason. We know
that it is nonsense or an impossibility for a sensation-state to
be a neuro-physiological state as clearly as we know anything.'
One had the majority that knew that sensation-states couldn't
be brain-states and a minority that knew the majority was
wrong. Each knew the other was wrong a priori. And there was
no really significant possibility of argument or movement from
this frozen state in the debate.
In 1951 W. V. Quine published a paper titled 'Two Dogmas
of Empiricism'.4 From that time on, there has been a steady ero-
sion in philosophical confidence in the notion of an 'a priori'
4
'Two Dogmas of Empiricism' first appeared in The Philosophical
Review, 1951. It is reprinted in Quine's From a Logical Point of View.
(New York, 1961).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 83
truth. Quine pointed out that many things we thought we knew

a priori have had to be revised. Thus, consider the following:
suppose someone had suggested to Euclid that this could hap-
pen: that one could have two straight lines which are perpendic-
ular to a third straight line and which meet. Euclid would have
said that it was a necessary truth that this couldn't happen.
According to the physical theory we accept today, it does hap-
pen. Light passing near the sun behaves as it does not because
the light travels in curved lines, but because the light continues
to travel in straight lines and the straight lines behave in that
way in our non-Euclidean world.
Once we accept that, then some philosopher was bound to ask
the question, 'What is left of the a priori*', and Quine did.
(Quine also showed convincingly that the standard empiricist
accounts of a priority - e.g. the notion of 'truth by convention' -
were incoherent, but I shall not review his arguments.)
In some ways, I think, Quine went too far. Quine's assertion
that 'no statement is immune from revision' suggests that for
every statement there are circumstances under which it would be
rational to reject it. But this is pretty clearly false: under what
circumstances, after all, would it be rational to reject 'Not every
statement is true', i.e. to accept 'All statements are true'?5
But if Quine does overstate the case against the a priori, what
he is nonetheless right about is this: our notions of rationality
and of rational revisability are not fixed by some immutable
book of rules, nor are they written into our transcendental
natures, as Kant thought, for the very good reason that the
whole idea of a transcendental nature, a nature that we have
noumenally, apart from any way in which we can conceive of
ourselves historically or biologically, is nonsensical. Since our
notions of rationality and of rational revisability are the product
of our all too limited experience and all too fallible biology, it is
to be expected that even principles we regard as 'a priori', or
'conceptual', or whatever, will from time to time turn out to
need revision in the light of unexpected experiences or unantici-
pated theoretical innovations. Such revision cannot be unlim-
ited: otherwise we would no longer have a concept of anything
5
I discuss Quine's attack on the notion of the a priori in 'Analyticity and
Apriority: Beyond Wittgenstein and Quine', Midwest Studies in
Philosophy, Vol. IV, 1979 (Minnesota).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
84 Mind and body
we could call rationality; but the limits are not in general possi-
ble for us to state. Apart from trivial cases (e.g. 'Not every state-
ment is true') we cannot be sure that it would never be rational
in any context to give up a statement that is regarded (and legit-
imately so, in a given context) as a 'necessary' truth. In general,
we have to admit that considerations of simplicity, overall util-
ity, and plausibility may lead us to give up something that was
formerly regarded as a priori, and that this is reasonable. Philo-
sophy has become anti-aprioristic. But once we have recognized
that most of what we regard as a priori truth is of a contextual
and relative character, we have given up the only good 'argu-
ment' there was against mind—body identity. Identity theorists
were bound to point this out, and they did. So there was a
changed situation.
I have been using the notion of a property; but it seems to me

that there are at least two notions of a 'property' that have
become confused in our minds.6 There is a very old notion for
which the term 'predicate' used to be employed (e.g. in the
famous question, 'Is existence a predicate?'), and there is the
notion that we use today when we speak of 'physical properties',
'fundamental magnitudes', etc. When a philosopher has the
older notion in mind, he frequently regards talk of properties as
interchangeable with talk of concepts. For such a philosopher,
properties cannot be the same unless it is a conceptual truth that
they are the same; in particular, the property of having a sensa-
tion with a certain qualitative character cannot be the same as
the property of being in a certain brain-state, since the corre-
sponding predicates are not synonymous (in the wide sense of
'analytically equivalent'), and the principle of individuation for
predicates is just that being P is one and the same predicate as
being Q just in case 'is P' is synonymous with 'is Q'.
Consider, however, the situation which arises when a scientist
asserts that temperature is mean molecular kinetic energy. On
the face of it, this is a statement of identity of properties. What
is being asserted is that the property of having a particular tem-
perature is really (in some sense of 'really') the same property as
the property of having a certain molecular energy; or (more gen-
6
See 'On Properties', Chapter 19 of my Mathematics, Matter and
Method, Philosophical Papers, Vol. 1 (Cambridge, 1975).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 85
erally) that the physical magnitude temperature is one and the

same physical magnitude as mean molecular kinetic energy. If
this is right, then since (x has such-and-such a temperature' is
not synonymous with 'x has blah-blah mean molecular kinetic
energy', even when 'blah-blah' is the value of molecular energy
that corresponds to the value 'such-and-such' of the tempera-
ture, it must be that what the physicist means by 'physical mag-
nitude' is something quite other than what philosophers have
called a 'predicate' or a 'concept'.
To be specific, the difference is that, whereas synonymy of the
expression fX is P' and fX is Q5 is required for the predicates P
and Q to be the 'same', it is not required for the property P to
be the same as the property Q. Properties, as opposed to predi-
cates, can be 'synthetically identical'.
If there is such a thing as synthetic identity of properties, then

why shouldn't it be the case that the property of being in a cer-
tain brain-state is the same property as the property of having a
sensation of a certain qualitative character (very much in line
with Spinoza's thinking) - even though it is not a conceptual
truth that it is, even though, in fact, it seems to many to be a
priori false? This is the argument that was made. In short, we
had a wave of anti-apriorism, we had the new machinery of the
synthetic identity of properties, and with these two the identity
theorist and in particular the functionalist seem automatically to
be in business.
Now, I want to consider what happens when to these two

things we add a third. What happens if a philosopher is (1) an
anti-aprioristic naturalist, who (2) allows that there is such a
thing as the synthetic identity of properties, and (3) also has a
hard-line realist view of truth? I wish to claim that such a phi-
losopher will find himself confronted with serious epistemologi-
cal difficulties.
Split brains
Let us consider a particular kind of experiment that neurologists
have performed in the last twenty years. This is the famous 'split
brain', or brain disassociation experiment. I want to discuss the
relevance of this kind of experiment to the identity theory and to
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
86 Mind and body
what has so far been taken for granted in the whole discussion,
the notion that there is a correlation.
On the model of the brain as a cognitive system resembling a
computer, the brain has a language, an internal language (which
may be innate, or which may be a mixture of an innate 'lan-
guage', or system of representation, and a public language).
Some philosophers have even invented a name for this hypothet-
ical brain language, 'mentalese'. Let us consider what happens
when one has a visual sensation on such a model (and I shall
make up my neurology, since I don't know enough, but I don't
think anyone really knows enough). Here is one possible story:
When one has a sensation a 'judgment' is made; the brain has
to 'print' something like 'red presented at 12 o'clock'. So the
quality (call it 'Q') corresponds, among other things, to a record
in mentalese. Also, there is an input to the verbal processing cen-
ter, the center which is connected with the voice box, which
accounts for the brain's ability to report in the public language,
'red now'. It may be that the judgment in mentalese has to be
transmitted from one location to another before there is an input
to the speech center. There are also events in the visual cortex
(which have been studied by the neurologists Hubel and Wiesel),
which I am imagining as on the road to the 'record in mentalese',
and the verbal process. These 'records', 'inputs', and other
events may take place in different lobes of the brain: if the cor-
pus collosum is split, the person's right lobe (the lobe that
doesn't have speech) can see red (or at least it will affirmatively
signal in response to a written query visible only to that lobe),
but if one asks the subject what color the card is, he will reply
'I can't see the card.' And, finally, there is at some point the
formation of a memory trace or of memory traces (one could
break this up into short-term memory and long-term memory).
There almost certainly is not a linear causal chain; there are
probably branchings and rejoinings, a causal network.
The problem is that psychology divides up mental events in a
fairly discrete way. Here is a sensation of blue. Now it started;
now it stopped. Causal networks are not discrete. There isn't a
unique physical event which is the correlate of the sensation.
If the identity theory is right, then the sensation-state Q is
identical with some brain-state or other. A metaphysical realist
cannot regard it as in any way a matter of convention or deci-
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 87
sion, or as having a conventional component, which brain-state

Q is identical with. The position is that as a matter of fact we
live in a world in which what we experience as the qualitative
characters of sensations really are one and the same properties
as some of the properties that we encounter in other ways as
physical properties of brain events. (Or better put, in which the
property of having a sensation of a certain qualitative character
is really just the property of being in a certain brain-state.)
Let us stop for a moment and see what the view actually says.
Suppose that red is the subjective quality we're attending to (pro-
duced, say, by staring at a green disk and then removing the disk
to get an afterimage). Suppose that when I experience this red,
the sensation-state I am in is identical to a disjunction of brain
states. It can't be identical to one maximally specified brain state,
because we know that one can take away any one neuron, or
whatever, and one can still have the experience. But the property
might be a disjunctive one, say (implausibly), the even numbered
neurons in area blah-blah are firing or the prime numbered ones
are firing. Actually, it would be a much bigger disjunction than
that. There would be a huge collection of neurological states
such that their disjunction would be the property of experiencing
red.
But now we go a little further. If the even numbered neurons
in area blah-blah are firing, I experience red. If the cerebroscope
says, 'no, the prime numbered neurons in area blah-blah are fir-
ing', I still experience red. That is, I can't tell which of these
brain-states I'm in. If I experience red I have to be in one of
them. But I can't distinguish between them. The even numbered
neurons in area blah-blah are firing is not an observable prop-
erty. Even with the knowledge that the identity theory is true, I
can't tell from my sensations that I have this property. Call this
property T / and call the property that the odd numbered neu-
rons are firing 'P2'. The sensation-state is identical with the dis-
junction (Px or P2), where this is, of course, a third property. Pt
is not a sensation-state, and P2 is not a sensation-state; it is only
their disjunction that is a sensation-state. In other words, in this
ontology, the disjunction of two properties which are themselves
unobservable can be observable. It is a complicated logical func-
tion of unobservable properties that I experience as a simple
given. That is the position.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
88 Mind and body
It may be that I have made the view sound silly. Thus, a friend
of mine has remarked, 'Suppose the only device we have for
detecting muons doesn't distinguish between muons and anti-
muons. Then muon isn't an observable property, and antimuon
isn't an observable property, but the disjunction of them is. This
only seems to be paradoxical to those who take observationality
to be less of a pragmatic notion than it is.' My purpose, however,
is not to ridicule the view, which, indeed, constitutes a very
important and legitimate research program in neurophysiology,
but to make clear what it commits one to. What leads to diffi-
culties, I shall argue, is not the identity theory by itself but the
identity theory taken in conjunction with metaphysical realism -
i.e. taken in conjunction with what I called the 'externalist' per-
spective on the nature of truth.
One can avoid committing oneself to such a perspective. Thus,
Carnap would have said (at least in a certain period) that talk
about physical objects is highly derived talk about sensations,
and that the decision to say that a particular brain state is iden-
tical with a sensation-state Q is really a decision to modify the
language of talk about physical properties in a certain way, to
change our concept of the physical property in question.
Since physical object and physical property talk is only highly
derived talk about sensations, we can modify the rules. But that
standpoint isn't the standpoint of metaphysical realism, at least
with respect to material objects and physical properties. Some-
body who thinks like that might be a metaphysical realist about
sensations, but he is not a metaphysical realist about material
objects, and since he regards material object talk as somewhat
soft, he can adopt the identity theory by simply saying 'I adopt
it as a kind of convention, as a further meaning stipulation.'
Since the meanings were not totally fixed beforehand, since there
was some openness of texture, there is no problem about 'how
can you know that the sensation-state is identical with this prop-
erty and not some other?' If what this property is is somewhat
vague, then we're allowed to simply postulate the identity as a
meaning specification. But I'm talking to someone who really
thinks there is a material world out there, and it is not just highly
derived talk about sensations; who really thinks that there are
physical properties; and who holds that such expressions as 'the
neurons in such and such a channel are firing' predicate definite
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 89
physical properties of us, and either those properties are or

aren't identical to this sensation-state.
Similarly for a philosopher like Daniel Dennett, who thinks
that sensation talk is highly vague, who doesn't think there is a
well defined subjective property of being in this sensation-state,
of having a sensation with this qualitative character. I think he
too could adopt an identity theory as a meaning stipulation, fix-
ing not the meaning of the physical object terms this time, but
the meaning of the psychological terms. But again, that wouldn't
be the position of a full-blown metaphysical realist.
I am considering a full-blown realist who thinks 'yes, I know
what this psychological property (the sensation-state) is. I've had
it. I can recognize it. I think it's a definite psychological property
to which I refer. I know what Pj and P 2 are, therefore what (Px
or P2) is, and either the sensation-state is identical to this or it
isn't.' Just in the way that a naive physicist might say, 'there's no
element of convention' (I think he'd be wrong by the way);
'there's no element of convention in the decision that tempera-
ture is mean molecular kinetic energy, either temperature is
mean molecular kinetic energy, or it's some other property'.
That is the standpoint I want to examine.
The problem is that if one takes this metaphysical realist
standpoint, then there are many more possibilities than people
are wont to consider. The possibility that first comes to mind is
that the sensation-state is identical with the property of having
the appropriate events take place in the visual cortex and having
the 'record' in 'mentalese' appropriately registered and having
the input to the speech center and having the memory traces
formed — i.e. the sensation-state is thought of as identical to the
conjunction of these several properties. But as soon as we con-
sider the possibility of disassociation, then we become unsure
that we really want the whole conjunction. Perhaps the sensation
is just the event in the visual cortex? (I.e. the property of having
the sensation is 'really' the property of having the event take
place in the visual cortex.)
Let us make the assumption that it is, for the moment. Now
let us suppose we can cut off the process that produces the
record in mentalese, or at least cut off the input to the speech
center. Let us imagine that we have shown the subject a red card
on the left side of his visual field (so that the card is only 'visible
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
90 Mind and body
to the right lobe', as neurologists say). The appropriate event in

the visual cortex will then take place in the right lobe, but if we
say to the subject, 'Do you see anything red?', the subject will
say, 'No'.
Now, by one criterion we employ to decide whether or not
someone has a sensation, the criterion of sincere verbal reports,
we should have to say that he didn't have the sensation of red,
and therefore that we have refuted the theory that Q (the rele-
vant qualitative character) is identical with the visual cortex
property in question. But someone could object, 'No, you
haven't refuted that theory at all. Because, what kind of an
observer is this? The chap's brain is cut in two.' As far as any
observer in a normal condition can tell, Q is identical with this
property of the visual cortex. And observers who are not in a
normal condition don't count. They cant count.
The difficulty is that there are identity theories which are
observationally indistinguishable,1 by which I mean that they
lead to the same predictions with respect to the experience of all
observers in a normal condition.
Consider the view that one doesn't have the sensation of red
unless one has the input to the speech center. How could one
prove that or refute it? One might think, if we split the corpus
colossum but there is some memory that doesn't go through the
verbal processing unit, then there is a way; namely, we first ask
the chap whether he has a sensation of red. He says, 'No'. Then
we sew the corpus colossum back together (a neat trick if you
can do it!), and ask, 'Did you have a sensation of red?' He might
say, 'Yes, but it was crazy, you know I had this sensation of red
and you asked me whether I had it, and I heard myself sincerely
7
The notion of 'Observational Indistinguishability' was introduced in
papers on space-time theory by Clark Glymour and David Malament in
Foundations of Space—Time Theories, Earman, Glymour, and Stachel
(eds.), Minnesota Studies in the Philosophy of Science, Vol. VIII
(Minnesota University 1977). The analogous problem in space-time
theory is the existence of 'possible' space-times (i.e. space-times allowed
by relativity theory) which differ in their global topological properties,
but in which observers would have exactly the same experiences. Such
examples are often dismissed on the grounds that 'simplicity
considerations' would tell one which space—time one is living in; the
trouble with this (as Malament points out) is that the physical theory
(general relativity) doesn't say we live in the simplest space—time
compatible with its laws.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 91
telling you I didn't.' (I am told that actually it is more customary

for patients to 'reconcile' or rationalize situations of this kind
than to describe them as I have just imagined.) Would such a
report show that there was a sensation of red without there
being an input to the speech center?
It would not. If Daniel Dennett (who at one time held the view
that the sensation is the input to the speech center, or a view
close to this8) wished to reconcile this subject's report with his
theory, all he would have to say is, 'I don't deny that at the later
time the psychological event of remembering having had the sen-
sation earlier took place. I deny that the sensation took place at
the earlier time.' On either theory the subject later has the expe-
rience of remembering rightly or wrongly that he had the sensa-
tion of red earlier.
The disagreement here is an actual one. Most neurologists do
believe that in the 'split brain' patients the right lobe is 'con-
scious'. In effect, this amounts to saying that there is sometimes
a sensation of red, or whatever, even though there is no input to
the speech center. ('There are two loci of consciousness', is the
way it is frequently put.) At least one famous neurologist, Eccles,
holds, however, that the disassociated right lobe (or left lobe, in
the case of patients who have the speech center on the right) is
not conscious. There is a unitary consciousness on Eccles' view;
that the disassociated right lobe can 'simulate' conscious behav-
ior does not show that it is a second 'locus of consciousness', he
would say.
Nor will it help to appeal to methodological maxims, e.g.
'choose the simpler theory'; for there does not appear to be any
relevant kind of 'simplicity' which is possessed by the 'unitary'
view and lacked by the 'two loci' view, or possessed by the 'two
loci' view and lacked by the 'unitary' view. Perhaps the 'two loci'
view is simpler in one respect; it says that certain behavioral
capacities (which the right lobe possesses, even if it does not pos-
sess speech) are sufficient for consciousness, and this agrees with
the fact that we call animals (who do not possess speech either)
conscious. But there are many dissimilarities between an animal
with an intact brain, whose brain processes are still 'integrated',
8
Dennett's model of consciousness is presented in 'Towards a Cognitive
theory of Consciousness', reprinted in his Brainstorms (Bradford Books,
1978).
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
92 Mind and body
even if they do not involve speech, and one piece of a 'split

brain'. If the case were not one which touches us so nearly, if we
did not have such a strong tendency to metaphysical realism
about sensations, would it not be in keeping with our best meth-
odological intuitions to regard this as a case to be legislated
rather than fought over?
In short, there are a number of observationally indistinguish-
able identity theories. If the identity theorist is right, it would
seem that there is no way on earth in which one could know
which way he is right; know with which brain-state a given
sensation-state is identical {or correlated with). The point is suf-
ficiently important to deserve further illustration.
Thomas Nagel9 has made the plausible claim that one cannot
imagine what it would be like to be a bat. But why should this
be a plausible claim? Some years ago I read a delightful book
about bats by Donald Griffin. I came to realize that bats are not
basically different from any other mammals. We mostly do think
we can imagine what sensations our dogs and cats have. What is
the difficulty with bats?
Well, bats can hear sounds several octaves higher than we can.
I cannot imagine what it would be like to be a bat in the sense of
imagining what the echolocation sensation would be. But need
this be so difficult? I used to be able to hear sounds an octave
higher than the highest sounds I can hear in middle age. But the
subjective pitches have not changed: the highest sounds I can
hear may be an octave lower than the highest sounds I could
hear when I was ten years old in objective pitch, but the highest
sounds I hear now have the same thin, squeaky quality that the
sounds on the threshold of being too high to hear always did for
me. Perhaps that's how a sound five octaves higher than those
we hear sounds to a bat: like a short high squeak.
Now, imagine a debate between two philosophers or psychol-
ogists, one of whom says no bat quale is at all like any human
quale. Bat qualia are unimaginably different from human qualia.
You will never be able to imagine what it feels like to be a bat
(or even a dog or cat). The other philosopher, we may imagine,
replies, 'Nonsense! Perhaps there are some bat sensations I can't
imagine. There are some sensations of other humans (e.g. some
9
'What is it like to be a bat?', reprinted in N. Block, op. cit.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 93
sensations of the other sex) I probably can't imagine, but that

doesn't mean I regard the psychological space of those other
humans as unimaginably different from my own. Why shouldn't
I think of the bat's visual field, for example, as very much like
my visual field? (N.B. Bats see very well, contrary to folklore.)
Allowing for some adjustments for the optics of the bat eye, or
the bat's hearing within the range that overlaps with mine, it's
like my hearing, and its pains are like my pains.' Now could we
settle this?
Because the number of neurons is different, and because the
arrangement is different (the acoustic center of the bat's brain is
enlarged to become 7/8ths of the brain), the properties at the
most completely specified neurological level — number of neu-
rons firing where — which are identical with a quale in the case
of a bat on the assumption that the identity theory is correct,
cannot literally be the same as the properties which are identical
with any quale in the case of a human. Or can they? Suppose
that when a bat has a certain visual sensation (produced by
seeing red objects), that the bat's brain has the disjunctive prop-
erty (Pt or P2), where Px and P2 are maximally specified states of
the bat's brain. (It would really be a much more complicated
disjunctive property with thousands of cases, but let us sim-
plify.) And let us suppose that when I have a certain visual sen-
sation (produced by seeing red objects), my brain has the dis-
junctive property (P/ or P2'). Consider the following two
theories: (1) that the qualitative character of the bat's sensation
(call it, 'reds') is identical with (or at least correlated with) the
disjunctive property (Pj or P2) and the qualitative character of
the human sensation (call it, (redH') is identical with (or at least
correlated with) the different property (P/ or P2'). (2) That the
qualitative character of the bat's sensation is identical with the
qualitative character of my sensation (i.e. redB = redH) and both
are identical with (or correlated with) the more complex disjunc-
tive property (Px or P2 or P / or P 2 ').
On the first theory the bat and I have different experiences,
while on the second we have the same experience; but these two
theories lead to the same predictions with respect to what human
observers, normal and abnormal will experience. Once again,
they are observationally indistinguishable.
Will methodological maxims ('choose the simpler theory')
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
94 Mind and body
help? Once again, it is not clear that they can. Ned Block has
pointed out that the first theory is simpler in one respect (the
quale is identified with a simpler physical property in each case),
but the second is simpler in another respect (the second theory is
'non-chauvinist'; it allows that one doesn't have to have exactly
our physical constitution to have our qualia). And once again,
we lack principles for determining a unique preferred trade-off.
Indeed, what reason is there to think there should or must be
such principles? Why should we not, as Wittgenstein urged we
do, abandon our metaphysical realism about sensations and
about 'same' (as applied to sensations), and treat this too as a
case to be legislated rather than fought over?
Finally, I want to present three theories which I am sure are
false, but which it is difficult or impossible to rule out if meta-
physical realism is right. These are: (1) thatred#is identical with
a functional (or quasi-functional) state after all, namely the state
of being in whatever material (e.g. physical) state earliest in your
life played the functional role of normally signalling the presence
of objective red. (2) that rocks have qualia (i.e. events qualita-
tively similar to, as it might be, visual sensations, take place in
rocks). (3) that nations are conscious.
Let us first consider (1). Recall the argument I used to show
that redH could not be a functional state. That argument was
that if we identified redH with the functional state of being in
whatever material state {e.g. brain-state) normally signals the
presence of objective red, then I would not have undergone a
spectrum inversion (at least in the 'amnesia' case), since I am in
that functional state when I see something objectively red both
before the spectrum inversion and after the spectrum inversion
(allowing time for linguistic adjustment to take place, and, if
necessary, postulating an attack of amnesia). But on a metaphys-
ical realist position it is certainly possible that I have undergone
a spectrum inversion (even though I don't remember it because
of the attack of amnesia). The case is even stronger if I don't
have an attack of amnesia and recall that my spectrum has been
inverted; even in this case, if the linguistic adjustments have
become automatic, there is a sense in which what used to be 'the
sensation of green' now plays the functional role of 'signalling
the presence of objective red in the environment'.
This argument only shows that redH is not identical with the
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 95
functional state mentioned. It does not show that it is not iden-

tical with a more complicated functional state, such as the state
of being in whatever material state earliest in your life realized
the above functional state. One might object that this is a funny
property, a complicated logical function of functional proper-
ties. But why is a complicated logical function of functional
properties less likely to be identical with redH than a disjunction
of complicated physical properties? Does the world prefer dis-
junctions of physical properties to conjunctions of functional
properties?
Let us consider (2). Let P 3 be the property of being a rock, and
consider the hypothesis that redH is identical with the disjunctive
property (Pl or P2 or P / or P2' or P3). Of course, rocks have this
property all the time. So on this hypothesis, events of the quali-
tative character redH are taking place in rocks all the time. (They
are not experiencing red in the functional sense of experiencing
red, but an event of the qualitative character of the event that
plays the functional role of being the sensation of red in us is
taking place in them all the time.) Or consider more complicated
hypotheses on which rocks are having different qualia at differ-
ent times. Or just the hypothesis that some one of these
hypotheses (which not specified) is correct. We might say 'Well,
these hypotheses are crazy.' Yes, they are. But each of them leads
to the same predictions with respect to all human observers as
the 'sane' theory. None of them can be ruled out on observa-
tional or experimental grounds, because each of them is obser-
vationally indistinguishable from the more standard view.
We might think that these theories can be ruled out by an
appeal to the methodological principle that one shouldn't attrib-
ute a property to an object with no reason. Of course this prin-
ciple doesn't say these theories are false (sometimes things we
have no reason to believe are true), but at least it says we are
justified in taking them to be false. But is there really no reason
to hold the least specific of these theories (the theory that some
such theory is correct, and rocks have qualia) ? What of the argu-
ment that if we have qualia and physicalism is true (and many
philosophers think there are many good reasons for accepting
physicalism), then there is at least one physical object in which
events with a qualitative character take place: so why shouldn't
such events take place in all physical objects? If we could show
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
96 Mind and body
that there is something about the quale itself which requires that
it have the particular functional 'role' that it does in the case of
humans then this move would be blocked; but this is just what
believers in qualia as metaphysically real objects tell us we can't
do.
Last but not least, let us consider (3). Consider the hypothesis
that pain is identical with an appropriate functional state which
can be exhibited by either organisms or nations. In other words,
suppose that when the United States announces that 'the United
States is pained by . . .' it really is. We would, of course, never
know. Perhaps the reader is at this moment finding it interesting
and mildly amusing that a group can behave in ways which
resemble the ways in which something that really does feel pain
behaves when it manifests its pain; but the reader does not think
that the United States really feels pain. On this hypothesis, the
reader would be wrong: the national Geist would really be feel-
ing pain.
This hypothesis connects with an interesting discussion in the
philosophy of mind. An argument that functionalists (including
me) like to employ is the following 'anti-chauvinism' argument:
in principle, the differences between a robot and a human (in
functional organization, anyway) could be reduced to small
details of the physics and chemistry. One might even have a
robot that corresponded to us down to the neuron level. (It could
even have a 'flesh and blood' body, apart from the brain.) The
difference would be that whereas we have neurons made of car-
bon and hydrogen and proteins and so on, it would have neu-
rons made of electronics, but from the neuron level up all the
circuitry would exactly correspond. Now, unless you are a
'hydrogen-carbon chauvinist' who thinks that carbon and
hydrogen are intrinsically more conscious, why shouldn't you
say that this robot is a person whose brain happens to have more
metal in it and less hydrogen and carbon?
This argument has provoked the following reply: 'Well,
instead of these electronic gadgets, electronic neurons wired
together in the same circuits that human neurons are wired in,
let us suppose you have miniature people, little girl scouts and
boy scouts.' We don't even have to imagine that these little peo-
ple even know what the whole scheme is for, or that they see
anything except a dimly lit room, or a lot of dimly lit rooms, in
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 97
which they pass notes to one another. (Their time would have to
pass very fast relative to 'our' time, of course.) They could be
alienated workers. 'Now,' the reply continues, 'you wouldn't
call that thing "conscious" because you know that it is really
only these little people moving the body. And that shows that an
appropriate functional organization (one like ours) is not suffi-
cient to justify the application of such predicates as
"conscious".'
One reply to this reply (the one I actually made) was to deny
that the 'hydra-headed robot' (as this last thing has been called)
does have the same functional organization we do. But there is
a more radical reply I might have made. I might have said, 'Why
shouldn't we call the hydra-headed robot conscious? If the first
argument is right (and I think it is), if the robot with the posi-
tronic brain would be conscious, why would the fact that the
neurons of the hydra-headed robot are more conscious mean
that the whole thing is less conscious? After all, we are in a sense
a society of small animals. Our cells are in a sense individual
animals. And perhaps they have some little bit of feeling, who
knows? Over and above our feeling.' Now, if we move that way,
if we decide that the hydra-headed robot is conscious (even
though its neurons are boy scouts and girl scouts), then why not
the United States?
I don't, of course, claim that the United States has the same
functional organization as homo sapiens. Clearly it doesn't. But
there are many similarities. The United States has defensive
organs. It has ingesting organs, it eats oil and copper and so on.
It excretes (pollution) in vast quantities. Is it not perhaps as sim-
ilar in functional organization to a mammal as is a wriggling fly,
to which we do attribute pain?
How well-defined is 'qualitative character'?

So far we have not questioned the idea that it is perfectly clear
what it means to say that two of one's own sensations have or
do not have the same 'qualitative character'. Even at an intro-
spective level, this is not the case, however. For one thing, just
what one's experiences seem to one to be is notoriously depen-
dent on antecedent conceptualization, as when we report seeing
a round table top even when we view it from an angle.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
98 Mind and body
In the case of the round table top, psychologists and philoso-

phers have argued since the nineteenth century about whether
one has 'eliptical sense data' and thinks that they are round
(unless one is a 'trained introspectionist') or has round 'Gestalts'
and only thinks that they are elliptical because of optical theory.
One can have experiences which fit each description; and many
experiences will fit either description. Nor is neurology likely to
settle this dispute: the elliptical image on the retina doubtless
produces events in the brain itself, and if we identify these with
the 'visual sensation', then we may well get something like the
classical story of 'elliptical sense-data plus unconscious infer-
ences'; the judged character of the experience ('I see a round
table top') also corresponds to 'records' and 'inputs' in the brain,
and if we identify these with the visual sensation, we may well
get a story in which one doesn't have elliptical sense-data unless
one judges that something looks elliptical. Why should we not
say that these two versions are both legitimate? As Goodman
says about the case of the subject who is asked to describe appar-
ent motion,
The best we can do is to specify the sort of terms, the
vocabulary, he is to use, telling him to describe what he
sees in perceptual or phenomenal rather than physical
terms. Whether or not this yields different responses, it
casts an entirely different light on what is happening.
That the instrument to be used in fashioning the facts
must be specified makes pointless any identification of the
physical with the real and of the perceptual with the
merely apparent. The perceptual is no more a rather
distorted version of the physical facts than the physical is
a highly artificial version of the perceptual facts.10
If I see a red tablecloth at two different times during the day,
do I have the same sensation of red? Or do I have different sen-
sations and not notice the difference (unless I happen to be a
painter) ?
An especially baffling case is the case of accommodation. If a
subject is given glasses which turn the image upside down, after
a time things will again look normal to the subject. Have the
10
Ways of Worldmaking, pp. 92-3.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 99
sense-data 'flipped back'? Or has he gotten used to altered sense

data, and reinterpreted 'up' and 'down'? Very likely the subject
himself cannot say at what point things became normal or which
of these things happened. (Readers who like me wear bifocals
can ask themselves: does the lower half of the visual field look
different even when one isn't noticing the difference?) While
there are transformations to which subjects never accommodate,
(in fact, it is only relatively simple changes to which one accom-
modates), and I have assumed that one would not accommodate
to a color-inversion, the phenomenon of accommodation cer-
tainly casts doubt on the extent to which 'same qualitative char-
acter' is a well-defined notion.
Realism about qualia

We have considered a set of sceptical difficulties. What they pur-
port to show is not that the identity theory is wrong (or that the
correlation theory is wrong — note that they can all be stated as
difficulties for a 'correlation' view just as much as for an identity
view), but that, if it is true, then there are a vast number of alter-
native ways of specifying the details such that one can never
know which one of them is true. And not knowing which one of
them is true means not knowing what the answer is to a great
many traditional sceptical questions, such as whether rocks and
other inanimate objects have qualia, whether bats and other spe-
cies have the kind of qualia we have or don't have the kind of
qualia we have, whether groups can feel pain, and so on.
But why should any philosopher think it is even a logical pos-
sibility that a rock can have a pain (i.e. that an event of the same
'qualitative character' as a human pain can take place 'in' a
rock) ? Perhaps Russell gives us some clue to the nature of this
kind of metaphysical realism. Russell was a realist about qualia
and a realist about universals. Moreover, he took qualia to be
paradigmatic universals. A universal is, above all, a way in
which things can be similar; and to Russell it seemed that the
qualitative similarities of one's own sensations are the episte-
mologically most primitive and most fundamental examples of
'ways in which things can be similar'. Qualia, for Russell, are
universals par excellence.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
100 Mind and body
Universals, however, are thought of as totally well-defined by

a traditional realist: words may be vague, but universals them-
selves can't be vague. (A vague word is vague because it stands
for a vague set of concepts, Godel once said in a conversation;
but the concepts are perfectly well-defined.)
So, if qualia are universals and universals are by nature well-
defined, it must be perfectly well-defined whether any given
thing or event — including a half of a split brain or some event in
it; including a rock or some event in it; including a nation or
group or some event in it— does or doesn't exhibit any given
quale. And if the quale is thought of as independent of the func-
tional role it plays, if it is thought to be wholly contingent that
the qualitative character of a sensation of red is the qualitative
character of something which has that particular functional role,
then it does seem to be a logical possibility that the split brain or
the rock has that quale.
A philosopher like myself who wishes to deny that every one
of these possibilities makes sense (although some of them may -
there is a temptation to treat the right lobe of the split brain as
a 'locus of consciousness', and I have suggested that it would be
legitimate to decide to do this) has to be careful to make clear
that he is not espousing some form of behaviorism. Saying that
'qualia' are not well defined entities is not the same thing as say-
ing they don't exist, that it is all just behavior, or whatever.
Many notions are vague and still have some clear referents. The
notion of a house, for example, is ill-defined in the case of igloos
(is an igloo a house?), in the case of hogans, perhaps in other
cases as well. Buttfte fact that there is no fact of the matter as to
whether or not an igloo is a house doesn't mean that houses
don't exist. And, similarly, the fact that there is no fact of the
matter as to whether or not the right lobe is 'conscious' doesn't
mean that conscious beings don't exist.
Qualitative similarity of sensations is defined to some extent:
if I have a sensation of red followed by a sensation of green, I
know that I have had dissimilar sensations (and I know this
without comparing their functional roles), and if I have a sensa-
tion of red followed by the 'same' sensation of red, I know (up
to the vagueness we discussed above) that I have had similar
sensations. But, for someone with an 'internalist' perspective on
truth, it does not follow that there is a fact of the matter in every
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Mind and body 101
case as to whether two sensations (let alone two arbitrary events)

are qualitatively similar or dissimilar.
Let E be the event of my having a particular sensation at a
particular time and E' be some physical event in a rock. The
suggestion that the qualitative character of E (say, redH) might
be identical with or correlated to some such property as (P1 or
P2 or P3) (where P 3 is the property of being a rock) offends any
sane human sensibility. The suggestion that E and E' might be
'qualitatively similar' events is absurd. We have already dis-
cussed one explanation of this absurdity: the explanation that
the hypothesis is absurd because it violates the methodological
maxim 'do not ascribe properties to an object without a reason'.
Even if this explanation worked, what it would yield is far less
than the impossibility of rocks having qualia (or the incoherence
of the notion that they do). If this is all that is wrong with the
'hypothesis' that rocks have qualia then we are in the position of
having to say: it is possible for all we know that rocks have
qualia, but it is a priori highly improbable that they do.
In fact, the hypothesis that rocks have qualia is incoherent in
much the way that the Brain in a Vat hypothesis is incoherent:
like the Brain in a Vat hypothesis, this 'hypothesis' presupposes
a magical theory of reference. Any sane human being regards E
and £' as so dissimilar that the question of 'qualitative similar-
ity' (in the sense in which two sensations can be qualitatively
similar, i.e. feel the same way) does not even arise. But the meta-
physical realist, while not denying this at all, thinks that E and
Ef might (logically possibly) be similar in this way, even though
it is 'crazy' to think so. And he thinks this because he is under
the illusion that by having the sensation in question, with its
qualitative character, its 'the way it feels', with its functional
role, with the accompanying thoughts and judgments, he has
somehow brought it about that the expression 'the way this sen-
sation feels' (or some technical substitute, e.g. 'the qualitative
character of this sensation', or (redH\ or 'this quale') refers to
one definite 'universal', one absolutely well-defined property of
metaphysical individual events. But this is not the case.
If there actually were robots functionally isomorphic to us and
we worked with them, argued with them, had some of them as
friends, we would quickly feel sure that they were conscious.
(We might still be puzzled as to whether they had the same
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
102 Mind and body
qualia we do; but we would not think of this any more often
than we think of the question whether bats or dogs have the
same qualia we do.) Suppose, however, we encountered hydra-
headed robots. (Imagine that they actually evolved by some
biological process somewhere, just as animals in symbiotic rela-
tionships evolve on earth.) What would we feel about them?
While one cannot really feel sure about so bizarre a case, it
seems that even here (if we interacted mostly with the whole
robot and only rarely with its conscious 'neurons' — the 'boy
scouts and girl scouts' of my story) we might begin to attribute
consciousness; but probably we would always be divided in our
opinions. If we came to be sure that the hydra-headed robots
were conscious, then might we begin to be ever-so-slightly
queasy about the United States? I do not know.
The perspective I urge with respect to all of these cases is that
there is nothing hidden here, no noumenal fact of the entities'
really being conscious or really not being conscious, or of the
qualities' really being the same or really being different. There
are only the obvious empirical facts: that rocks and nations are
grossly dissimilar from people and animals; that robots of var-
ious kinds are in between sorts of objects; and so on. Rocks and
nations aren't conscious; that is a fact about the notion of con-
sciousness we actually have.
What makes this line seem so disturbing is that it makes our
standards of rational acceptability, justification and ultimately
of truth, dependent on standards of similarity which are clearly
the product of our biological and cultural heritage (e.g. whether
we have or haven't interacted with 'intelligent robots'). But
something like this is true of most of the language we use in
everyday life, of such words as 'person', 'house', 'snow', and
'brown', for example. A realist who accepted this resolution of
the puzzles about qualia would be likely to express it by saying
that 'qualia don't really exist', or that qualia belong to our 'sec-
ond class conceptual system'; but what is the point of a notion
of 'existence' that puts houses on the side of the non-existent?
Our world is a human world, and what is conscious and not
conscious, what has sensations and what doesn't, what is quali-
tatively similar to what and what is dissimilar, are all dependent
ultimately on our human judgments of likeness and difference.
Cambridge
https://doi.org/10.1017/CBO9780511625398.006
Two conceptions of rationality
In the preceding chapters I have spoken of rationality and of

'rational acceptability'. But rationality is not an easy thing to
give an account of.
The problem is not without analogues in other areas. Some
years ago I studied the behavior of natural kind words, for
example, gold, and I came to the conclusion that the extension
of the term is not simply determined by a 'battery of semantical
rules', or other institutionalized norms. The norms may deter-
mine that certain objects are paradigmatic examples of gold; but
they do not determine the full extension of the term, nor is it
impossible that even a paradigmatic example should turn out
not to really be gold, as it would be if the norms simply defined
what it is to be gold.
We are prepared to count something as belonging to a kind
even if our present tests do not suffice to show it is a member of
the kind if it ever turns out that it has the same essential nature
as (or, more vaguely, is 'sufficiently similar' to) the paradigmatic
examples (or the great majority of them). What the essential
nature is, or what counts as sufficient similarity, depends both
on the natural kind and on the context (iced tea may be 'water'
in one context but not in another); but for gold what counts is
ultimate composition, since this has been thought since the
ancient Greeks to determine the lawful behavior of the sub-
stance. Unless we say that what the ancient Greeks meant by
chrysos was whatever has the same essential nature as the para-
digmatic examples, then neither their search for new methods of
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
104 Two conceptions of rationality
detecting counterfeit gold (which led Archimedes to the density

test) nor their physical speculations will make sense.
It is tempting to take the same line with rationality itself, and
to say that what determines whether a belief is rational is not the
norms of rationality of this or that culture, but an ideal theory
of rationality, a theory which would give necessary and suffi-
cient conditions for a belief to be rational in the relevant circum-
stances in any possible world. Such a theory would have to
account for the paradigmatic examples, as an ideal theory of
gold accounts for the paradigmatic examples of gold; but it
could go beyond them, and provide criteria which would enable
us to understand cases we cannot presently see to the bottom of,
as our present theory of gold enables us to understand cases the
most brilliant ancient Greek could not have understood. A gen-
eral difficulty with the proposal to treat 'rational', 'reasonable',
'justified', etc., as natural kind terms is that the prospects for
actually finding powerful generalizations about all rationally
acceptable beliefs seem so poor. There are powerful universal
laws obeyed by all instances of gold, which is what makes it
possible to describe gold as the stuff that will turn out to obey
these laws when we know them; but what are the chances that
we can find powerful universal generalizations obeyed by all
instances of rationally justified belief?
That the chances are poor does not mean that there are no
analogies between scientific inquiry into the nature of gold and
moral inquiry or philosophical inquiry. In ethics, for example,
we start with judgments that individual acts are right or wrong,
('observation reports', so to speak) and we gradually formulate
maxims (not exceptionless generalizations) based on those judg-
ments, often accompanied by reasons or illustrative examples, as
for instance 'Be kind to the stranger among you, because you
know what it was like to be a stranger in Egypt' (a 'low level
generalization'). These maxims in turn affect and alter our judg-
ments about individual cases, so that new maxims supplement-
ing or modifying the earlier ones may appear. After thousands
of years of this dialectic between maxims and judgments about
individual cases, a philosopher may come along and propose a
moral conception (a 'theory'), which may alter both maxims and
singular judgments and so on.
The very same procedure may be found in all of philosophy
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
Two conceptions of rationality 105
(which is almost coextensive with theory of rationality). In a

publication a few years1 ago I described the desiderata for a
moral system, following Grice and Baker, and I included (1) the
desire that one's basic assumptions, at least, should have wide
appeal; (2) the desire that one's system should be able to with-
stand rational criticism; (3) the desire that the morality recom-
mended should be livable.
The way to develop a better understanding of the nature of
rationality — the only way we know — is, likewise, to develop
better philosophical conceptions of rationality. (An unending
process; but that is as it should be.) It is striking that the desid-
erata I listed for a moral system, unchanged, could be listed as
the desiderata for a methodology or a system of rational proce-
dure in any major area of human concern. In analytical philos-
ophy the main attempts to better understand the nature of
rationality in this way have come from philosophers of science,
and two important tendencies have resulted from these efforts.
Logical positivism
In the past fifty years the clearest manifestation of the tendency
to think of the methods of 'rational justification' as given by
something like a list or canon (although one that philosophers of
science have admittedly not yet succeeded in fully formalizing)
was the movement known as Logical Positivism. Not only was
the list or canon that the positivists hoped 'logicians of science'
(their term for philosophers) would one day succeed in writing
down supposed to exhaustively describe the 'scientific method';
but, since, according to the logical positivists, the 'scientific
method' exhausts rationality itself, and testability by that
method exhausts meaningfulness ('The meaning of a sentence is
its method of verification'), the list or canon would determine
what is and what is not a cognitively meaningful statement.
Statements testable by the methods in the list (the methods of
mathematics, logic, and the empirical sciences) would count as
meaningful; all other statements, the positivists maintained, are
'pseudo-statements', or disguised nonsense.
1
'Literature, Science, and Reflection', New Literary History, vol. VII,
1975—6, reprinted in my Meaning and the Moral Sciences, Routledge
and Kegan Paul, 1978.
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
An obvious rejoinder was to say that the Logical Positivist

criterion of significance was self-refuting: for the criterion itself
is neither (a) 'analytic' (a term used by the positivists to account
for logic and mathematics), nor (b) empirically testable.
Strangely enough this criticism had very little impact on the log-
ical positivists and did little to impede the growth of their move-
ment. I believe that the neglect of this particular philosophical
gambit was a great mistake; that the gambit is not only correct,
but contains a deep lesson, and not just a lesson about Logical
Positivism.
The point I am going to develop will depend on the following
observation: the forms of 'verification' allowed by the logical
positivists are forms which have been institutionalized by mod-
ern society. What can be 'verified' in the positivist sense can be
verified to be correct (in a non-philosophical or prephilosophical
sense of 'correct'), or to be probably correct, or to be highly
successful science, as the case may be; and the public recognition
of the correctness, or the probable correctness, or the 'highly
successful scientific theory' status, exemplifies, celebrates, and
reinforces images of knowledge and norms of reasonableness
maintained by our culture.
On the face of it, the original positivist paradigm of verifica-
tion was not this publicly institutionalized one. In Carnap's Der
Logische Aufbau der Welt (The Logical Construction of the
World) verification was ultimately private, based on sensations
whose subjective quality or 'content' was said to be 'incommu-
nicable'. But, under the urgings of Neurath, Carnap soon shifted
to a more public, more 'intersubjective', conception of verifica-
tion.
Popper has stressed the idea that scientific predictions are con-
fronted with 'basic sentences', sentences such as 'the right pan of
the balance is down' which are publicly accepted even if they
cannot be 'proved' to the satisfaction of a sceptic. He has been
criticized for using 'conventionalist' language here, for speaking
as if it were a convention or social decision to accept a basic
sentence; but I think that what sounds like a conventionalist ele-
ment in Popper's thought is simply a recognition of the institu-
tionalized nature of the implicit norms to which we appeal in
ordinary perceptual judgments. The nature of our response to a
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
sceptic who challenges us to 'prove' such statements as 'I am

standing on the floor' testifies to the existence of social norms
requiring agreement to such statements in the appropriate cir-
cumstances.
Wittgenstein argued that without such public norms, norms
shared by a group and constituting a 'form of life', language and
even thought itself would be impossible. For Wittgenstein it is
absurd to ask if the institutionalized verification I have been
speaking of is 'really' justificatory. In On Certainty Wittgenstein
remarks that philosophers can provide one with a hundred epis-
temological 'justifications' of the statement 'cats don't grow on
trees' - but none of them starts with anything which is more sure
(in just this institutionalized sense of 'sure') than the fact that
cats don't grow on trees.
Sceptics have doubted not only perceptual judgments but
ordinary inductions. Hume, whose distinction between what is
rational and what is reasonable I am not observing, would have
said there is no rational proof that it will snow (or even that it
will probably snow) in the United States this winter (although he
would have added that it would be most unreasonable to doubt
that it will). Yet our response to a sceptic who challenges us to
'prove' that it will snow in the United States this winter testifies
that there are social norms requiring agreement to such 'induc-
tions' just as much as to ordinary perceptual judgments about
people standing on floors and about equal arm balances.
When we come to high-level theories in the exact sciences,
people's reactions are somewhat different. Ordinary people can-
not 'verify' the special theory of relativity. Indeed, ordinary peo-
ple do not at the present time even learn the special theory, or
the (relatively elementary) mathematics needed to understand it,
although it is beginning to be taught in freshman physics courses
in some of our colleges. Ordinary people defer to scientists for
an informed (and socially accepted) appraisal of a theory of this
type. And because of the instability of scientific theories, a sci-
entist is not likely to refer to even so successful a theory as spe-
cial relativity as 'true' tout court. But the judgment of the scien-
tific community is that special relativity is a 'successful' — in fact,
like quantum electrodynamics, an unprecedentedly successful —
scientific theory, which yields 'successful predictions' and which
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
is 'supported by a vast number of experiments'. And these judg-

ments are, in fact, deferred to by other members of the society.
The difference between this case and the cases of institutional-
ized norms of verification previously referred to (apart from the
hedging of the adjective 'true') is the special role of experts and
the institutionalized deference to experts that such a case
involves; but this is no more than an instance of the division of
intellectual labor (not to mention intellectual authority relations)
in the society. The judgment that special relativity and quantum
electrodynamics are 'the most successful physical theories we
have' is one which is made by authorities which the society has
appointed and whose authority is recognized by a host of prac-
tices and ceremonies, and in that sense institutionalized.
Recently it occurred to me that Wittgenstein may well have
thought that only statements that can be verified in some such
'institutionalized' way can be true (or right, or correct, or justi-
fied) at all. I don't mean to suggest that any philosopher ever
held the view that all things which count in our society as 'justi-
fications' really are such. Philosophers generally distinguish
between institutions which are constitutive of our concepts
themselves and those which have some other status, although
there is much controversy about how to make such a distinction.
I mean to suggest that Wittgenstein thought that it was some
subset of our institutionalized verification norms that determines
what it is right to say in the various 'language games' we play
and what is wrong, and that there is no objective Tightness or
wrongness beyond this. Although such an interpretation does fit
much that Wittgenstein says — for instance, the stress on the
need for 'agreement in our judgments' in order to have concepts
at all - I do not feel sure that it is right. It is just too vague who
the 'we' is in Wittgenstein's talk of 'our' judgments; and I don't
know whether his 'forms of life' correspond to the institutional-
ized norms I have mentioned. But this interpretation occurred to
me upon reading Wittgenstein's Lectures and Conversations. In
this Wittgenstein rejects both psychoanalysis and Darwin's the-
ory of evolution (although unlike the positivists he does not
regard such language as meaningless, and he has admiration for
Freud's 'cleverness'). Wittgenstein's view about psychoanalysis
(which he calls a 'myth') does not signify much, since so many
people have the view - mistakenly in my opinion - that psycho-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
analysis is more or less nonsense. But his rejection of evolution

is quite striking.2 Wittgenstein contrasts Darwin's theory unfa-
vorably with theories in physics ('One of the most important
things about an explanation is that it should work, that it should
enable us to predict something. Physics is connected with Engi-
neering. The bridge must not fall down' (Lectures on Aesthetics,
p. 25)). And he says people were persuaded 'on grounds which
were extremely thin'. 'In the end you forget entirely every ques-
tion of verification, you are just sure it must have been like that.'
Again, the great discussions about 'analyticity' that went on
in the 1950s seem to me to be connected with the desire of phi-
losophers to find an objective, uncontroversial foundation for
their arguments. 'Analyticity', i.e. the doctrine of truth by virtue
of meaning alone, came under attack because it had been over-
used by philosophers. But why had philosophers been tempted
to announce that so many things which are in no intelligible
sense 'rules of language', or consequences of rules of language,
were analytic or 'conceptually necessary', or whatever? The
answer, I think, is that the idea that there is a definite set of rules
of language and that these can settle what is and is not rational,
had two advantages, as philosophers thought: (1) the 'rules of
2
Concerning evolution what Wittgenstein said was 'People were certain
on grounds which were extremely thin. Couldn't there have been an
attitude which said: "I don't know. It is an interesting hypothesis which
may eventually be well confirmed" ', Lectures on Aesthetics, p. 26, in
Cyril Burret (ed.) L. W. Wittgenstein: Lectures and Conversations,
Berkeley, University of California Press, 1967. What it would be like for
evolution to be 'well confirmed' Wittgenstein does not say, but the
paragraph suggests that actually seeing speciation occur is what he has
in mind ('Did anyone see this process happening? No. Has anyone
seen it happening now? No. The evidence of breeding is just a drop in
the bucket.')
It is instructive to contrast Wittgenstein's attitude with Monod's:
the selective theory of evolution, as Darwin himself had stated it,
required the discovery of Mendelian genetics, which of course
was made. This is an example, and a most important one,
of what is meant by the content of a theory, the content of an
idea . . . [A] good theory or a good idea will be much wider and
much richer than even the inventor of the idea may know at
his time. The theory may be judged precisely on this type
of development, when more and more falls into its lap, even
though it was not predictable that so much would come of it (J.
Monod, 'On the Molecular Theory of Evolution', in Harre,
R. (ed.), Problems of Scientific Revolution: Progress and
Obstacles to Progress in the Sciences, Oxford, 1975).
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
language' are constitutive institutionalized practices (or norms

which underlie such practices), and as such have the 'public' sta-
tus I have described; (2) at the same time, it was claimed that
only philosophers (and not linguists) could discover these mys-
terious things. It was a nice idea while it lasted, but it was bound
to be exploded, and it was.
I shall call any conception according to which there are insti-
tutionalized norms which define what is and is not rationally
acceptable a criterial conception of rationality. The logical posi-
tivists, Wittgenstein, at least on the admittedly uncertain inter-
pretation I have essayed, and some though not all of the 'ordi-
nary language' philosophers3 at Oxford shared a criterial
conception of rationality even if they differed on other issues,
such as whether to call unverifiable statements 'meaningless',
and over whether or not some ethical propositions could be
'conceptually necessary'.
3
One might develop an 'ordinary language' philosophy which was not
committed to the public and 'criterial' verification of philosophical
theses if one could develop and support a conception in which the norms
which govern linguistic practices are not themselves discoverable by
ordinary empirical investigation. In Must We Mean What We
Say, Stanley Cavell took a significant step in this direction, arguing that
such norms can be known by a species of 'self knowledge' which he
compared to the insight achieved through therapy and also to the
transcendental knowledge sought by phenomenology. While I agree with
Cavell that my knowledge as a native speaker that certain uses are
deviant or non-deviant is not 'external' inductive knowledge — I
can know without evidence that in my dialect of English one says 'mice'
and not 'mouses' — I am inclined to think this fact of speaker's
privileged access does not extend to generalizations about correctness
and incorrectness. If I say (as Cavell does) that it is part of the rule
for the correct use of locutions of the form X is voluntary that
there should be something 'fishy' about X, then I am advancing a theory
to explain my intuitions about specific cases, not just reporting those
intuitions. It is true that something of this sort also goes on in
psychotherapy; but I am not inclined to grant self-knowledge any kind
of immunity from criticism by others, including criticisms which
depend on offering rival explanations, in either case. And if one allows
the legitimacy of such criticism, then the activity of discovering such
norms begins to look like social science or history — areas in which, I
have argued, traditional accounts of 'The Scientific Method' shed little
light. (See my Meaning and the Moral Sciences, Routledge and
Kegan Paul, 1978.)
In any case, whatever their status, I see no reason to believe that the
norms for the use of language are what decide the extension of
'rationally acceptable', 'justified', 'well confirmed', and the like.
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
The gambit I referred to at the outset, the gambit that refutes

the logical positivists' verification principle, is deep precisely
because it refutes every attempt to argue for a criterial concep-
tion of rationality, that is because it refutes the thesis that noth-
ing is rationally verifiable unless it is criterially verifiable.
The point is that although the philosophers I mentioned often
spoke as if their arguments had the same kind of finality as a
mathematical proof or a demonstration experiment in physics;
that although the logical positivists called their work logic of
science; although the Wittgensteinians displayed unbelievable
arrogance towards philosophers who could not 'see' that all
philosophical activity of a pre-Wittgensteinian or non-Wittgen-
steinian kind is nonsensical; and although ordinary language
philosophers referred to each other's arguments and those of
non-ordinary language philosophers as 'howlers' (as if philo-
sophical errors were like mistakes on an arithmetic test); no phil-
osophical position of any importance can be verified in the con-
clusive and culturally recognized way I have described. In short,
if it is true that only statements that can be criterially verified
can be rationally acceptable, that statement itself cannot be cri-
terially verified, and hence cannot be rationally acceptable. If
there is such a thing as rationality at all - and we commit our-
selves to believing in some notion of rationality by engaging in
the activities of speaking and arguing — then it is self-refuting to
argue for the position that it is identical with or properly con-
tained in what the institutionalized norms of the culture deter-
mine to be instances of it. For no such argument can be certified
to be correct, or even probably correct, by those norms alone.
I don't at all think that rational argumentation and rational
justification are impossible in philosophy, but rather I have been
driven to recognize something which is probably evident to lay-
men if not to philosophers, namely that we cannot appeal to
public norms to decide what is and is not rationally argued and
justified in philosophy. The claim which is still often heard that
philosophy is 'conceptual analysis', that the concepts themselves
determine what philosophical arguments are right, is, when
combined with the doctrine that concepts are norms or rules
underlying public linguistic practices, just a covert form of the
claim that all rational justification in philosophy is criterial, and
that philosophical truth is (barring 'howlers') as publicly demon-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
strable as scientific truth. Such a view seems to me to be simply

unreasonable in the light of the whole history of the subject,
including the recent history.
What goes for philosophical argument goes for arguments
about religion and about secular ideology as well. An argument
between an intelligent liberal and an intelligent Marxist will
have the same character as a philosophical dispute at the end,
even if more empirical facts are relevant. And we all do have
views in religion, or politics, or philosophy, and we all argue
them and criticize the arguments of others. Indeed, even in 'sci-
ence', outside of the exact sciences, we have arguments in his-
tory, in sociology and in clinical psychology, of exactly this char-
acter. It is true that the logical positivists broadened their
description of the 'scientific method' to include these subjects;
but so broadened it cannot be shown to clearly exclude anything
whatsoever. (See Chapter 8.)
The positivists, I will be reminded, conceded that the verifica-
tion principle was 'cognitively meaningless'. They said it was a
proposal and as such not true or false. But they argued for their
proposal, and the arguments were (and had to be) non-starters.4
So the point stands.
4
The weakest argument offered in defense of the Verification Principle
construed as a proposal was that it 'explicated' the 'pre-analytic' notion
of meaningfulness. (For a discussion of this claim, see my 'How
Not to Talk about Meaning', in my Mind, Language and Reality,
Philosophical Papers, Vol. 2, Cambridge University Press, 1975.)
Reichenbach defended a form of the Verification Principle (in Experience
and Prediction) as preserving all differences in meaning relevant to
behavior. Against an obvious objection (that the non-empirical belief in
a divinity — Reichenbach used the example of Egyptian cat wor-
shippers - could alter behavior) Reichenbach replied by proposing to
translate 'Cats are divine animals' as 'cats inspire feelings of awe in
cat-worshippers'. Clearly the acceptance of this substitute would
not leave behavior unchanged in the case of a cat worshipper!
The most interesting view was that of Carnap. According to Carnap,
all rational reconstructions are proposals. The only factual questions
concern the logical and empirical consequences of accepting this or that
rational reconstruction. (Carnap compared the 'choice' of a rational
reconstruction to the choice of an engine for an airplane.) The con-
clusion he drew was that in philosophy one should be tolerant of
divergent rational reconstructions. However, this principle of Tolerance,
as Carnap called it, presupposes the Verification Principle. For the
doctrine that no rational reconstruction is uniquely correct or corre-
sponds to the way things 'really are', the doctrine that all 'external
questions' are without cognitive sense is just the Verification Principle.
To apply the Principle of Tolerance to the Verification Principle
itself would be circular.
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
In sum, what the logical positivists and Wittgenstein (and per-

haps the later Quine as well) did was to produce philosophies
which leave no room for a rational activity of philosophy. This
is why these views are self-refuting; and also why the little gam-
bit I have been discussing represents a significant argument of
the kind philosophers call a 'transcendental argument': arguing
about the nature of rationality (the task of the philosophers par
excellence) is an activity that presupposes a notion of rational
justification wider than the positivist notion, indeed wider than
institutionalized criterial rationality.
Anarchism is self-refuting
Let me now discuss a very different philosophical tendency.
Thomas Kuhn's The Structure of Scientific Revolutions (SSR)
enthralled vast numbers of readers, and appalled most philoso-
phers of science because of its emphasis on what seemed to be
irrational determinants of scientific theory acceptance and by
its use of such terms as 'conversion' and 'Gestalt switch'. In fact,
Kuhn made a number of important points about scientific theo-
ries and about how scientific activity should be viewed. I have
expressed a belief in the importance of the notions of paradigm,
normal science, and scientific revolution elsewhere; at this point
I want to focus on what I do not find sympathetic in Kuhn's
book, what I described elsewhere as 'Kuhn's extreme relativism'.
The reading that enthralled Kuhn's more sophomoric readers
was one according to which he is saying that there is no such
thing as rational justification in science, it's just Gestalt switches
and conversions. Kuhn has rejected this interpretation of the
SSR, and has since introduced a notion of 'non-paradigmatic
rationality' which may be closely related to if not the same as
what I just called 'non-criterial rationality'.
The tendency that most readers thought they detected in
Kuhn's SSR certainly manifested itself in Paul Feyerabend's
Against Method. Feyerabend, like Kuhn, stressed the manner in
which different cultures and historic epochs produce different
paradigms of rationality. He suggests that the determinants of
our conceptions of scientific rationality are largely what we
would call irrational. In effect, although he does not put it this
way, he suggests that the modern scientific-technological con-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
ception of rationality is fraudulent by its own standards. (I think

I detect a similar strain in Michel Foucault.) And he goes far
beyond Kuhn or Foucault in suggesting that even the vaunted
instrumental superiority of our science may be somewhat of a
hoax. Faith healers can do more to relieve your pain than doc-
tors, Feyerabend claims.
It is not those terrifyingly radical claims that I want to talk
about, although they are the reason Feyerabend calls his position
'anarchism'. I wish to discuss a claim Kuhn does make in both
the SSR and subsequent papers, and that Feyerabend made both
in Against Method and in technical papers. This is the thesis of
incommensurability. I want to say that this thesis, like the logical
positivist thesis about meaning and verification, is a self-refuting
thesis. In short, I want to claim that both of the two most influ-
ential philosophies of science of the twentieth century, certainly
the two that have interested scientists and non-philosophers gen-
erally, the only two the educated general reader is likely to have
even heard of, are self-refuting. Of course, as a philosopher of
science I find it a bit troublesome that this should be the case.
We shall shortly come to the question of what to make of this
situation.
The incommensurability thesis is the thesis that terms used
in another culture, say, the term 'temperature' as used by a
seventeenth-century scientist, cannot be equated in meaning or
reference with any terms or expressions we possess. As Kuhn
puts it, scientists with different paradigms inhabit 'different
worlds'. 'Electron' as used around 1900 referred to objects
in one 'world'; as used today it refers to objects in quite a
different 'world'. This thesis is supposed to apply to observa-
tional language as well as to so-called 'theoretical language';
indeed, according to Feyerabend, ordinary language is simply a
false theory.
The rejoinder this time is that if this thesis were really true
then we could not translate other languages - or even past stages
of our own language - at all. And if we cannot interpret orga-
nisms' noises at all, then we have no grounds for regarding them
as thinkers, speakers, or even persons. In short, if Feyerabend
(and Kuhn at his most incommensurable) were right, then mem-
bers of other cultures, including seventeenth-century scientists,
would be conceptualizable by us only as animals producing
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
responses to stimuli (including noises that curiously resemble

English or Italian). To tell us that Galileo had 'incommensura-
ble' notions and then to go on to describe them at length is
totally incoherent.
This problem is posed in a sympathetic essay on Feyerabend's
view by Smart:5
Surely it is a neutral fact that in order to see Mercury

we have to point the telescope over the top of that
tree, say, and not, as predicted by Newtonian theory,
over the top of that chimney pot. And surely one can
talk of trees, chimney pots, and telescopes in a way
which is independent of the choice between Newton-
ian and Einsteinian theory. However Feyerabend could
well concede that we use Euclidean geometry and non-
relativistic optics for the theory of our telescope. He
would say that this is not the real truth about our tele-
scope, the tree, and the chimney pot, but nevertheless
it is legitimate to think in this way in order to discuss
the observational tests of general relativity, since we know
on theoretical grounds that our predictions will be un-
affected (up to the limits of observational error) if we
avail ourselves of this computational convenience.
But the trouble with Smart's rescue move is that I must under-
stand some of the Euclidean non-relativists' language to even say
the 'predictions' are the same. If every word has a different sig-
nificance, in what sense can any prediction be 'unaffected'? How
can I even translate the logical particles (the words for 'if-then',
'not', and so on) in seventeenth-century Italian, or whatever, if
I cannot find a translation manual connecting seventeenth-
century Italian and modern English that makes some kind of sys-
tematic sense of the seventeenth-century corpus, both in itself
and in its extra-linguistic setting? Even if I am the speaker who
employs both theories (as Smart envisages) how can I be justified
5
J. J. C. Smart, 'Conflicting Views about Explanation', in R. Cohen and
M. Wartofsky (eds.), Boston Studies in the Philosophy of Science,
Volume II: in Honor of PhHipp Frank (New York, Humanities Press,
Inc., 1965).
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
in equating any word in my Newtonian theory with any word in

my general relativistic theory?
The point I am making comes into even sharper focus when
we apply to it some of Quine's and Davidson's observations
about meaning and translation practice. Once it is conceded that
we can find a translation scheme which 'works' in the case of a
seventeenth century text, at least in the context fixed by our
interests and the use to which the translation will be put, what
sense does it have in that context to say that the translation does
not 'really' capture the sense or reference of the original? It is
not, after all, as if we had or were likely to have criteria for
sameness of sense or reference apart from our translation
schemes and our explicit or implicit requirements for their
empirical adequacy. One can understand the assertion that a
translation fails to capture exactly the sense or reference of the
original as an admission that a better translation scheme might
be found; but it makes only an illusion of sense to say that all
possible translation schemes fail to capture the 'real' sense or
reference. Synonymy exists only as a relation, or better, as a fam-
ily of relations, each of them somewhat vague, which we employ
to equate different expressions for the purposes of interpreta-
tion. The idea that there is some such thing as 'real' synonymy
apart from all workable practices of mutual interpretation, has
been discarded as a myth.
Suppose someone tells us that the German word 'Rad' can be
translated as 'wheel'. If he goes on to say his translation is not
perfect, we naturally expect him to indicate how it might be
improved, supplemented by a gloss, or whatever. But if he goes
on to say that 'Rad' can be translated as 'wheel', but it doesn't
actually refer to wheels, or indeed to any objects recognized in
your conceptual system, what do we get from this? To say that
a word A can be translated as 'wheel', or whatever, is to say
that, to the extent that the translation can be relied upon, A
refers to wheels.
Perhaps the reason that the incommensurability thesis
intrigues people so much, apart from the appeal which all inco-
herent ideas seem to have, is the tendency to confuse or conflate
concept and conception. To the extent that the analytic/synthetic
distinction is fuzzy, this distinction too is fuzzy; but all interpre-
tation involves such a distinction, even if it is relative to the inter-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
pretation itself. When we translate a word as, say, temperature

we equate the reference and, to the extent that we stick to our
translation, the sense of the translated expression with that of
our own term 'temperature', at least as we use it in that context.
(Of course, there are various devices we can use, such as special
glosses, to delimit or delineate the way we are employing 'tem-
perature', or whatever the word may be, in the context.) In this
sense we equate the 'concept' in question with our own 'concept'
of temperature. But so doing is compatible with the fact that the
seventeenth-century scientists, or whoever, may have had a dif-
ferent conception of temperature, that is a different set of beliefs
about it and its nature than we do, different 'images of knowl-
edge', and different ultimate beliefs about many other matters as
well. That conceptions differ does not prove the impossibility of
ever translating anyone 'really correctly' as is sometimes sup-
posed; on the contrary, we could not say that conceptions differ
and how they differ if we couldn't translate.
But, it may be asked, how do we ever know that a translation
scheme 'works' if conceptions always turn out to be different?
The answer to this question, as given by various thinkers from
Vico down to the present day, is that interpretative success does
not require that the translatees' beliefs come out the same as our
own, but it does require that they come out intelligible to us.
This is the basis of all the various maxims of interpretative char-
ity or 'benefit of the doubt', such as 'interpret them so they come
out believers of truths and lovers of the good', or 'interpret them
so that their beliefs come out reasonable in the light of what they
have been taught and have experienced', or Vico's own directive
to maximize the humanity of the person being interpreted. It is
a constitutive fact about human experience in a world of differ-
ent cultures interacting in history while individually undergoing
slower or more rapid change that we are, as a matter of universal
human experience, able to do this; able to interpret one
another's beliefs, desires, and utterances so that it all makes
some kind of sense.
Kuhn and Feyerabend, not surprisingly, reject any idea of con-
vergence in scientific knowledge. Since we are not talking about
the same things as previous scientists, we are not getting more
and more knowledge about the same microscopic or macro-
scopic objects. Kuhn argues that science 'progresses' only instru-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
mentally; we get better and better able to transport people from

one place to another, and so on. But this too is incoherent.
Unless such locutions as 'transport people from one place to
another' retain some degree of fixity of reference, how can we
understand the notion of instrumental success in any stable way?
The argument I have just employed is essentially related to
Kant's celebrated arguments about preconditions for empirical
knowledge. Replying to the contention that the future might be
wholly lawless, might defeat every 'induction' we have made,
Kant pointed out that if there is any future at all - any future for
us, at any rate, any future we can grasp as thinkers and concep-
tualize to say if our predictions were true or false - then, in fact,
many regularities must not have been violated. Else why call it a
future? For example, when we imagine balls coming from an urn
in some 'irregular' order, we forget that we couldn't even tell
they were balls, or tell what order they came out in, without
depending on many regularities. Comparison presupposes there
are some commensurabilities.
There is a move Kuhn and Feyerabend could make in reply to
all these criticisms, but it is not one they would feel happy mak-
ing, and that would be to introduce some kind of observa-
tional/theoretical dichotomy. They could concede commensura-
bility, translatability, and even convergence with respect to
observational facts, and restrict the incommensurability thesis to
the theoretical vocabulary. Even then there would be problems
(why shouldn't we describe the meanings of the theoretical terms
via their relations to the observational vocabulary a la Ramsey?)
But Kuhn and Feyerabend reject this alternative with reason, for
in fact the need for principles of interpretative charity is just as
pervasive in 'observational language' as in 'theoretical language'.
Consider, for example, the common word 'grass'. Different
speakers, depending on where and when they live have different
perceptual prototypes of grass (grass has different colors and dif-
ferent shapes in different places) and different conceptions of
grass. Even if all speakers must know that grass is a plant, on
pain of being said to have a different concept altogether, the
conception of a plant today involves photosynthesis and the con-
ception of a plant two hundred years ago did not. Without inter-
pretative charity which directs us to equate 'plant' 200 years ago
with 'plant' today (at least in ordinary contexts) and 'grass' 200
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
years ago with 'grass' today, no statement about the reference of

this word 200 years ago could be made. Nor is it only natural
kind words that are so dependent for interpretation on principles
of charity; the artifact word 'bread' would pose exactly the same
problems. Indeed, without interpretative charity we could not
equate a simple color term such as 'red' across different speak-
ers. We interpret discourse always as a whole; and the interpre-
tation of 'observation' terms is as dependent on the interpreta-
tion of 'theoretical' terms as is the interpretation of the latter on
the former.
What I have given is, once again, a transcendental argument.
We are committed by our fundamental conceptions to treating
not just our present time-slices, but also our past selves, our
ancestors, and members of other cultures past and present, as
persons; and that means, I have argued, attributing to them
shared references and shared concepts, however different the
conceptions that we also attribute. Not only do we share objects
and concepts with others, to the extent that the interpretative
exercise succeeds, but also conceptions of the reasonable, of the
natural, and so on. For the whole justification of an interpreta-
tive scheme, remember, is that it renders the behavior of others
at least minimally reasonable by our lights. However different
our images of knowledge and conceptions of rationality, we
share a huge fund of assumptions and beliefs about what is rea-
sonable with even the most bizarre culture we can succeed in
interpreting at all.
Why relativism is inconsistent

That (total) relativism is inconsistent is a truism among philos-
ophers. After all, is it not obviously contradictory to hold a point
of view while at the same time holding that no point of view is
more justified or right than any other? Alan Garfinkel has put
the point very wittily. In talking to his California students he
once said, aping their locutions: 'You may not be coming from
where I'm coming from, but I know relativism isn't true for me'
. . . If any point of view is as good as any other, then why isn't
the point of view that relativism is false as good as any other?
The plethora of relativistic doctrines being marketed today
(and marketed by highly intelligent thinkers) indicates this sim-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
pie refutation will not suffice. Why should an intelligent relativ-

ist concede that every view is as true {for him) as any other? He
cannot prevent you (or Alan Garfinkel) from saying that his view
is not true for you (or justified for you, or whatever): but if he
has his wits about him, he can retort that truth for you is far less
salient {for him) than is truth for him. What concept of anything
is more salient than one's own, after all? Is it then really incon-
sistent to treat true, justified, etc., as relative notions?
The answer is that it is inconsistent but it does require a more
elaborate argument than the (nonetheless very nice) one-liner
produced by Garfinkel. The important point to notice is that if
all is relative, then the relative is relative too. But this takes a bit
of explaining!
Plato was perhaps the first to employ the sort of argument I
have in mind (against Protagoras). Protagoras (a deep-dyed rel-
ativist, apparently) claimed that when I say X, I really should say
'I think that X'. Thus when I say 'Snow is white', Protagoras
would say that I really mean that Hilary Putnam thinks that
snow is white, and that what Robert Nozick means by the same
utterance is that Robert Nozick thinks that snow is white. A
more sophisticated statement of the same idea would be that
when I say 'Snow is white', I am using this utterance to claim
that snow is white is true-for-me, whereas when Robert Nozick
says the same words he would normally be claiming that snow
is white is true-ior-him (or at least he would count his statement
as having been correct just in case it turned out to be true-for-
him). It follows (on Protagoras' view) that no utterance has the
same meaning for me and for anyone else; there is, as we saw
before, an intimate connection between relativism and incom-
mensurability. Plato's counter-argument was that, if every state-
ment X means 'I think that X\ then I should (on Protagoras'
view) really say
(1) I think that I think that snow is white.
But the process of adding 'I think' can always be iterated! On
Protagoras' view, the ultimate meaning of'Snow is white' is then
not (1) but
(2) I think that I think that I think that I . . . (with
infinitely many 'I thinks') that snow is white.
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
This Plato took to be a reductio ad absurdum. However,

Plato's argument is not a good one as it stands. Why should
Protagoras not agree that his analysis applies to itself? It doesn't
follow that it must be self-applied an infinite number of times,
but only that it can be self-applied any finite number of times.
But Plato had noticed something very deep.
When one first encounters relativism, the idea seems simple
enough. The idea, in a natural first formulation is that every per-
son (or, in a modern 'sociological' formulation, every culture, or
sometimes every 'discourse') has his (its) own views, standards,
presuppositions, and that truth (and also justification) are rela-
tive to these. One takes it for granted, of course, that whether X
is true (or justified) relative to these is itself something 'absolute'.
Modern Structuralists like Foucault write as if justification rel-
ative to a discourse is itself quite absolute — i.e. not at all relative.
But if statements of the form fX is true (justified) relative to per-
son P' are themselves true or false absolutely, then there is, after
all, an absolute notion of truth (or of justification) and not only
of truth-for-me, truth-for-Nozick, truth-for-you, etc. A total rel-
ativist would have to say that whether or not X is true relative
to P is itself relative. At this point our grasp on what the position
even means begins to wobble, as Plato observed.
Plato's line of attack on relativism does not seem to have been
followed up until recently. But it was brilliantly extended by
Wittgenstein in, of all places, the Private Language Argument
(alluded to in Chapter 3).
Most commentators read the Private Language Argument as
simply an argument against the 'copy theory' of truth. And Witt-
genstein's brilliant demonstration that the similitude theory of
reference does not work even for reference to sensations is cer-
tainly part of a sustained attack on metaphysical realism. But I
prefer to read the argument as a pair of quite traditional argu-
ments (at least Kant would have approved of both of them!)
against two positions, one a realist position and one a relativist
position: for the attempt to read the whole argument as an anti-
realist one makes it come out looking rather contrived.
The form of relativism Wittgenstein was concerned to attack
is known as 'methodological solipsism'. A 'methodological
solipsist' is a non-realist or 'verificationist' who agrees that truth
is to be understood as in some way related to rational accept-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
ability, but who holds that all justification is ultimately in terms

of experiences that each of us has a private knowledge of. Thus,
I have my knowledge of what experiences of mine would verify
that snow is white and Bob Nozick has his knowledge of what
experiences of his would verify that snow is white: every state-
ment has a different sense for every thinker.
Wittgenstein's argument seems to me to be an excellent argu-
ment against relativism in general. The argument is that the rel-
ativist cannot, in the end, make any sense of the distinction
between being right and thinking he is right; and that means that
there is, in the end, no difference between asserting or thinking,
on the one hand, and making noises {or producing mental
images) on the other. But this means that (on this conception) I
am not a thinker at all but a mere animal. To hold such a view
is to commit a sort of mental suicide.
To see that Wittgenstein was right, let us consider, as Wittgen-
stein does not, how the relativist might attempt to draw the dis-
tinction that Wittgenstein denies him, the distinction between
being right and thinking he is right.
The relativist might borrow the idea that truth is an idealiza-
tion of rational acceptability. He might, hold that X is true-for-
me if SX is justified-for-me' would be true provided I observed
carefully enough, reasoned long enough, or whatever. But sub-
junctive conditionals of the form 'If I were to . . . , then I would
think such-and-such', are, like all statements, interpreted differ-
ently by different philosophers.
A metaphysical realist can regard statements about what
would be the case if as themselves true or false in an absolute
sense, independently of whether we ever will be justified in
accepting or rejecting them. If the relativist interprets statements
about what he would believe under such-and-such conditions in
this realist way, then he has recognized one class of absolute
truths, and so has given up being a relativist.
A non-realist or 'internal' realist regards conditional state-
ments as statements which we understand (like all other state-
ments) in large part by grasping their justification conditions.
This does not mean that the 'internal' realist abandons the dis-
tinction between truth and justification, but that truth {idealized
justification) is something we grasp as we grasp any other con-
cept, via a (largely implicit) understanding of the factors that
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
make it rationally acceptable to say that something is true. Can

the relativist interpret statements about what he would believe
under ideal conditions in this non-realist or 'internal' realist
way?
Let us recall that the non-realist position, as I described it
(in Chapter 3), assumes an objective notion of rational accepta-
bility. The non-realist rejects the notion that truth is corres-
pondence to a 'ready-made world'. That is what makes him a
«o«-(metaphysical)-realist. But rejecting the metaphysical 'cor-
respondence' theory of truth is not at all the same thing as regard-
ing truth or rational acceptability as subjective. Nelson Good-
man, who regards truth and rational acceptability as species of
a more general predicate 'rightness', applicable to works of art
as well as to statements, has put the point succinctly:
Briefly, then, truth of statements and rightness of descrip-
tions, representations, exemplifications, expressions — of
design, drawing, diction, rhythm - is primarily a matter of
fit: fit to what is referred to in one way or another, or to
other renderings, or to modes and manners of organiza-
tion. The differences between fitting a version to a world,
a world to a version, and a version together or to other
versions fade when the role of versions in making the
worlds they fit is recognized. And knowing or under-
standing is seen as ranging beyond the acquiring of true
beliefs to the discovering and devising of fit of all sorts.
The whole purpose of relativism, its very defining characteristic,
is, however, to deny the existence of any intelligible notion of
objective 'fit'. Thus the relativist cannot understand talk about
truth in terms of objective justification-conditions.
The attempt to use conditionals to explicate the distinction
between being right and thinking one is right fails, then, because
the relativist has no objective notion of rightness for these con-
ditionals any more than he does for any other sort of statement.
Finally, if the relativist of today, like the ancient Protagoras,
simply decides to bite the bullet and say that there is no differ-
ence between 'I am right' and 'I think I am right' — that a dis-
tinction between being justified and thinking one is justified can-
not be drawn in one's own case — then what is speaking, on such
a conception - beyond producing noises in the hope that one
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
will have the feeling of being right? What is thinking — beyond

producing images and sentence-analogues in the mind in the
hope of having a subjective feeling of being right? The relativist
must end by denying that any thought is about anything in either
a realist or non-realist sense; for he cannot distinguish between
thinking one's thought is about something and actually thinking
about that thing. In short, what the relativist fails to see is that
it is a presupposition of thought itself that some kind of objective
'tightness' exists.
There is an interesting relation between the argument I just
analyzed (Plato—Wittgenstein) and the argument against incom-
mensurability I attributed to Quine and Davidson: Quine and
Davidson argue, in effect, that a consistent relativist should not
treat others as speakers (or thinkers) at all (if their 'noises' are
that 'incommensurable', then they are just noises), while Plato
and Wittgenstein argue, in effect, that a consistent relativist can-
not treat himself as a speaker or thinker.
What to make of this?

The arguments I just set before you convinced me that the two
most widely known philosophies of science produced in this cen-
tury are both incoherent. (Of course, neither of them is just a
'philosophy of science'.) This naturally led me to reflect on the
meaning of this situation. How did such views arise?
Logical positivism, I recalled, was both continuous with and
different from the Machian positivism which preceded it.
Mach's positivism, or 'empirio-criticism', was, in fact, largely a
restatement of Humean empiricism in a different jargon. Mach's
brilliance, his dogmatic and enthusiastic style, and his scientific
eminence made his positivism a large cultural issue (Lenin, afraid
that the Bolsheviks would be converted to 'empirio-criticism',
wrote a polemic against it). Einstein, whose interpretation of
special relativity was operationalist in spirit (in marked contrast
to the interpretation he gave to general relativity), acknowledged
that his criticism of the notion of simultaneity owed much to
Hume and to Mach, although, to his disappointment, Mach
totally rejected special relativity.
But the most striking event that led up to the appearance of
logical positivism was the revolution in deductive logic. By 1879
Frege had discovered an algorithm, a mechanical proof proce-
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
dure, that embraces what is today standard 'second order logic'.

The procedure is complete for the elementary theory of deduc-
tion ('first order logic'). The fact that one can write down an
algorithm for proving all of the valid formulas of first order
logic —an algorithm which requires no significant analysis and
simulation of full human psychology — is a remarkable fact. It
inspired the hope that one might do the same for so called
'inductive logic' — that the 'scientific method' might turn out to
be an algorithm, and that these two algorithms - the algorithm
for deductive logic (which, of course, turned out to be incom-
plete when extended to higher logic) and the algorithm-to-be-
discovered for inductive logic - might exhaustively describe or
'rationally reconstruct' not just scientific rationality, but all
rationality worthy of the name.
When I was just starting my teaching career at Princeton Uni-
versity I got to know Rudolph Carnap, who was spending two
years at the Institute for Advanced Studies. One memorable
afternoon, Carnap described to me how he had come to be a
philosopher. Carnap explained to me that he had been a gradu-
ate student in physics, studying logic in Frege's seminar. The text
was Principia Mathematica (imagine studying Russell and
Whitehead's Principia with Frege!) Carnap was fascinated with
symbolic logic and equally fascinated with the special theory of
relativity. So he decided to make his thesis a formalization of
special relativity in the notation of Principia. It was because the
Physics Department at Jena would not accept this that Carnap
became a philosopher, he told me.
Today, a host of negative results, including some powerful
considerations due to Nelson Goodman, have indicated that
there cannot be a completely formal inductive logic. Some
important aspects of inductive logic can be formalized (although
the adequacy of the formalization is controversial), but there is
always a need for judgments of 'reasonableness', whether these
are built in via the choice of vocabulary (or, more precisely, the
division of the vocabulary into 'projectible' predicates and 'non-
projectible' predicates) or however. Today, virtually no one
believes that there is a purely formal scientific method (on this,
see Chapter 8).
The story Carnap told me supports the idea that it was the
sucess of formalization in the special case of deductive logic that
played a crucial role. If that success inspired the rise of logical
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
positivism, could it not have been the failure to formalize induc-

tive logic, the discovery that there is no algorithm for empirical
science, that inspired the rise of 'anarchism'?
I won't press this suggestion; in any case, additional factors
are probably at work. While Kuhn has increasingly moderated
his view, both Feyerabend and Michel Foucault have tended to
push it to extremes. There is something political in their minds:
both Feyerabend and Foucault link our present institutionalized
criteria of rationality with capitalism, exploitation, and even
with sexual repression. Clearly there are many divergent reasons
why people are attracted to extreme relativism today, the idea
that all existing institutions and traditions are bad being one of
them.
Another reason is a certain scientism. The scientistic character
of logical positivism is quite overt and unashamed; but I think
there is also a scientism hidden behind relativism. The theory
that all there is to 'rationality' is what your local culture says
there is is never quite embraced by any of the 'anarchistic' think-
ers, but it is the natural limit of their tendency: and this is a
reductionist theory. That rationality is defined by an ideal com-
puter program is a scientistic theory inspired by the exact sci-
ences; that it is simply defined by the local cultural norms is a
scientistic theory inspired by anthropology.
I will not discuss here the expectation aroused in some by
Chomskian linguistics that cognitive psychology will discover
innate algorithms which define rationality. I myself think that
this is an intellectual fashion which will be disappointed as the
logical positivist hope for a symbolic inductive logic was disap-
pointed.
All this suggests that part of the problem with present day
philosophy is a scientism inherited from the nineteenth century —
a problem that affects more than one intellectual field. I do not
deny that logic is important, or that formal studies in confirma-
tion theory, in semantics of natural language, and so on are
important. I do tend to think that they are rather peripheral to
philosophy, and that as long as we are too much in the grip of
formalization we can expect this kind of swinging back and
forth between the two sorts of scientism I described. Both sorts
of scientism are attempts to evade the issue of giving a sane and
human description of the scope of reason.
Cambridge
https://doi.org/10.1017/CBO9780511625398.007
Fact and value
Understood in a sufficiently wide sense, the topic of fact and

value is a topic which is of concern to everyone. In this respect,
it differs sharply from many philosophical questions. Most edu-
cated men and women do not feel it obligatory to have an opin-
ion on the question whether there really is a real world or only
appears to be one, for example. Questions in philosophy of lan-
guage, epistemology, and even in metaphysics may appear to be
questions which, however interesting, are somewhat optional
from the point of view of most people's lives. But the question
of fact and value is a forced choice question. Any reflective per-
son has to have a real opinion upon it (which may or may not
be the same as their notional opinion). If the question of fact and
value is a forced choice question for reflective people, one partic-
ular answer to that question, the answer that fact and value are
totally disjoint realms, that the dichotomy 'statement of fact or
value judgment' is an absolute one, has assumed the status of a
cultural institution.
By calling the dichotomy a cultural institution, I mean to sug-
gest that it is an unfortunate fact that the received answer will
go on being the received answer for quite some time regardless
of what philosophers may say about it, and regardless of
whether or not the received answer is right. Even if I could con-
vince you that the fact-value dichotomy is without rational
basis, that it is a rationally indefensible dichotomy, or even if
some better philosopher than I could show this by an absolutely
conclusive argument (of course there are no such in philosophy),
still the next time you went out into the street, or to a cocktail
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
128 Fact and value
party, or had a discussion at some deliberative body of which

you happen to be a member, you would find someone saying to
you, 'Is that supposed to be a statement of fact or a value judg-
ment?' The view that there is no fact of the matter as to whether
or not things are good or bad or better or worse, etc. has, in a
sense, become institutionalized.
The strategy of my argument is not going to be a new one. I'm
going to rehabilitate a somewhat discredited move in the debate
about fact and value, namely the move that consists in arguing
that the distinction is at the very least hopelessly fuzzy because
factual statements themselves, and the practices of scientific
inquiry upon which we rely to decide what is and what is not a
fact, presuppose values.
The reason this is a somewhat discredited move is that there is
an obvious rejoinder to it. The rejoinder to the view that science
presupposes values is a protective concession. The defenders of
the fact—value dichotomy concede that science does presuppose
some values, for example, science presupposes that we want
truth, but argue that these values are not ethical values. I shall
imagine a somewhat strawman opponent who takes the view
that science presupposes one value, namely the value of truth
itself.
As we have seen, truth is not a simple notion. The idea that
truth is a passive copy of what is 'really' (mind-independently,
discourse-independently) 'there' has collapsed under the cri-
tiques of Kant, Wittgenstein, and other philosophers even if it
continues to have a deep hold on our thinking.
Some philosophers have appealed to the equivalence principle,
that is the principle that to say of a statement that it is true is
equivalent to asserting the statement, to argue that there are no
real philosophical problems about truth. Others appeal to the
work of Alfred Tarski, the logician who showed how, given a
formalized language (a formal notation for expressing certain
statements, employing symbolic logic), one can define 'true' for
that language in a stronger language (a so-called 'meta-
language').1
Tarski's work was itself based on the equivalence principle: in
1
For a non-technical account of Tarski's work see my Meaning and the
Moral Sciences, Part I, Lecture I.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 129
fact his criterion for a successful definition of 'true' was that it

should yield all sentences of the form 'P' is true if and only if P,
e.g.
(T) 'Snow is white' is true if and only if snow is white
as theorems of the meta-language (where P is a sentence of the
formal notation in question).
But the equivalence principle is philosophically neutral, and so
is Tarski's work. On any theory of truth, 'Snow is white' is
equivalent to ' "Snow is white" is true.'
Positivist philosophers would reply that if you know (T)
above, you know what " 'Snow is white" is true' means: it
means snow is white. And if you don't understand 'snow' and
'white', they would add, you are in trouble indeed! But the prob-
lem is not that we don't understand 'Snow is white'; the problem
is that we don't understand what it is to understand 'Snow is
white.' This is the philosophical problem. About this (T) says
nothing.
And indeed does this not accord with our intuitions about
these matters? If someone approaches us with a gleam in his eye
and says, 'Don't you want to know the "Truth"?', our reaction
is generally to be pretty leery of this person. And the reason that
we are leery (apart from the gleam in the eye) is precisely because
someone's telling us that they want us to know the truth tells us
really nothing as long as we have no idea what standards of
rational acceptability the person adheres to: what they consider
a rational way to pursue an inquiry, what their standards of
objectivity are, when they consider it rational to terminate an
inquiry, what grounds they will regard as providing good reason
for accepting one verdict or another on whatever sort of ques-
tion they may be interested in. Applied to the case of science, I
would say that to tell us that science 'seeks to discover the truth'
is really a purely formal statement. It is to say no more than that
scientists don't want to assert that snow is white if snow is not
white, that they don't want to assert that there are electrons
flowing through a wire if electrons are not flowing through the
wire, and so on. But these purely formal statements are quite
empty as long as we don't have some idea wrhat the system of
criteria of rational acceptability is which distinguishes scientific
ways of attempting to determine whether snow is white from
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
130 Fact and value
other ways of attempting to determine whether snow is white,

scientific ways of attempting to determine whether electrons are
flowing through a wire from other ways of attempting to deter-
mine whether there are electrons flowing through a wire, and so
on.
If the notion of comparing our system of beliefs with uncon-
ceptualized reality to see if they match makes no sense, then the
claim that science seeks to discover the truth can mean no more
than that science seeks to construct a world picture which, in the
ideal limit, satisfies certain criteria of rational acceptability. That
science seeks to construct a world picture which is true is itself a
true statement, an almost empty and formal true statement; the
aims of science are given material content only by the criteria of
rational acceptability implicit in science. In short I am saying
that the answer to the 'strawman' position I considered, that the
only aim of science is to discover truth (besides pointing out that
science has additional aims, which is of course true), is that truth
is not the bottom line: truth itself gets its life from our criteria of
rational acceptability, and these are what we must look at if we
wish to discover the values which are really implicit in science.
For the purpose of an example let me now imagine an extreme
case of disagreement. The disagreement I'm going to imagine is
not an ordinary scientific disagreement, although I hope our
response to it will enable us to discover something about the
nature of scientific values.
The hypothesis that the disagreement is going to be about, in
the case I am about to describe, is just the hypothesis we dis-
cussed in Chapter 1, the hypothesis that we are all Brains in a
Vat. We have argued that this hypothesis cannot possibly be
true; but we shall suppose that our arguments have failed to
convince one side in this disagreement (which is not improbable,
since philosophical arguments never convince everyone). In
short, the hypothesis is that everything is a collective hallucina-
tion in the way we described before.
Of course, if it were all one collective hallucination in this
way, there are many people to whom this need not make any
difference. It would make little or no difference to lovers, for
example.2 And I imagine it would make no difference at all to
2
But I keep changing my mind about whether it would or not.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 131
economists. (Why should an economist care if all the money in

the world isn't physically real? Most of it isn't physically real on
any theory!)
I want the reader to imagine that this crazy (and, I would
claim, incoherent) theory, the theory that we are all brains in a
vat, is held not by an isolated lunatic, but by virtually all the
people in some large country, say, Australia. Imagine that in
Australia only a small minority of the people believe what we do
and the great majority believe that we are Brains in a Vat. Per-
haps the Australians believe this because they are all disciples of
a Guru, the Guru of Sydney, perhaps. Perhaps when we talk to
them they say, 'Oh if you could talk to the Guru of Sydney and
look into his eyes and see what a good, kind, wise man he is, you
too would be convinced.' And if we ask, 'But how does the Guru
of Sydney know that we are brains in a vat, if the illusion is as
perfect as you say?', they might reply, 'Oh, the Guru of Sydney
just knows/
As I said before, this is not a scientific disagreement in the
ordinary sense. We can imagine that the Australians are just as
good as we are at anticipating experiences, at building bridges
that stay up (or seem to stay up), etc. They may even be willing
to accept our latest scientific discoveries, not as true, but as cor-
rect descriptions of what seems to go on in the image. We may
or may not imagine that they disagree with us about some pre-
dictions concerning the very distant future (for example, they
might expect that some day the automatic machinery will break
down and then people will begin to have collective hallucina-
tions of a kind which will give evidence that their view is right),3
but whether they do make such predictions, or whether they
commit themselves to no predictions different from the ones
afforded by standard theory, will not affect my argument. The
point is that here I've imagined a case where a vast number of
people have a self-contained belief system which violently dis-
agrees with ours.
There is no question of a disagreement in 'ethical' values here;
3
If they do make such predictions, then it does make this much difference:
their view is no longer incoherent in the way we criticized in Chapter
1, since they are making a claim that could be justified (eventually), and
hence one that does not require a view of truth as 'transcendent' (or
independent of justification) to be understood.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
132 Fact and value
the Australians can have ethics just as similar to ours as you like.
(Although an ancient Greek would have said that being wise is
an ethical value; Judaism and Christianity have, in fact, nar-
rowed the notion of the ethical because of a certain conception
of Salvation.)
The first thing I want to observe about the hypothetical Aus-
tralians is that their world view is crazy. Sometimes, to be sure,
'crazy9 is used almost as a term of approval; but I don't mean it
in that sense here. I think we would regard a community of
human beings who held so insane a world view with great sad-
ness. The Australians would be regarded as crazy in the sense of
having sick minds; and the characterization of their minds as
sick is an ethical one, or verges on the ethical. But how, other
than by calling them names, could one argue with the Austra-
lians? (Or try to argue with them, for I shall suppose that they
are not to be convinced.)
One argument that one can immediately think of has to do
with the incoherence of their view. I don't just mean the inco-
herence that we found in the view in Chapter 1. That is a deep
incoherence, which requires a philosophical (and hence contro-
versial) argument to expose. But the Australian's view is inco-
herent at a much more superficial level. One of the things that
we aim at is that we should be able to give an account of how
we know our statements to be true. In part we try to do this by
developing a causal theory of perception, so that we can account
for what we take to be the reliability of our perceptual knowl-
edge, viewed from within our theory itself, by giving an account
within the theory of how our perceptions result from the opera-
tion of transducing organs upon the external world. In part we
try to do this by a theory of statistics and experimental design,
so that we can show, within our theory itself, how the proce-
dures that we take to exclude experimental error really do have
a tendency in the majority of cases to exclude experimental
error. In short, it is an important and extremely useful constraint
on our theory itself that our developing theory of the world
taken as a whole should include an account of the very activity
and processes by which we are able to know that that theory is
correct.
The Australians' system, however, does not have this property
of coherence (at least as we judge it, and 'coherence' is not some-
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 133
thing that we have an algorithm for, but something that we ulti-

mately judge by 'seat of the pants' feel). The Australians, remem-
ber, have themselves postulated an illusion so perfect that there
is no rational way in which the Guru of Sydney can possibly
know that the belief system which he has adopted and persuaded
all the others to adopt is correct. Judged by our standards of
coherence, their belief system is totally incoherent.
Other methodological virtues could be listed which their belief
system lacks. Their belief system, as I described it, agrees with
ours concerning what the laws of nature are in the image; but
does it tell us whether or not the laws of nature that appear to
hold in the image are the laws of nature that actually hold out-
side the vat? If it fails to, then it lacks a certain kind of compre-
hensiveness which we aim after, for it does not, even in its own
terms, tell what the true and ultimate laws of nature are. Cer-
tainly it violates Ockham's razor. Again, Ockham's razor seems
difficult or impossible to formalize as an algorithm, but the very
fact that the Brain in a Vatist theory postulates all kinds of
objects outside the vat which play no role in the explanation of
our experiences, according to the theory itself, makes it clear
that this is a case in which we can definitely say that the maxim
. . . 'don't multiply entities without necessity' is violated. Let us
call a theory which obeys Ockham's razor, in spirit as opposed
to just in letter, functionally simple.
What I have been saying is that the procedures by which we
decide on the acceptability of a scientific theory have to do with
whether or not the scientific theory as a whole exhibits certain
'virtues'. I am assuming that the procedure of building up scien-
tific theory cannot be correctly analyzed as a procedure of veri-
fying scientific theories sentence by sentence. I am assuming that
verification in science is a holistic matter, that it is whole theo-
retical systems that meet the test of experience 'as a corporate
body', and that the judgment of how well a whole system of
sentences meets the test of experience is ultimately somewhat of
an intuitive matter which could not be formalized short of for-
malizing total human psychology. But let us come back to our
original question. What are the values implicit in science?
I've been arguing that if we take the values to which we appeal
in our criticism of the Brain-in-a-Vatists, and add, of course,
other values which are not at issue in this case, e.g. our desire for
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
134 Fact and value
instrumental efficacy, which we presumably share with the

Brain-in-a-Vatists, then we get a picture of science as presuppos-
ing a rich system of values. The fact is that, if we consider the
ideal of rational acceptability which is revealed by looking at
what theories scientists and ordinary people consider rational to
accept, then we see that what we are trying to do in science is to
construct a representation of the world which has the character-
istics of being instrumentally efficacious, coherent, comprehen-
sive, and functionally simple. But why?
I would answer that the reason we want this sort of represen-
tation, and not the 'sick' sort of notional world possessed by the
Australians, possessed by the Brain-in-a-Vatists, is that having
this sort of representation system is part of our idea of human
cognitive flourishing, and hence part of our idea of total human
flourishing, of Eudaemonia.
Of course, if metaphysical realism were right, and one could
view the aim of science simply as trying to get our notional
world to 'match' the world in itself, then one could contend that
we are interested in coherence, comprehensiveness, functional
simplicity, and instrumental efficicacy only because these are
instruments to the end of bringing about this 'match'. But the
notion of a transcendental match between our representation
and the world in itself is nonsense. To deny that we want this
kind of metaphysical match with a noumenal world is not to
deny that we want the usual sort of empirical fit (as judged by
our criteria of rational acceptability) with an empirical world.
But the empirical world, as opposed to the noumenal world,
depends upon our criteria of rational acceptability (and, of
course, vice versa). We use our criteria of rational acceptability
to build up a theoretical picture of the 'empirical world' and then
as that picture develops we revise our very criteria of rational
acceptability in the light of that picture and so on and so on
forever. The dependence of our methods on our picture of the
world is something I have stressed in my other books; what I
wish to stress here is the other side of the dependence, the depen-
dence of the empirical world on our criteria of rational accepta-
bility. What I am saying is that we must have criteria of rational
acceptability to even have an empirical world, that these reveal
part of our notion of an optimal speculative intelligence. In
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 135
short, I am saying that the 'real world' depends upon our values
(and, again, vice versa).
At least some values must be objective

The fact that science is not 'value neutral', as has been thought,
does not, to be sure, show that 'ethical' values are objective, or
that ethics could be a science. In fact, there is no prospect of a
'science' of ethics, whether in the sense of a laboratory science or
of a deductive science. As Aristotle long ago remarked,4
We must be content, then, in speaking of such subjects
and with such premisses to indicate the truth roughly and
in outline, and in speaking about things which are only
for the most part true, and with premisses of the same
kind, to reach conclusions which are no better. In
the same spirit, therefore, should each kind of statement
be received, for it is the mark of an educated man to
look for precision in each class of things just so far as the
nature of the subject admits; it is evidently foolish to
accept probable reasoning from a mathematician and to
demand from a rhetorician scientific proofs.
But the fact that rational acceptability in the exact sciences
(which are certainly central examples of rational thinking) does
depend on such cognitive virtues as 'coherence' and 'functional
simplicity' shows that at least some value terms stand for prop-
erties of the things they are applied to, and not just for feelings
of the person who uses the terms.
If the terms 'coherent' and 'simple' do not stand tor properties
of theories, not even fuzzy or imperfectly defined ones, but only
for 'attitudes' that some people have towards theories, then such
terms for rational acceptability as 'justified', 'well confirmed',
'best of the available explanations' must also be entirely subjec-
tive: for rational acceptability cannot be more objective than the
parameters upon which it depends. But, as we argued in the pre-
ceding chapter, the view that rational acceptability itself is sim-
ply subjective is a self-refuting one. So we are compelled to con-
4
Ethica Nicomachea, Book I, Ch. 3.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
136 Fact and value
elude that at least these value-terms have some sort of objective

application, some sort of objective justification conditions.
Of course, one might attempt to avoid conceding that there
are objective values of any kind by choosing to deny that
'coherent', 'simple', 'justified', and the like are value terms. One
might hold that they stand for properties which we do value, but
that there is no objective Tightness about our doing so. But this
line runs into difficulties at once. 'Coherent' and 'simple' have
too many characteristics in common with the paradigmatic value
words. Like 'kind', 'beautiful', and 'good', 'coherent' and 'sim-
ple' are often used as terms of praise. Our conceptions of coher-
ence, simplicity, and justification are just as historically condi-
tioned as our conceptions of kindness, beauty, and goodness;
these epistemic terms figure in the same sorts of perennial
philosophical controversies as do the terms for ethical and aes-
thetic values. The conception of rationality of a John Cardinal
Newman is obviously quite different from that of a Rudolf Car-
nap. It is highly unlikely that either could have convinced the
other, had they lived at the same time and been able to meet.
The question: which is the rational conception of rationality
itself is difficult in exactly the way that the justification of an
ethical system is difficult. There is no neutral conception of
rationality to which to appeal.
One might attempt various conventionalist moves here, e.g.
saying that 'justifiedCarnap' is one 'property' and justifiedNewman'
is a different 'property', and that a 'subjective value judgment'
is involved in the decision to mean 'justifiedcarnap' or 'justi-
fied Newman' by the word 'justified' but that no value judgment
is involved in stating the fact that a given statement S is justi-
fied camap o r JustifiedNewman. But from whose standpoint is the
word fact' being used? If there is no conception of rationality one
objectively ought to have, then the notion of a 'fact' is empty.
Without the cognitive values of coherence, simplicity, and
instrumental efficacy we have no world and no 'facts', not even
facts about what is so relative to what, for those are in the same
boat with all other facts. And these cognitive values are arbitrary
considered as anything but a part of a holistic conception of
human flourishing. Bereft of the old realist idea of truth as 'cor-
respondence' and of the positivist idea of justification as fixed by
public 'criteria', we are left with the necessity of seeing our
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 137
search for better conceptions of rationality as an intentional

human activity, which, like every activity that rises above habit
and the mere following of inclination or obsession, is guided by
our idea of the good.
Rationality in other areas

If the values implicit in science, especially in the exact sciences,
reveal a part of our idea of the good, I think that the rest of our
idea of the good can be read off from our standards of rational
acceptability in yet other areas of knowledge. At this point, how-
ever, it is necessary to broaden the notion of standards of
rational acceptability.
So far, we have only considered standards of rational accept-
ability in the literal sense: standards which tell us when we
should and when we should not accept statements. But stan-
dards of rationality in the wide sense have to do not only with
how we judge the truth or falsity of systems of statements, but
also with how we judge their adequacy and perspicuousness.
There are ways - purely cognitive ways - in which a system of
statements can fall short of giving us a satisfactory description
other than by being false.
Had I chosen I could have made this point even in connection
with theoretical science. I could have pointed out that the con-
cern of exact science is not just to discover statements which are
true, or even statements which are true and universal in form
('laws'), but to find statements that are true and relevant. And
the notion of relevance brings with it a wide set of interests and
values. But this would have only been to argue that our knowl-
edge of the world presupposes values, and not to make the more
radical claim that what counts as the real world depends upon
our values.
When we come to perceptual rationality, that is to the implicit
standards and skills on the basis of which we decide whether
someone is able to give a true, adequate, and perspicuous
account of even the simplest perceptual facts, then we see a large
number of factors at play. Recently psychologists have stressed
just how much theory construction is involved in even the sim-
plest cases of perception. Not only is this true at the neurophys-
iological level, but it is also true at the cultural level. Someone
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
138 Fact and value
from a culture which had no furniture might be able to come

into a room and give some kind of description of the room, but,
if he did not know what a table or a chair or a desk was, his
description would hardly convey the information that a member
of this culture would wish to have about the room. His descrip-
tion might consist only of true statements but it would not be
adequate.
What this simple example shows is that the requirement that
a description be adequate is implicitly a requirement that the
describer have available a certain set of concepts; we expect
rational describers with respect to certain kinds of descripta to
be capable of acquiring certain concepts and of seeing the need
to use them; the fact that the describer did not employ a certain
concept may be a ground for criticizing both him and his
description.
What is true at the simple level of talk about tables and chairs
in a room without people in it is also true at the level of descrip-
tion of interpersonal relations and situations. Consider the terms
we use every day in describing what other people are like, e.g.
considerate or inconsiderate. Considerate and inconsiderate may
of course be used to praise or blame; and one of the many dis-
tinctions which have gotten confused together under the general
heading 'fact—value distinction' is the distinction between using
a linguistic expression to describe and using that linguistic
expression to praise or blame. But this distinction is not a dis-
tinction which can be drawn on the basis of vocabulary. The
judgment that someone is inconsiderate may indeed be used to
blame; but it may be used simply to describe, and it may also be
used to explain or to predict.
For example I may say to you, 'Don't let Jones hurt your feel-
ings. You're likely to think that he's taken a dislike to you from
the way he will talk, but that's a common misimpression. No
matter what he feels about you he'll likely behave in such a way
that your feelings will be hurt. He's just a rather inconsiderate
man, but don't think that it has anything to do with you.'
In this little imaginary speech someone is using the word
'inconsiderate' not for the purpose of blaming Jones, but with
the intention of predicting and explaining Jones' behavior to
someone else. And both the prediction and the explanation may
be perfectly correct. And similarly, 'jealous' may be a term of
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 139
blame and may be used without any intention of blaming at all.

(Sometimes one has a perfect right to be jealous.)
The use of the word 'inconsiderate' seems to me a very fine
example of the way in which the fact/value distinction is hope-
lessly fuzzy in the real world and in the real language. The
importance of terms like 'inconsiderate', 'pert', 'stubborn',
'pesky', etc., in actual moral evaluation, has been emphasized by
Iris Murdoch in The Sovereignty of 'Good'.5 Even though each
of the statements 'John is a very inconsiderate man', 'John
thinks about nobody but himself, 'John would do practically
anything for money' may be simply a true description in the
most positivistic sense (and notice 'John would do practically
anything for money' does not contain any value term), if one has
asserted the conjunction of these three statements it is hardly
necessary to add 'John is not a very good person'. When we
think of facts and values as independent we typically think of
'facts' as stated in some physicalistic or bureaucratic jargon, and
the 'values' as being stated in the most abstract value terms, e.g.
'good', 'bad'. The independence of value from fact is harder to
maintain when the facts themselves are of the order of 'incon-
siderate', 'thinks only about himself, 'would do anything for
money'.
Just as we criticize a describer who does not employ the con-
cepts of table and chair when their use is called for, so also,
someone who fails to remark that someone is considerate or
spontaneous may open himself to the criticism that he is imper-
ceptive or superficial; his description is not an adequate one.
The super-Benthamites
Let me go back and modify my previous example of the 'Brain-
in-a-Vatists'. This time let us imagine that the continent of Aus-
tralia is peopled by a culture which agrees with us on history,
geography and exact science, but which disagrees with us in eth-
ics. I don't want to take the usual case of super-Nazis or some-
thing of that kind, but I want to take rather the more interesting
case of super-Benthamites. Let us imagine that the continent of
Australia is peopled with people who have some elaborate sci-
5
Routledge and Kegan Paul, 1970.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
140 Fact and value
entific measure of what they take to be 'hedonic tone', and who

believe that one should always act so as to maximize hedonic
tone (taking that to mean the greatest hedonic tone of the great-
est number). I will assume that the super-Benthamites are
extremely sophisticated, aware of all the difficulties of predicting
the future and exactly estimating the consequences of actions
and so forth. I will also assume that they are extremely ruthless,
and that while they would not cause someone suffering for the
sake of the greatest happiness of the greatest number if there
were reasonable doubt that in fact the consequence of their
action would be to bring about the greatest happiness of the
greatest number, that in cases where one knows with certainty
what the consequences of the actions would be, they would be
willing to perform the most horrible actions - willing to torture
small children or to condemn people for crimes which they did
not commit - if the result of these actions would be to increase
the general satisfaction level in the long run (after due allowance
for the suffering of the innocent victim in each case) by any pos-
itive e, however small.
I imagine that we would not feel very happy about this sort of
super-Benthamite morality. Most of us would condemn the
super-Benthamites as having a sick system of values, as being
bureaucratic, as being ruthless, etc. They are the 'new man' in
his most horrible manifestation. And they would return our
invective by saying that we are soft-headed, superstitious, pris-
oners of irrational tradition, etc.
The disagreement between us and the super-Benthamites is
just the sort of disagreement that is ordinarily imagined in order
to make the point that two groups of people might agree on all
the facts and still disagree about the 'values'. But let us look at
the case more closely. Every super-Benthamite is familiar with
the fact that sometimes the greatest satisfaction of the greatest
number (measured in 'utils') requires one to tell a lie. And it is
not counted as being 'dishonest' in the pejorative sense to tell lies
out of the motive of maximizing the general pleasure level. So
after a while the use of the description 'honest' among the super-
Benthamites would be extremely different from the use of that
same descriptive term among us. And the same will go for 'con-
siderate', 'good citizen', etc. The vocabulary available to the
super-Benthamites for the description of people-to-people situa-
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 141
tions will be quite different from the vocabulary available to us.

Not only will they lack, or have altered beyond recognition,
many of our descriptive resources, but they will very likely
invent new jargon of their own (for example, exact terms for
describing hedonic tones) that are unavailable to us. The texture
of the human world will begin to change. In the course of time
the super-Benthamites and we will end up living in different
human worlds.
In short, it will not be the case that we and the super-Bentham-
ites 'agree on the facts and disagree about values'. In the case of
almost all interpersonal situations, the description we give of the
facts will be quite different from the description they give of the
facts. Even if none of the statements they make about the situa-
tion are false, their description will not be one that we will count
as adequate and perspicuous; and the description we give will
not be one that they could count as adequate and perspicuous.
In short, even if we put aside our 'disagreement about the val-
ues', we could not regard their total representation of the human
world as fully rationally acceptable. And just as the Brain-in-a-
Vatists' inability to get the way the world is right is a direct result
of their sick standards of rationality — their sick standards of
theoretical rationality - so the inability of the super-Benthamites
to get the way the human world is right is a direct result of their
sick conception of human flourishing.
Subjectivism about goodness

It has often been claimed that the step from 'John is considerate,
truthful, kind, courageous, responsible, etc' to 'John is morally
good' involves at least one unproved (and unprovable) 'premiss',
namely, 'Consideration is morally good.' And it has been held
that the need for moral 'premisses' before one can draw moral
conclusions from 'factual' statements shows that ethical state-
ments are not rationally justifiable.
This picture of ethics as a sort of inverted pyramid, with the
tip (which is itself unsupported) consisting of 'ethical axioms'
which support our whole body of moral belief and thinking, is
naive. No one has ever succeeded in imposing an axiomatic
structure upon ethics (as Aristotle remarked in the passage I
cited a few pages ago, such moral maxims as we are able to list
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
142 Fact and value
are almost always true only 'for the most part'). And the same
trick, of picturing a body of thinking one wishes to cast into
doubt as resting upon unsupportable 'axioms' is one which scep-
tics have employed in every area. Sceptics who doubt the exis-
tence of material objects, for example, argue that the principle
that 'if our sensations occur as they would if there were a mate-
rial world, then there probably is a material world' is a rationally
unsupportable premiss which we tacitly invoke whenever we
claim to 'observe' a material object, or try otherwise to justify
belief in their existence. In fact, ethics and mathematics and talk
of material objects presuppose concepts not 'axioms'. Concepts
are used in observation and generalization, and are themselves
made legitimate by the success we have in using them to describe
and generalize.
A more sophisticated attack on the idea of ethical objectivity
concedes that our ethical beliefs rest on observations of specific
cases, 'intuitions', general maxims, etc., and not on some collec-
tion of arbitrary 'ethical axioms', but makes the charge that eth-
ical 'observation' itself is infected with an incurable disease: pro-
jection.
According to this account, humans are naturally, if intermit-
tently, compassionate. So when we see something terrible hap-
pening, as it might be, someone torturing a small child just for
his own sadistic pleasure, we are (sometimes) horrified. But the
psychological mechanism of 'projection' leads us to experience
the feeling quality as a quality of the deed itself: we say 'the act
was horrible' when we should really say 'my reaction was to be
horrified'. Thus we build up a body of what we take to be 'ethi-
cal observations', which are really just observations of our own
subjective ethical feelings.
This story has more sophisticated forms (like any other).
Hume postulated a human tendency he called 'sympathy', which
has gradually become wider under the influence of culture. Con-
temporary sociobiologists postulate an instinct they call 'altru-
ism', and speak of 'altruistic genes'. But the key idea remains the
same: there are ethical feelings, but no objective value proper-
ties.
We have already seen that this is not right: there are at least
some objective values, for example, justification. It could still be
claimed that the ethical values are subjective while the cognitive
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 143
values are objective; but the argument that there can't be any
objective values at all has been refuted.
In order to show what is wrong with arguments for moral

subjectivism, I must now recall the arguments that were used
against metaphysical realism in Chapter 2. This may seem queer:
isn't subjectivism the opposite of metaphysical realism? If one
thinks so, then it will seem that any argument against metaphys-
ical realism must support subjectivism; the strategy I am going
to follow of using the same argument against both metaphysical
realism and subjectivism will seem an impossible one.
But in fact, metaphysical realism and subjectivism are not sim-
ple 'opposites'. Today we tend to be too realistic about physics
and too subjectivistic about ethics, and these are connected ten-
dencies. It is because we are too realistic about physics, because
we see physics (or some hypothetical future physics) as the One
True Theory, and not simply as a rationally acceptable descrip-
tion suited for certain problems and purposes, that we tend to be
subjectivistic about descriptions we cannot 'reduce' to physics.
Becoming less realistic about physics and becoming less subjec-
tivistic about ethics are likewise connected.
The argument at the end of Chapter 2 was directed against the
'physicalistic' or naturalistic version of metaphysical realism. To
recall it, let us suppose that the standard interpretation / (Under
which 'cat' refers to cats, 'cherry' to cherries, etc.) is either coex-
tensive with or identical with physicalistic relation R. So R
holds between tokens of 'cat' (or physical events of someone's
using those tokens suitably) and cats, etc. The non-standard
interpretation / we described will then also be co-extensive with
a certain relation R', definable in terms of R and the possible
worlds and permutations used in constructing/ (see Appendix).
So Rf holds between tokens of 'cat' (or the physical events of
someone's using those tokens in the standard way) and cherries,
etc. R and R' are both 'correspondences': The same sentences
are 'true' under both correspondences. The actions called for by
the R '-truth of a sentence (i.e. the actions which will 'succeed',
from the agent's point of view) are the same as the actions called
for when the sentence is R-true. If R is 'identical with reference';
if R, R', and all the other relations which assign extensions to
our words in ways which satisfy our operational and theoretical
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
144 Fact and value
constraints are not equally correct; if R, Rf and the others are

not equally correct because one of them — R — just is reference;
then that fact itself is an inexplicable fact from a physicalist per-
spective.
This argument is not just an argument against (the physicalist
version of) metaphysical realism, but an argument against
reductionism. If there is nothing in the physicalist world-picture
that corresponds to the obvious fact that 'cat' refers to cats and
not to cherries, then that is a decisive reason for rejecting the
demand that all the notions we use must be reduced to physical
terms. For reference and truth are notions we cannot consis-
tently give up. If I think 'a cat is on a mat', then I am committed
to believing that 'cat' refers to something (though not to a meta-
physical realist account of 'reference') and to believing that 'a
cat is on a mat' is true (though not to a metaphysical realist
account of truth).
Having reviewed the argument of Chapter 2, let us now see
how it bears on the arguments for moral subjectivism. The 'pro-
jection' theory gave one account of moral experience: moral
experience is, so to speak, mislocated subjective feeling. Contrast
the 'projection' theory with the following account: 'all humans
have, to some extent, a sense of justice and some idea of the
good. So we respond (intermittently) to such appeals as "be kind
to the stranger among you, because you know what it was like
to be a stranger in Egypt"'. Our sympathy becomes broader,
partly because we are persuaded that it ought to be broader; we
feel that an atrocity is wrong (sometimes) even when we don't
easily or spontaneously find the victim a person we can sympa-
thize with. We come to see similarities between injuries to others
and injuries to ourselves, and between benefits to others and
benefits to ourselves. We invent moral words for morally rele-
vant features of situations, and we gradually begin to make
explicit moral generalizations, which lead to still further refine-
ment of our moral notions, and so on.'
This account is, on the face of it, simpler and more sophisti-
cated than the 'projection' theory. (For one thing, it acknowl-
edges the role of argument in shaping moral attitudes.) Never-
theless, many intelligent people feel that today we must reject
talk of a 'sense of justice' and talk of 'having an idea of the good'
(where this is not taken in a purely subjective sense), as 'unscien-
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 145
tific'. So moral knowledge becomes problematical; perhaps

downright impossible.
But what does 'unscientific' mean here? A belief that there is
such a thing as justice is not a belief in ghosts, nor is a 'sense of
justice' a para-normal sense which enables us to perceive such
ghosts. Justice is not something anyone proposes to add to the
list of objects recognized by physics as eighteenth-century chem-
ists proposed to add 'phlogiston' to the list of objects recognized
by chemical theory. Ethics does not conflict with physics, as the
term 'unscientific' suggests; it is simply that 'just' and 'good' and
'sense of justice' are concepts in a discourse which is not reduc-
ible to physical discourse. As we have just seen, other kinds of
essential discourse are not reducible to physical discourse and
are not for that reason illegitimate. Talk of 'justice', like talk of
'reference', can be wow-scientific without being wwscientific.
As a way of seeing what is going on, let us consider any basic

principle of logic or mathematics, say, the principle that the
series of whole numbers can always be continued ('every number
has a successor'), or the principle that a non-empty set of whole
numbers must contain a smallest member. Suppose someone put
forward the following view: 'These principles are true for the
numbers and sets of numbers we deal with in practice. So they
come to seem necessary. By the mechanism called "projection",
we attach this feeling of necessity to the principles themselves;
we feel the statements have a mysterious "necessity". But in real-
ity this has no justification. For all we know, these principles
may not even be true.'
Virtually no one would agree with this. Virtually every math-
ematician would say, instead, something like this: 'Most humans
have mathematical intuition to some extent. So we intuitively
"see", or can be brought by examples (or by skillful questioning,
like the slave-boy in Plato's dialogue) to "see" that the principles
are necessarily true.'
Kurt Godel believed that 'mathematical intuition' was analo-
gous to perception: mathematical objects (which he called 'con-
cepts') are out there, and our intuition enables us to intellectually
perceive these Platonic entities; but few mathematicians would
commit themselves to such a Platonic metaphysics. Godel's com-
parison of mathematical intuition to perception reveals an over-
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
146 Fact and value
simple idea of perception. Vision does not really give us direct

access to a ready made world, but gives us a description of
objects which are partly structured and constituted by vision
itself. If we take the physicist's rainbow to be the rainbow 'in
itself, then the rainbow 'in itself has no bands (a spectroscopic
analysis yields a smooth distribution of frequencies); the red,
orange, yellow, green, blue and violet bands are a feature of the
perceptual rainbow, not the physicist's rainbow. The perceptual
rainbow depends on the nature of our perceptual aparatus itself,
on our visual 'world making' as Nelson Goodman has termed it.
(The physicist's 'objects' also depend on our worldmaking, as is
shown by the plethora of radically different versions physics
constructs of the 'same' objects.) Yet we do not consider vision
as defective because it sees bands in the rainbow; someone who
couldn't see them would have defective vision. Vision is certified
as good by its ability to deliver a description which fits the
objects for us, not metaphysical things-in-themselves. Vision is
good when it enables us to see the world 'as it is' — that is, the
human, functional world which is partly created by vision itself.
A proposed new axiom of set theory, such as the 'Axiom of
choice', may be adopted partly because of its agreement with the
'intuition' of expert mathematicians, and partly for its yield.
If the axiom of choice did not deliver results which count as
successful mathematics the fact that some people find it 'intu-
itive' would have little interest. Mathematical intuition itself is
demonstrated or tested by grasping mathematical principles and
by following proofs. In short, mathematical intuition is good
when it enables us to see mathematical facts 'as they are' — that
is, as they are in the mathematical world which is constructed by
human mathematical practice (including the application of
mathematics to other subject matters).
A physiological or psychological description of vision cannot
tell us whether seeing bands in the rainbow counts as seeing 'cor-
rectly' or not. Even less could a physiological or psychological
description of the brain-process which goes on when one 'grasps'
the Principle of Mathematical Induction tell us whether that
principle is true or not. Once one sees this, it should be no sur-
prise that a description of the brain process which goes on when
one 'sees' that an action is unjust cannot tell us whether the
action really is unjust.
Talk of moral 'perception', like talk of mathematical intuition,
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 147
or of reference and understanding, is not reducible to the lan-

guage or the world-picture of physics. That does not mean phys-
ics is 'incomplete'. Physics can be 'complete' — that is, complete
for physical purposes. The completeness physics lacks is a com-
pleteness all particular theories, pictures, and discourses lack.
For no theory or picture is complete for all purposes. If the irre-
ducibility of ethics to physics shows that values are projections,
then colors are also projections. So are the natural numbers. So,
for that matter, is 'the physical world'. But being a projection in
this sense is not the same thing as being subjective.6
Authoritarianism and pluralism

I have been arguing that it is necessary to have standards of
rational acceptability in order to have a world at all; either a
world of 'empirical facts' or a world of 'value facts' (a world in
which there is beauty and tragedy). It should go without saying
that it is not possible both to have standards of rational accept-
ability and not to accept them, or to stand at arm's length from
them. (The kind of scepticism which consists in refusing to have
any standards of rational acceptability commits one to not hav-
ing any concepts at all. As Sextus Empiricus recognized, that
kind of empiricism ultimately is unexpressible in language.) We
have just as much right to regard some 'evaluationaP casts of
mind as sick (and we all do) as we do to regard some 'cogni-
tionaP casts of mind as sick. But to say this is not to reject plu-
ralism or to commit oneself to authoritarianism.
Even in science, holding that science is an objective enterprise
(by a standard of 'objectivity' which is admittedly anthropocen-
tric, but, as David Wiggins once remarked, the only standard of
6
An unintentionally funny version of the projection theory is cited by
C. S. Lewis in The Abolition of Man (Macmillan, 1947). Lewis quotes a
secondary school English text (which he does not identify, out of
charity). 'You remember that there were two tourists present [Lewis is
talking about the well known story of Coleridge at the waterfall]-, that
one called it "sublime" and the other "pretty"; and that Coleridge
mentally endorsed the first judgment and rejected the second with
disgust. Gaius and Titius [Lewis' pseudonyms for the unidentified
authors of the text] comment as follows: "When the man said That is
sublime, he appeared to be making a remark about the waterfall . . .
Actually he was not making a remark about the waterfall, but about his
own feelings. What he was saying was really / have feelings in my mind
associated with the word 'Sublime', or shortly, I have sublime feelings." '
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
148 Fact and value
objectivity available to us) is not to hold that every scientific

question has a determinate answer. Some scientific questions
may have objectively indeterminate answers, i.e. there may be no
convergence with respect to an answer to them even in the ideal
limit of scientific inquiry; and some scientific questions may have
determinate but context-relative answers (e.g. 'What was the
cause of John's heart attack?' may have different correct answers
depending on who asks the question and why). And, similarly,
holding that ethical inquiry is objective in the sense that some
'value judgments' are definitely true and some are definitely
false, and more generally that some value positions (and some
'ideologies') are definitely wrong, and some are definitely infe-
rior to some others, is not the same thing as holding the silly
position that there are no indeterminate cases at all. (One espe-
cially important kind of indeterminate case has been emphasized
by Bernard Williams: this is the case where all the alternatives
are so horrible that there is no one of the alternatives that would
clearly be chosen by an ideally rational and wise person.) And
that there are context relativities in ethics goes without saying.
If today we differ with Aristotle it is in being much more plu-
ralistic than Aristotle was. Aristotle recognized that different
ideas of Eudaemonia, different conceptions of human flourish-
ing, might be appropriate for different individuals on account of
the difference in their constitution. But he seemed to think that
ideally there was some sort of constitution that every one ought
to have; that in an ideal world (overlooking the mundane ques-
tion of who would grow the crops and who would bake the
bread) everyone would be a philosopher. We agree with Aris-
totle that different ideas of human flourishing are appropriate
for individuals with different constitutions, but we go further
and believe that even in the ideal world there would be different
constitutions, that diversity is part of the ideal. And we see some
degree of tragic tension between ideals, that the fulfillment of
some ideals always excludes the fulfillment of some others. But
to emphasize the point again, belief in a pluralistic ideal is not
the same thing as belief that every ideal of human flourishing is
as good as every other. We reject ideals of human flourishing as
wrong, as infantile, as sick, as one-sided.
Nor should commitment to ethical objectivity be confused
with what is a very different matter, commitment to ethical or
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
Fact and value 149
moral authoritarianism. It is perhaps this confusion that has lead

one outstanding philosopher7 to espouse what he himself
regards as a limited version of 'non-cognitivism', and to say
'Concerning what "living most fully" is for each man, the final
authority must be the man himself.' (Notice the ambiguity in
'the final authority': does he mean the final political authority?
The final epistemological authority? Or does he mean that there
is no fact of the matter, as his use of the term 'non-cognitivism'
suggests?) Respect for persons as autonomous moral agents
requires that we accord them the right to choose a moral stand-
point for themselves, however repulsive we may find their
choice. According to the philosophy of political liberalism, it
also requires that we also insist the government not preempt
individual moral choices by setting up a state religion or a state
morality. But diehard opposition to all forms of political and
moral authoritarianism should not commit one to moral relativ-
ism or moral scepticism. The reason that it is wrong for the gov-
ernment to dictate a morality to the individual citizen is not that
there is no fact of the matter about what forms of life are fulfill-
ing and what forms of life are not fulfilling, or morally wrong in
some other way. (If there were no such thing as moral wrong,
then it would not be wrong for the government to impose moral
choices.) The fact that many people fear that if they concede any
sort of moral objectivity out loud then they will find some gov-
ernment shoving its notion of moral objectivity down their
throats is without question one of the reasons why so many peo-
ple subscribe to a moral subjectivism to which they give no real
assent.
7
David Wiggins in 'Truth, Invention, and the Meaning of Life', Proceed-
ings of the British Academy, vol. LXII, 1976.
Cambridge
https://doi.org/10.1017/CBO9780511625398.008
7
Reason and history
With the rise of science has come the realization that many ques-
tions cannot be settled by the methods of the exact sciences, ide-
ological and ethical questions being the most obvious examples.
And with the increase in our admiration and respect for the
physicist, the cosmologist, the molecular biologist, has come a
decrease in respect and trust for the political thinker, the moral-
ist, the economist, the musician, the psychiatrist, etc.
In this situation some have gone with the cultural tide and
argued that, indeed, there is no knowledge to be found outside
of the exact sciences (and the social sciences to the extent that
they succeed in aping the exact sciences, and only to this extent).
This view may take the form of positivism or materialism, or
some combination of these. Others have tried to argue that sci-
ence too is 'subjective' and arbitrary — this is the popular reading
of Kuhn's immensely successful book The Structure of Scientific
Revolutions, even if it is not the one Kuhn now says he intended.
Others — e.g. the Marxist philosophers and the religious philos-
ophers - adopt a sort of double-entry bookkeeping, leaving
technical questions to the exact sciences and engineering and ide-
ological or ethical questions to a different tribunal: the Party,
the Utopian future, the church. But few can feel comfortable
with any of these stances - with extreme scientism in either its
positivist or materialist forms, with subjectivism and radical rel-
ativism, or with any of the species of double-entry bookkeeping.
It is just because we feel uncomfortable that there is a real prob-
lem for us in this area.
To be sure, the problem is in one way wwreal. The same person
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
Reason and history 151
who argues that ethical and political opinions are unverifiable

argues with passion for his ethical and political opinions. Hume
said that he left his scepticism whenever he left his study; and
relativists are likely to do the same with their relativism. But this
only shows that no one can consistently live by relativism; if this
is all that can be said in response to relativism, then we are just
pushed from relativism to 1945 style existentialism ('it's all
absurd, but you have to choose'). And is that so different?
In order to fix our ideas, let us recall a remark by a philoso-
pher of the last century whose Utilitarianism actually covered a
good bit of relativism. I am thinking of Bentham, and of Ben-
tham's challenging judgment that 'prejudice aside, the game of
pushpin is of equal value with the arts and sciences of music and
of poetry'. Prejudice aside, pushpin is as good as poetry.
What makes this so shocking to the modern reader is how
deeply it conflicts with our current cultural values. The arts have
been exalted by us to a place much higher than any they occu-
pied in Plato's day or in the middle ages. As a number of authors
have remarked, for a certain sort of educated person, art today
is religion, i.e. the closest thing to salvation available.
Bentham is saying that a preference for 'the arts and sciences
of music and poetry' over the childish game of pushpin is merely
subjective, like a preference for vanilla ice cream over chocolate
ice cream. He does not wish to deny that music and poetry do
have greater value than pushpin ('prejudice aside' is an impor-
tant part of the sentence); in the context of his Utilitarianism,
the very fact that a large majority do prefer music and poetry to
pushpin gives music and poetry greater 'utility' and hence
greater value. But the value is, as it were, the product of 'preju-
dice' (i.e. purely subjective interest); there is no fact of the matter
about the relative value of pushpin and poetry apart from the
fact that people prefer poetry to pushpin. We don't prefer poetry
to pushpin because poetry has greater value than pushpin, Ben-
tham is saying, rather, it's the other way around, and poetry has
greater value than pushpin because people prefer poetry to push-
pin. (For no reason apparently.)
Stating the position so baldly already makes it look implausi-
ble. Let us consider for the moment a really 'subjective' prefer-
ence.
One model that people sometimes seem to have in mind for
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
152 Reason and history
subjective preference is this. There is something C which is the

taste of chocolate ice cream; there is something V which is the
taste of vanilla ice cream. There are two feelings L, D which are
'liking' and 'disliking'. And what goes on and all that goes on,
when Jones likes and Smith dislikes vanilla (and Smith likes and
Jones dislikes chocolate), is that Jones experiences V + L when
he eats vanilla and C +D when he eats chocolate whereas Smith
experiences V + D when he eats vanilla and C +L when he eats
chocolate.
Such an account is naive psychologically, however, as Kohler
long ago argued. What vanilla tastes like to Jones, who likes
vanilla ice cream, is not what it tastes like to Smith, who can't
stand vanilla ice cream. Rather it's like this: Call the quality
vanilla has for Smith Vs. Vs is an 'unpleasant' taste; it may be
imaginable that one could experience Vs and like it, but just
barely, and, even if one did, there would be some kind of disas-
sociation or repression. In short, psychologically if not meta-
physically, Vs is 'intrinsically' unpleasant. And Smith feels D
(dislike for the taste) when he eats vanilla because vanilla has the
taste quality Vs (for him). Similarly, Vj9 the taste quality of
vanilla for Jones is intrinsically 'pleasant' (which is why Jones
feels L, liking). In the language of G. E. Moore, the taste Vj and
the positive value are an organic unity for Jones, and the taste Vs
and the negative value are an organic unity for Smith. Phenom-
enologically, they cannot really ever be separated into two parts
in the way the notation (VS + D', {Vj + U suggests. Almost cer-
tainly (barring special factors of repression or disassociation),
Smith would like vanilla ice cream too if it evoked Vj in his
mouth and not Vs t and Jones would dislike vanilla ice cream too
if it evoked Vs in his mouth and not Vj.
Why do we regard the preference for vanilla over chocolate as
'subjective' then? I mean, why do we regard it as subjective even
when we don't think all value judgments are subjective or agree
with Bentham that, prejudice aside, pushpin is as good as
poetry? Obviously, if we think all preferences are subjective, we
will think this one is too, but the interesting question is why this
judgment doesn't even seem objective to us unless, perhaps, we
are Smith or Jones, why it doesn't have the kind of objectivity
that many value judgments do undeniably have.
It isn't just that there is disagreement. If we think there are
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
objective (or warranted) value judgments at all, very likely we

think some hotly disputed judgments are objectively right. The
Nazis disputed the judgment that wanton killing of Jews just
because of their racial affiliation is wrong, but anti-Nazis did not
regard their disagreement with the Nazis over this judgment as
'subjective'. Those who think homosexuals should have full
rights in our society violently disagree with those who think
homosexual activity or civil rights of homosexuals should be
legally proscribed; but neither side in this dispute regards its own
position as 'subjective'. Indeed, disagreement frequently makes
people more sure that their moral position is warranted. So it
isn't just the fact that 'some people prefer chocolate and some
people prefer vanilla' that makes the Smith/Jones disagreement
in preference subjective.
Part of the story may be that most people don't have strong
preferences between vanilla and chocolate, but this cannot be
decisive. If half of the population couldn't stand chocolate but
liked vanilla, and the other half couldn't stand vanilla but liked
chocolate, we would still (if we were reasonable about our pref-
erences) regard this as a 'matter of taste', i.e. as subjective. It
isn't the existence of 'neutrals' that is decisive.
What is decisive, in my opinion, is that whatever biological or
psychological idiosyncrasies are responsible for Smith's and
Jones' preferences are not correlated with important traits of
mind and character. If we try the thought experiment of imag-
ining the contrary, of imagining that there was a caste of char-
acter that we regarded as good, both for its own sake and
because of its effects on feeling, judgment and action, and
another caste of character that we regarded as bad, both in itself
and for its effects, and that everyone knew that the good caste
of mind and character always revealed itself in a preference for
vanilla and the bad in a preference for chocolate, then I think we
will find that the more vividly we can succeed in making this
case real to ourselves, the more we will feel that in such a world
the first preference would be seen as 'normal' and 'right' and the
second as 'perverse' or 'monstrous' or something of that sort.
I don't mean to claim that all preferences are judged morally
by the traits of character they are thought to express. Some 'pref-
erences' are terribly important in themselves: someone who
thought it was just wonderful to torture small children for the
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
fun of it would (if he was serious) be condemned on the basis of

that one attitude. But if the matter preferred is not regarded as
important in itself, then whether we make an issue of the pref-
erence or take it to be 'a matter of taste' will generally depend
on what, if anything, we think the preference shows. Value judg-
ments often come in clumps; and clumps of value judgments fre-
quently express durable traits of mind, personality, and charac-
ter. The independence of 'I prefer vanilla to chocolate ice
cream' from any interesting and significant 'clump' of this kind
is just what makes it 'subjective' (along, of course, with the
absence of any intrinsic importance to the choice itself).
Even if Smith's preference for vanilla is 'subjective', that does
not make it irrational or arbitrary. Smith has a reason - the best
possible reason — for liking vanilla, namely the way it tastes to
him. Values can be 'subjective' in the sense of being relative and
still be objective; it is objective that vanilla tastes better than
chocolate to Smith. In The Sovereignty of Good, Iris Murdoch
pointed out that philosophers as different as the French Existen-
tialists and the logical positivists actually shared a common
model of value judgment, the model of reason as supplying the
mind with neutral 'facts' on the basis of which the will must
arbitrarily choose 'values' — the choice of values must be arbi-
trary, precisely because 'facts' are (by definition) neutral. But,
since the will is given no clues by reason as to how to judge
(reason only supplies 'facts', on this picture) it has no reason for
its arbitrary choice; which is why the French philosophers called
it 'absurd', and why more naturalistically inclined philosophers
see instinct and emotion (the historic successors to Bentham's
all-purpose category of pleasure) as the ultimate basis of moral
choice.
In the case we have just examined, the Existentialist— Positivist
model does notfithowever. The 'fact' — the taste itself— and the
'value' — the goodness of the taste — are one, at least psycholog-
ically. Presented experiential qualities aren't, in general, neutral
and they frequently seem to demand responses and attitudes.
One may override these felt demands for good and sufficient rea-
son, as when a child learns to bear the pain of an injection for
the sake of the benefits conferred by the immunizing agent
injected, but the prima facie goodness and badness of particular
experiences can hardly be denied. (Interestingly enough, this
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
point was recognized by Plato and the medievals — we are per-

haps the first culture to conceive of experience as neutral).
The non-neutrality of experience also bears on the
pushpin/poetry case. We find it virtually impossible to imagine
that someone who really appreciates poetry, someone who is
capable of distinguishing real poetry from mere verse, capable of
responding to great poetry, should prefer a childish game to arts
which enrich our lives as poetry and music do. We have a reason
for preferring poetry to pushpin, and that reason lies in the felt
experience of great poetry, and of the after effects of great
poetry - the enlargement of the imagination and the sensibility
through the enlargement of our repertoire of images and meta-
phors, and the integration of poetic images and metaphors with
mundane perceptions and attitudes that takes place when a
poem has lived in us for a number of years. These experiences
too are prima facie good- and not just good, but enobling, to
use an old fashioned word.
That there can be reasons for value judgments — reasons
which really are good reasons for particular people to make par-
ticular value judgments—does not mean that all value judg-
ments are rational, of course. Value judgments, judgments that
people have cared passionately about, and in whose name people
have killed and tortured other people, have often enough been
based on an unwholesome mixture of aggressive impulses and
narcissistic ideas. Not surprisingly, when a relativist
historian/philosopher like Michel Foucault writes about the past,
he often focusses our attention on these irrational ideas and
value judgments. But it is important to see why he does this.
Foucault writes about the early modern era (the sixteenth and
seventeenth centuries) and about ideology and culture in general.
His knowledge of fact is legendary, even though many specialists
dispute Foucault's 'facts'. While some of his books are highly
abstract (e.g. The Archeology of Knowledge), some are rather
specific, e.g. The Birth of the Clinic and The Birth of the Prison.
The Birth of the Clinic is perhaps Foucault's best book, and it
makes an important part of the case for Foucault's more abstract
theories.
What Foucault tries to show is that the 'clinic', i.e. the insti-
tution of the hospital and the related medical institutions, was
the reflection of the growth of a certain ideology about disease
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
and health as much as the result of any increase in scientific

knowledge and technique. This ideology, in turn, was connected
with wider ideological changes, especially with the growth of
individualism in the seventeenth century. And he suggests both
that the 'clinic' is not a very good way of treating most patients
and that our belief that it is is a kind of ideological prejudice, in
short, a kind of folly.
The wider suggestion that emerges is that ideological convic-
tions and the associated value judgments are a rather arbitrary
affair.1 There is no objective place to stand in ideological matters
(except of course, for the mysterious standpoint of Foucault's
own allegedly objective 'Archeology of Knowledge').
To see what Foucault is driving at, let us consider a more
familiar and less controversial example. In the Middle Ages, it
was believed that monarchy was the natural and proper form of
government. This belief was based partly on factual beliefs now
thought to be unwarranted (e.g. that democracy would inevita-
bly lead to anarchy and tyranny), and partly on the authority of
the Church. The view of the Church was in fact based partly on
political considerations (the Church was the state religion), but
this was not perceived because the Church itself was thought to
be the divinely inspired and divinely appointed interpreter of the
word and will of God. What Foucault is suggesting is that beliefs
held in the recent past, and, by implication, the beliefs we hold
right now, are no more rational than the medieval belief in the
Divine Right of Kings.
Let us consider the belief in the Divine Right of Kings for a
moment. If we don't think there is a good reason to believe in
the existence of a personal God, one who commands that we live
by certain kinds of social forms and structures, then this belief
will be immediately stamped as irrational (which is not to deny
that the belief answered to real psychological needs). Even if we
1
And also that we are determined in our thinking by the very language
we use. Foucault speaks of 'implicit systems which determine our
most familiar behavior without our knowing it' (see J. Simon, 'A
Conversation with Michel Foucault', Partisan Review, No. 2, 1971,
p. 201). French Structuralism, at least as represented by Foucault,
Althusser, Lacan, Deleuze, etc., often seems to amount to (1) deter-
minism; (2) relativism; and (3) claims that Structuralism is 'linguistic
science'.
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
believe in God, if we don't believe that the Church has special

access to his wishes, we will think the Divine Right of Kings was
and is an irrational doctrine. And finally, even believing Catho-
lics will concede that the Church's support for monarchy in the
Middle Ages was based as much on political considerations as
on revelation or sound theology. In short, the belief in the Divine
Right of Kings lacks, and always lacked, an adequate rational
basis.
How, then, did the belief arise? The usual answer would
appeal partly to political and economic factors (one does not
have to be a Marxist to concede that these factors are among the
determinants of ideology) and partly to psychological factors.
The comfort provided by belief in a personal God and a here-
after is obvious, and so, perhaps, is the comfort provided to the
believer in an infallible Church and a divinely appointed social
order. In short, narcissistic ego-gratification and social condi-
tioning were the real determinants of this belief. And, if this
belief is really typical, if it is really representative of all our 'ide-
ological' beliefs, then such factors are the real determinants of
all ideology.
It is because they believe something like this that so many
modern French thinkers hold Marx, Freud, and Nietzsche in
such high esteem. Marx, Freud, and Nietzsche have this in com-
mon: they see our cherished religious and ethical ideas as reflec-
tions of the irrational, of class interest (in the case of Marx), of
the unconscious (Freud and Nietzsche), of the Will to Power
(Nietzsche).2 Below what we are pleased to regard as our most
profound spiritual and moral insights lies a seething cauldron of
power drives, economic interests, and selfish fantasies. This is
the view that is the cutting edge of relativism today.
At the same time, no relativist can himself use the term 'irra-
tional' in the way I have just used it in describing the relativist
view. Such a use is ruled out by relativism itself. When I showed
these pages to a relativist friend he was indignant at my state-
ments about the Divine Right of Kings. Was I not aware that
intelligent men had been convinced by philosophical arguments
2
I am not accusing Marx, Freud, or Nietzsche of drawing relativist
conclusions from this.
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
that this doctrine was correct? Was I offering a cheap Marxist or

Freudian explanation? Of course, belief in the Divine Right of
Kings was rational!
My reply to him was that there may be a sense of 'rational' in
which any view that can be intelligently and persuasively
defended from the shared assumptions of a culture is a 'rational'
view, but that sense cannot be the only or the normatively
important one. The Jews accepted Moses as lawgiver and
prophet because his doctrine filled real religious, cultural, and
national needs; that is not the same thing as being convinced by
rational argument. Later, prophets anointed Jewish kings (after
trying to dissuade the Jews from having kings at all); that hardly
proves that later Christian kings are divinely appointed. Chris-
tianity, which shared the Jewish bible, became the religion of the
Roman Empire - hardly because the population or the Emperor
had rational proof that Christianity was true. Roman emperors
were then anointed (as Jewish Kings had been); that hardly
proves that they were divinely appointed. Finally, after the
assumptions of Christianity had been accepted, one could give
'rational arguments' for the Divine Rights of Kings from those
assumptions. But to express this by saying that in the late Roman
Empire or the Middle Ages, 'belief in the Divine Right of Kings
was perfectly rational' is to debase the notion of rationality.
Hegel, who introduced the idea that Reason itself changes in
history, operated with two notions of rationality: there is a sense
in which what is rational is measured by the level to which Spirit
has developed in the historical process at a given time; it is in
something like this sense that it is claimed by some that 'belief in
the Divine Right of Kings was rational at the time'. And there is
a limit notion of rationality in Hegel's system; the notion of that
which is destined to be stable, the final self-awareness of Spirit
which will not itself be transcended. When present day relativists
'naturalize' Hegel by throwing away the limit-concept of true
rationality, they turn the doctrine into a self-defeating cultural
relativism.
No relativist wants to be a relativist about everything how-
ever. How do these French thinkers put limits on their own rel-
ativism? The answer varies with the thinker. In the case of a
Marxist like Althusser, the answer adopted is a version of the
'class interest' theory: all 'ideologies' are the product of non-
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
rational factors, but those ideologies that are the product of the
interests of the working class are (in the present era) 'just', and
tend in the direction of human liberation, while those ideolo-
gies that spring from the interests of the exploiting class are
'unjust' and produce misery. But Althusser distinguishes himself
from previous expounders of this class-relativist view by refusing
to say that even Marxist ideology ('working class' ideology) is
true or closer to the truth than bourgeois ideology. Ideologies
can be 'just' or 'unjust' according to Althusser, but not true or
false.3 ('True' and 'false' apply, he says, in 'laboratory science',
and, presumably, to those ordinary empirical statements that
have clear empirical test conditions.) Foucault also seems to be
moving towards a class-interest view in his most recent work,
although it is hard to be sure. The point of such a view, at least
in its radical Althusserian form, is that it seeks to preserve the
radical relativist claim that no 'ideology' can be rational while
saving the idea that some ideologies (the preferred one — Marx-
ism—Leninism in the case of Althusser) can be good by distin-
guishing between good and bad or 'just' and 'unjust' ideologies
on grounds other than rational acceptability. The idea is that
although all ideologies are adopted for irrational or non-rational
causes, some non-rational causes (working class interests) are
good, and produce good ideologies (by definition?) and some
non-rational causes are bad and produce bad ideologies. Instead
of judging ideologies by their reasons (which are always ratio-
nalizations) we should judge them by their causes.
This way of limiting one's own relativism is clearly unwork-
able however. For on what is the judgment based that the victory
of 'working class interests' will lead to such manifestly desirable
consequences as a world free from war and racism, and not to
totalitarianism and imperialism disguised as 'socialism'? If the
3
According to Althusser, 'Philosophical propositions are Theses.'
'Philosophical Theses can be held negatively as dogmatic propositions,
insofar as they are not susceptible of demonstration in strict scientific
sense of the term (in which one talks of demonstration in mathematics
or in logic), nor of proof in the strict scientific sense (in which one talks
of proof in the experimental sciences) . . . Philosophical Theses, since
they can neither be demonstrated nor scientifically proved, cannot be
said to be "true" (demonstrated or proved, as in mathematics and in
physics). They can only be said to be "just" ', Philosophie et Philosophie
Spontanee des Savants, pp. 13—14, Maspero (1967).
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
answer is that the latter wouldn't be 'true' socialism, or reflect

'true' working class interests, then on what is the judgment
based that any particular institution (e.g. the French Communist
Party, of which Althusser is a leading member) or policy will
promote 'true' working class interests and 'true' socialism? If
these beliefs can be rationally justified, then not every ideology
is irrational; if they cannot be, then the claim that any institution
or policy is 'just' must be as utterly irrational as every other 'ide-
ological' claim is asserted to be. If all human thought about ide-
ological questions is self-serving folly, then thought about which
beliefs spring from 'working class interests' and which do not
must also be self-serving folly.
Coming back to Foucault, however, and ignoring the signs in
Foucault's most recent work that he too is becoming radicalized,
his motive for focussing on the cases he chooses is precisely to
suggest the utterly non-rational (and, in fact, irrational) character
of the real reasons that people have for adopting ideological
positions. And the notion of the ideological here is very wide; it
is not just Communism, Fascism, Democracy, the Divine Right
of Kings, etc., that are under discussion. The belief that someone
is 'diseased' and needs a 'cure', the belief that someone is a 'crim-
inal' and should, if possible, be 'rehabilitated', and many, many
more of our everyday beliefs are 'ideological' in the sense of
these thinkers. Indeed, to the eagle eye of the Marxist sociologist
or the French philosopher, almost every belief is 'ideological'.
Perhaps, 'if I drop this glass, it will break' is ideologically neu-
tral, but little else is.
It may seem that I have missed the real point of what Foucault
is saying. His real point, he himself would say, is not that the
ideological perspectives of the past were foolish or irrational at
all, but rather that all ideology in the very wide sense in which
he uses the term, including our present ideology, is culture-
relative. He is trying to show us how every culture lives, thinks,
sees, makes love, by a set of unconscious guiding assumptions
with non-rational determinants. If previous ideologies now seem
'irrational' it is because we judge them by our culture-bound
notion of rationality.
What is troubling about Foucault's account is that the deter-
minants he and other French thinkers point to are irrational by
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
our present lights. If our present ideology is the product of forces

that are irrational by its own lights, then it is internally incoher-
ent. The French thinkers are not just cultural relativists; they are
attacking our present notion of rationality from within, and this
is what the reader feels and is troubled by.
Cultural relativism is not a new doctrine at all. Anthropolo-
gists have been preaching cultural relativism to us as long as
there have been anthropologists. But it would be a mistake to
assimilate the relativism of Foucault to the older relativism.
When an anthropologist preaches relativism to us, normally
he cites practices and beliefs of an exotic tribe which initially
strike us as irrational or repulsive or both, and proceeds to show
that these actually promote welfare and social cohesion. In
short, he shows (to the extent that the example is a reasonable
one) that what is considered wrong or irrational in our society
may actually be reasonable and right in different natural and
social circumstances.
Of course, the wrong conclusion is frequently drawn by
anthropologists from their own examples (and some of the
examples are rather less clear than the anthropologist thinks).
Very often an anthropologist will say 'it's all relative', meaning
that there is no fact of the matter as to what is right and wrong
at all. And Richard Boyd has suggested to me that very often the
motive is a political one; to convince us to stop destroying pri-
mitive cultures by attacking our belief in the superior rationality
and morality of our own. Unfortunately, the argument is very
confused. The anthropologist's examples (when they are good
ones) show that right and wrong, for example, are relative to
circumstances, not that there is no right and wrong at all, even
in specified circumstances. His argument against cultural impe-
rialism amounts to this: other cultures are not objectively worse
than ours (because there is no such thing as objective better and
worse, according to him); therefore they are just as good as ours;
therefore it is wrong to destroy them.
This argument equivocates. The conclusion requires that 'just
as good as' means objectively just as good as (at least by our
lights); but what follows from the non-existence of objective val-
ues cannot be that everything is (in the required sense) 'just as
good' as anything else, but rather that there is no such thing as
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
'just as good as'. If values really were arbitrary, then why should
we not destroy whatever cultures we please?
Fortunately, there are better grounds for criticizing cultural
imperialism than the denial of objective values. The anthropol-
ogist's motive may be a good one, but he has chosen the wrong
argument. Another term on which he equivocates is the notion
of being 'relative'. What his examples actually confirm is
Dewey's 'objective relativism'. Certain things are right -objec-
tively right - i n certain circumstances and wrong- objectively
wrong — in others, and the culture and the environment consti-
tute relevant circumstances. About this the anthropologist is
right. But this is not the same thing as values being 'relative' in
the sense of being mere matters of opinion or taste.
Still, freed of its conceptual confusions, the anthropologist's
argument should not trouble us. We should welcome his obser-
vations, for they tend to widen our sensibilities and attack our
smug assumption of cultural superiority. But the very compari-
son of Foucault's argument with the anthropologist's brings out
their difference: Foucault is not arguing that past practices were
more rational than they look to be, but that all practices are less
rational, are, in fact, mainly determined by unreason and selfish
power. The similarity of this doctrine to the older cultural rela-
tivism is a superficial one.
The fact is that the position we have been discussing caters to
an intellectual temptation which is the product of our increased
knowledge about and sensitivity to psychological and sociologi-
cal mechanisms. The knowledge and the sensitivity are in part
pretense and in part real; the temptation is to fall into the trap
of concluding that all rational argument is mere rationalization
and then proceeding to try to argue rationally for this position.
If all 'rational argument' were mere rationalization, then not
only would it make no sense to try to argue rationally for any
view, but it would make no sense to hold any view. If I view my
own assent and dissent as crazy behavior, then I should stop
assenting and dissenting — something to which there can be no
rational assent or dissent, only crazy parody of rational discus-
sion, cannot be called a statement. Like Sextus Empiricus, who
eventually concluded that his own scepticism could not be
expressed by a statement (because even the statement, 'I do not
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
know' could not be one he knew), the modern relativist, were he
consistent (and how could one consistently hold a doctrine
which makes nonsense of the notion of consistency?) should end
by regarding his own utterances as mere expression of feeling.
To say this is not to deny that we can rationally and correctly
think that some of our beliefs are irrational. It is to say that there
are limits to how far this insistence that we are all intellectually
damned can go without becoming unintelligible. We do, for
example, discuss just such doctrines as those advanced by Fou-
cault; we make an effort to be impartial; we try to adopt what
Popper calls 'the critical attitude', and actively to seek evidence
and argumentation we might overlook, even when it bears
against our own views. None of this would make the slightest
sense if we did not think that these practices of discussion and
communication, and these virtues of criticism and impartiality
tend to weed out irrational beliefs, if not at once, then gradually,
over time, and to improve the warranted assertibility of our final
conclusions. Rationality may not be defined by a 'canon' or set
of principles, but we do have an evolving conception of the cog-
nitive virtues to guide us.
It will be objected that this conception does not 'get us very
far'. Rudolf Carnap and John Cardinal Newman were both
careful and responsible thinkers, and both were committed to
the cognitive virtues just mentioned, but no one thinks that one
could have convinced the other, had they lived at the same time
and been able to meet. But the fact that there is no way to resolve
all disputes to everyone's satisfaction does not show that there is
no better and worse in such a case. Most of us think that New-
man's Catholicism was somewhat obsessive; and most philoso-
phers think that, brilliant as he was, Carnap employed many
weak arguments. That we make these judgments shows that we
do have a regulative idea of a just, attentive, balanced intellect,
and we do think that there is a fact of the matter about why and
how particular thinkers fall short of that ideal. Some will say,
'So what; we are no better off when it comes to resolving an
actual dispute than if there were no notion of rational accepta-
bility external to the views under debate to which we could
appeal!' This is true when it comes to any one unresolvable dis-
pute such as the Carnap—Newman dispute just imagined; but it
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
is not true that we would be just as well off in the long run if we
abandoned the idea that there are really such things as impar-
tiality, consistency, and reasonableness, even if we only approx-
imate them in our lives and practice, and came to the view that
there are only subjective beliefs about these things, and no fact
of the matter as to which of these 'subjective beliefs' is right.
Perhaps the analogy I have (occasionally) drawn between phil-
osophical discussion and political discussion may be of help.
One of my colleagues is a well-known advocate of the view that
all government spending on 'welfare' is morally impermissible.
On his view, even the public school system is morally wrong. If
the public school system were abolished, along with the compul-
sory education law (which, I believe, he also regards as an imper-
missible government interference with individual liberty), then
the poorer families could not afford to send their children to
school and would opt for letting the children grow up illiterate;
but this, on his view, is a problem to be solved by private charity.
If people would not be charitable enough to prevent mass illit-
eracy (or mass starvation of old people, etc.) that is very bad,
but it does not legitimize government action.
In my view, his fundamental premisses — the absoluteness of
the right to property, for example — are counterintuitive and not
supported by sufficient argument. On his view I am in the grip
of a 'paternalistic' philosophy which he regards as insensitive to
individual rights. This is an extreme disagreement, and it is a
disagreement in 'political philosophy' rather than merely a
'political disagreement'. But much political disagreement
involves disagreements in political philosophy, although they are
rarely as stark as this.
What happens in such disagreements? When they are intelli-
gently conducted on both sides, sometimes all that can happen
is that one sensitively diagnoses and delineates the source of the
disagreement. Often, when the disagreement is less fundamental
than the one I described, both sides may modify their view to a
larger or smaller extent. If actual agreement does not result, per-
haps possible compromises may be classed as more or less
acceptable to one or another of the parties.
Such intelligent political discussion between people of differ-
ent outlooks is, unfortunately, rare nowadays; but it is all the
more enjoyable when it does happen. And one's attitude toward
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
one's co-disputant in such a discussion is interestingly mixed. On

the one hand, one recognizes and appreciates certain intellectual
virtues of the highest importance: open-mindedness, willingness
to consider reasons and arguments, the capacity to accept good
criticisms, etc. But what of the fundamentals on which one can-
not agree? It would be quite dishonest to pretend that one thinks
there are no better and worse reasons and views here. I don't
think it is just a matter of taste whether one thinks that the obli-
gation of the community to treat its members with compassion
takes precedence over property rights; nor does my co-disputant.
Each of us regards the other as lacking, at this level, a certain
kind of sensitivity and perception. To be perfectly honest, there
is in each of us something akin to contempt, not for the other's
mind — for we each have the highest regard for each other's
minds - nor for the other as a person -, for I have more respect
for my colleague's honesty, integrity, kindness, etc., than I do for
that of many people who agree with my 'liberal' political views -
but for a certain complex of emotions and judgments in the
other.
But am I not being less than honest here? I say I respect Bob
Nozick's mind, and I certainly do. I say I respect his character,
and I certainly do. But, if I feel contempt (or something in that
ballpark) for a certain complex of emotions and judgments in
him, is that not contempt (or something like it) for him?
This is a painful thing to explore, and politeness normally
keeps us from examining with any justice what exactly our atti-
tudes are towards those whom we love and disagree with. The
fact is that none of us who is at all grown up likes and respects
everything about anyone (least of all one's own self). There is no
contradiction between having a fundamental liking and respect
for someone and still regarding something in him as an intellec-
tual and moral weakness, just as there is no contradiction
between having a fundamental liking and respect for oneself and
regarding something in oneself as an intellectual and moral (or
emotional, etc.) weakness.
I want to urge that there is all the difference in the world
between an opponent who has the fundamental intellectual vir-
tues of open-mindedness, respect for reason, and self-criticism,
and one who does not; between an opponent who has an impres-
sive and pertinent store of factual knowledge, and one who does
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
not; between an opponent who merely gives vent to his feelings

and fantasies (which is all people commonly do in what passes
for political discussion), and one who reasons carefully. And the
ambivalent attitude of respectful contempt is an honest one:
respect for the intellectual virtues in the other; contempt for the
intellectual or emotional weaknesses (according to one's own
lights, of course, for one always starts with them). 'Respectful
contempt' may sound almost nasty (especially if one confuses it
with contemptuous respect, which is something quite different).
And it would be nasty if the 'contempt' were for the other as a
person, and not just for one complex of feelings and judgments
in him. But it is a far more honest attitude than false relativism;
that is, the pretense that there is no giving reasons, or such a
thing as better or worse reasons on a subject, when one really
does feel that one view is reasonable and the other is irrational.
It may be helpful to descend from the abstract level at which

we have been discussing and consider once more a relatively sim-
ple example. Consider the judgment that most ordinary people
are prepared to make at most times, that peace is preferable to
war. (Such judgments are never discussed by Foucault, just as
they are never described by Swift, and for the same reason: both
are satirists. Only social folly interests them, not — when it
exists — social sanity.) There are no doubts about the sources of
such a judgment. We are too familiar with the horrors of war,
with what war does to adults and children, to combatant and
non-combatant, to the very land and foliage. Even if this judg-
ment springs partly from self-interest, that does not make it irra-
tional, quite the contrary.
Yet whole populations can make the opposite judgment, that
war is preferable to peace, and not for reasons of legitimate self-
defense. Aggression and fantasy can whip people up to a
national blood thirst. But, again, what this shows is not that all
value judgments are irrational, but only that some are; and that
it is very hard to tell which are which when one is not able to
put aside partisanship or criticize one's own beliefs (which is
why we assign so much importance to impartiality and the crit-
ical attitude among the cognitive virtues).
That some value judgments are rational and objective does not
mean that our abstract talk about Capitalism, Democracy,
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
Socialism, Rights, Autonomy, and so on, is not frequently non-

sense. Even when what we mean to say about some general issue
is right, frequently we have trouble expressing it well, especially
if we are not trained in the expression of abstract ideas. The case
of the anthropologist who says there are no objective values
when what he means is that values are relative to circumstances
is a case in point. Even when we do succeed in expressing what
we mean to say effectively, there are powerful forces of a non-
rational kind tending to sway our judgment. My purpose here is
not to deny that power can corrupt our judgment and narcissism
seduce it; it is to deny that we are helpless in the face of these
powerful forces, so helpless that it would be idle (and in fact,
self-deception) to attempt to judge with intelligence and justice.
To say we can be rational is different from saying we can be
infallible. On the contrary, as Iris Murdoch points out, the striv-
ing for a reasoned and rational stance is essentially something
progressive, something 'infinitely perfectible'.4
What I have said so far might, perhaps, be conceded by an

intelligent moral relativist. A relativist need not be concerned to
undermine the rationality of all 'value' judgments, or to defend
Foucault's picture of history as a discontinuous series of 'dis-
courses' or 'ideologies' which succeed one another for no
rational reason. A more modest relativist might be happy to
agree with Dewey5 that some values are objectively relative - i.e.
rational given the circumstances, the nature and history, of those
who make them. What is important, such a modest relativist
maintains, is precisely the relativity of all values. The 'objectiv-
ity' he is denying is not the objectivity Dewey was affirming,
which is simply the objectivity of any judgment that is warranted
in its actual existential setting, but it is rather the objectivity a
Plato would affirm, the spurious objectivity (the relativist would
say) that purports to speak from an absolute point of view, apart
from all circumstances and valid for all circumstances.
If we are not content to accept such a modest relativism, if we
feel troubled by Dewey's own ethical writings, it is not, I think,
because we really do hanker for Absolutes. When 1 claim that
4
The Sovereignty of Good, p. 23.
5
See Dewey's Theory of Valuation, in The Encyclopedia of Unified
Science, vol. II, no. 4, University of Chicago (1939).
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
the murder and suffering of innocent people is wrong, I do not,

I think, really care about the question whether this judgment
would be valid for a being of a totally alien constitution and
psychology. If there are beings on, say, Alpha Centauri, who
cannot feel pain and who do not mind individual death, then
very likely our fuss about 'murder and suffering' will seem to
them to be much ado about nothing. But the very alienness of
such a life form means that they cannot understand the moral
issues involved. If our 'objectivity' is objectivity humanly speak-
ing, it is still objectivity enough.
What is of concern is that Dewey's doctrine of 'objective rel-

ativism' cannot handle the case of the Nazi (although Dewey
would have disputed this). We want to say that the Nazi's goals
were deeply wrong; and the claim that 'this is true relative to
your interests and false relative to the Nazi's interests' is just the
kind of 'moral relativism' we find repulsive. Objective relativism
seems the right doctrine for many moral cases; but not for cases
where rights and duties are manifest and sharp and the choice
seems to us to be between right and wrong, good and evil.
Indeed, there is a sense in which the modern instrumental
notion of rationality is itself 'objective relativist'. The core of this
notion is a deceptively simple dichotomy: the idea is that the
choice of 'ends' or 'goals' is neither rational nor irrational (pro-
vided some minimal consistency requirements are met); while
the choice of means is rational to the extent that it is efficient.
Rationality is a predicate of means, not ends, and it is totally
conflated with efficiency. Thus, Jones' preference for vanilla over
chocolate ice cream is neither rational or irrational, but the
action of choosing vanilla over chocolate on a particular occa-
sion would be rational for Jones, given his 'preference ordering'.
This conception, which goes back to Hume's dictum that 'reason
is and ought to be the slave of the passions', and which deeply
influenced Bentham, is widely assumed to be the right one by
modern social scientists. It has played a role in welfare econom-
ics and in many other areas. The modern economist's notion of
a Pareto optimum is an attempt to have a notion of economic
optimality which considers only efficiency of means, and
involves no 'value judgments' concerning the goals of the various
economic agents; this notion is of contemporary interest pre-
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
cisely because the assumption is made that the choice of means

is subject to rational criticism, while the choice of ends is not.
This whole conception loses much of its persuasive appeal,
however, when we see what an oversimplified psychological the-
ory it rests upon. In the Benthamite scheme, goals, ends, prefer-
ences are treated either as fixed individual parameters (i.e. the
individual's learning is pictured as a process of learning to better
estimate consequences and probable consequences of actions
and to attain ends more efficiently, but not as a process of
acquiring new ends) or as individual parameters which, if not
actually fixed, change only as a result of factors which have no
rational status and which the theorist cannot take account of.
Bernard Williams6 has pointed out that there are a number of
ways in which an individual's goals, and not just the means he
chooses to attain them, can be rationally criticized; ways which
become apparent the moment we pass beyond this narrow 'Ben-
thamite' psychology.
The 'Benthamite' conception does allow one case in which an
individual can be persuaded to abandon a goal (or, at any rate,
to abandon pursuit of the goal) by rational criticism: this is the
case in which he had misestimated consequences in the direction
of badly underestimating the costs of attaining the goal (relative
to other goals he has). This opens the door to a question which
has to do as much with imagination as with propositional intel-
ligence: the question of what it would actually be like, experien-
tially, to attain the goal. Many human beings pursue goals they
would not actually enjoy attaining or would not enjoy nearly as
long or nearly as much as they think. Even within a Benthamite
framework, it would be possible to improve the account of
rational decision making by taking into account the possibility
of misestimating the actual existential feel of various goals. And
this begins to introduce a sense in which goals themselves can be
criticized as irrational, and not just means.
Again, people often overlook goals they might pursue if they
thought of them. Or, even if they think of them (or someone
suggests them) they may lack the imagination (imagination
again!) to visualize what the attainment of these goals would
I am here summarizing a lecture titled 'Internal and External Reasons',

given at Harvard, Nov. 1978.
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
really be like — all the more if these goals are long-term traits of
character, such as developing an appreciation of poetry. The
man who prefers pushpin to poetry may not actually be able to
imagine what it would be like to have a developed sensitivity to
the nuances of real poetry, and if his intelligence could be raised
or his imagination improved he might be brought to see that he
is making a mistake.
It is significant that the ability to rationally criticize one's own
goals (and those of others) may depend just as much on one's
imagination as on one's ability to accept true statements and
disbelieve false ones. And it is significant that one's goal may be
a long-term trait of mind or character, and not a thing or event.
There are still further ways besides misestimating the real
experiential significance of one's goals or of possible alternative
goals in which one may make errors in the choice of goals. Wil-
liams points out (reviving an observation that goes back to Aris-
totle) that very often a goal is general (e.g. 'having a good time
this evening') and the problem is not so much to find a means to
the 'end', but to find an overall pattern of activity that will con-
stitute an acceptable specification of the goal (e.g. 'going to a
movie' or 'staying home and reading a book'). Whether one can
think of creative and novel specifications of one's goal or only of
commonplace and banal specifications will depend again on
imagination and not just propositional intelligence.
The problem, as Williams pointed out, is that even if one
replaces the narrow Benthamite psychology with an account that
does justice to all of these things, one still seems to be left with
a certain relativism. Williams' example was a hypothetical case
of a young man whose father wished him to undertake a military
career. The old man appeals to family traditions (the males have
been army officers for generations) and patriotism, but in vain.
Even when the young man makes as vivid to himself as he can
what it would be like to be an army officer, there is nothing in
this goal which appeals to him. It just is not his end; and not
because of some failure of intelligence or imagination.
Even the case of the Nazi could be like this. Suppose the Nazis
had won the war, so that we could not appeal to Germany's
defeat as a practical reason for not being a Nazi. Perhaps some
Nazis were simply lacking in knowledge of the actual conse-
quences of Nazism, the suffering brought about, and so on. Per-
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
haps some Nazis would not have been Nazis if they had had the
intelligence and imagination to appreciate these consequences,
or to appreciate more vividly the alternative life, the life of a
good man. But doubtless many Nazis would still have been
Nazis, because they did not care about the suffering their actions
caused and because no matter how vivid they might make the
alternative life seem to their imaginations, it would no more
speak to anything in them than the military life did to the young
man in Bernard Williams' story. There is no end in them to
which we can appeal, neither an actual end or even a potential
one, one which they would come to realize if they were more
intelligent and more imaginative. Even without 'Benthamite psy-
chology', we are faced again with the problem of moral relativ-
ism.
Let us consider a case less inflammatory than the Nazi case.
Imagine a society of farmers who, for some reason, have a total
disinterest in the arts, in science (except in such products as assist
them in farming), in religion, in short, in everything spiritual or
cultural. (I don't mean to suggest that actual peasant societies
are or ever have been like this.) These people need not be imag-
ined as being bad people; imagine them as cooperative, pacific,
reasonably kind to one another, if you like. What I wish the
reader to imagine is that their interests are limited to such mini-
mal goals as getting enough to eat, warm shelter, and such sim-
ple pleasures as getting drunk together in the evenings. In short,
imagine them as living a relatively 'animal' existence, and as not
wishing to live any other kind of existence.
Such people are not immoral. There is nothing impermissible
about their way of life. But our natural tendency (unless we are
entranced with Ethical Relativism) is to say that their way of life
is in some way contemptible. It is totally lacking in what Aris-
totle called 'nobility'. They are living the lives of swine — amiable
swine, perhaps, but still swine, and a pig's life is no life for a
man.
At the same time — and this is the rub — we are disinclined to
say the pig-men are in any way irrational. This may be the result
of our long acculturation in the Benthamite use of'rational' and
'irrational', but, be that as it may, it is our present disposition.
The lives of the pig-men are not as good as they might be, we
want to say, but they are not irrational.
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
We do not want to say that it is just a matter of taste whether

one lives a better or worse life. We don't see how we can say that
it is rational to choose the better life and irrational to choose the
worse. Yet not saying some such things seems precisely to be
saying that 'it's all relative'; the ground crumbles beneath our
feet.
Perhaps some of the corrections to Benthamite psychology
suggested by Bernard Williams will help with this case. Let us
assume the pig-men are born with normal human potential (if
they aren't, then their lives aren't 'worse than they might be',
and we are not justified in feeling contempt, but only, at most,
pity). Then they might be led to appreciate artistic, scientific, and
spiritual aspects of life; to live more truly human lives, so to
speak. And if any of them did this, they would doubtless prefer
those lives (even though they might be less fun) to the lives they
are now living. People who live swinish lives feel shame when
they come to live more human lives; people who live more
human lives do not feel ashamed that they did when they sink
into swinishness. These facts give one grounds for thinking that
the pig-men are making the sort of error, the sort of cognitive
short-fall, that Williams discussed; grounds for thinking that
they have overlooked alternative goals, and certainly grounds
for thinking that they have never made vivid to themselves what
realization of those alternative goals would be like. In short, one
cannot really say that they have chosen the worse life; for they
never had an adequate conception of the better.
While this might give us a handle on the notion that those
peoples' lives are open to rational criticism, it is not evident how
to apply it to the case of the Nazi. (One could make it a tautol-
ogy that anyone who doesn't actually choose the better life
hasn't 'adequately conceived' it; but such a maneuver would
clearly be no help.) Even in the case of the pig-men, if they were
ideologically fanatic pig-men, as opposed to mere pig-men, then
our point about the direction of shame might not hold. There
might not, in such a case, be any end that is theirs, even latently,
to which we could appeal.
Our reluctance to accuse the pig-men of a defect in reason
(unless we can point to some end that is theirs, at least latently,
which they are failing to achieve) is the product of the recent
vicissitudes of the notion of reason in our culture, as is easy to
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
establish. For neither ancient philosophers nor the medievals

saw anything strange about saying that if A is a better life than
B then that fact is a reason, the best possible reason, ior choosing
A over B. We have lost the ability to see how the goodness of an
end can make it rational to choose that end.
Of course, this is very largely explained by the fact that we
don't regard 'goodness' as anything objective. But now we are
confronted with a circle, or rather two curves. There is the mod-
ern circle: the instrumentalist conception of rationality supports
the claim that the goodness of an end doesn't particularly make
it irrational not to choose that end, or to choose an end which is
downright bad, which in turn supports the claim that goodness
and badness are not objective, which in turn supports the claim
that the instrumentalist notion of rationality is the only intelli-
gible one. And there is the traditional arch: reason is a faculty
which chooses ends on the basis of their goodness (as opposed
to the 'passions', which try to dictate ends on the basis of the
appetites; or 'inclination'); a claim which supports the view that
it is rational to choose the good, which in turn supports the
claim that goodness and badness are objective. Clearly we can-
not simply go back to the ancient or medieval world-view, what-
ever conservatives might wish; but is the Benthamite circle really
the only alternative left to us?
Cambridge
https://doi.org/10.1017/CBO9780511625398.009
8
The impact of science on
modern conceptions of rationality
If the discussion that we have reviewed - and it is a discussion

that has been going on for many decades - seems inconclusive,
it is perhaps because the discussion always assumes a kind of
priority of rationality over goodness. The question is always
whether there is any sense in which it can be called irrational to
choose a bad end, as if goodness were on trial and rationality
were the judge. To assume this stance, especially when one's
assumptions about rationality are a largely unexamined collec-
tion of cultural myths and prejudices, is to prejudge the question
of the status of value judgments in advance. In the remainder of
this essay I propose to reverse the terms of the comparison and
to ask not how rational is goodness, but why is it good to be
rational? Asking what value rationality itself has will both force
us to become clearer about the nature of rationality itself and
about the assumptions we are prone to make concerning ration-
ality and may enable us to see what is wrong with the way we
think about the former question.
Let us recall that when Max Weber introduced the modern
fact—value distinction, his argument against the objectivity of
value judgments was precisely that it is not possible to establish
the truth of a value judgment to the satisfaction of all possible
rational persons.1 From the very beginning it was the impossi-
1
Cf. especially 'Die Objektivitat sozialwissenschaftlicher Erkenntnis', in
Archiv fur Sozialwiss. und Sozialpolitik, vol. 19 (1904), pp. 24-87,
and 'Der Sinn der Wertfreiheit der soziologischen und okonomischen
Wissenschaften', in Logos, VII (1917), pp. 49-88 and 'Wissenschaft
als BeruP, Vortrag, 1919. All three texts were reprinted in The
Methodology of the Social Sciences, Glencoe, Illinois, 1949.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
The impact of science 175
bility, or alleged impossibility, of rational proof that cast value

judgments into a somewhat suspect light. Rationality has been
putting value on trial for a long time. And in this context, ration-
ality always means scientific rationality;2 it is the results of 'pos-
itive science' that are said to be such that they can be established
to the satisfaction of all rational persons. One reason for valuing
rationality is obvious. Scientific rationality undeniably aids us in
achieving various practical goals. While few educated people
would subscribe to the view that we ought to pursue science
solely for the sake of technological success, there is no doubt
that the technological success of science is overwhelming in the
most literal sense of the word. We live in an apparently endless
2
K. O. Apel reads Weber as I do in 'The Common Presuppositions of
Hermeneutics and Ethics: Types of Rationality Beyond Science and
Technology', Research in Phenomenology, No. IX, 1979. Thus Apel
writes (p. 36):
Max Weber, however, also proposed a strictly negative answer
with regard to my question as to possible types of rationality
beyond value-free science and technology. And this answer has
become paradigmatic, I suggest, for the present system of Western
ideology. Weber restricted the scope of methodical understanding
to 'value-free' understanding which he centered around the 'ideal
type' of 'purposive-rational understanding' of 'purposive-
rational actions'. Now 'purposive-rational actions' may also be
called 'instrumental actions'; and in those cases where these
actions are successful, they may be analyzed or reconstructed as
being based on successful transpositions of the if—then-rules of
nomological science into the if—then-rules of technological pre-
scriptions. Hence Max Weber thus restricted the business of
methodical understanding to the attempt of grasping the (value-
free) technological means-ends-rationality beyond the human
actions. And it is this idea of instrumental rationality which
indeed constituted Weber's paradigm of rationality in a restrictive
sense.
It has to be pointed out, though, that for a purposive—rational
understanding in sociology it is not necessary to fulfill the maxi-
mal requirement of making sure that the agent succeeded in
transposing nomological rules into his technological maxims
about means-ends-relations. In order to understand his actions in
the light of that type of instrumental rationality, it is enough to
make sure that it was rational for the agent to act as he did under
the presupposition of his aims and his beliefs about means or
ways or strategies as being suited to reaching his aims. Thus it
becomes the empiric—hermeneutic business of understanding to
hypothetically find out and verify those goal-intentions and
means-beliefs on the side of the agent, in the light of which his
actions can be understood as being rational in the sense of tech-
nological means—ends rationality.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
176 The impact of science
series of technological revolutions - 'the Industrial Revolution',

'the Electronic Revolution' - which constantly remind us how
momentous a force science is in shaping our lives. Even before
the Industrial Revolution the apparently unique success of New-
tonian physics impressed a number of minds. For example, when
the notion of 'progress' first began to be discussed in the seven-
teenth century, the progressivists clinched their case with the
claim that 'Newton knew more than Aristotle.' No one could
argue convincingly that Shakespeare is a better dramatist than
any of the ancient tragedians, or a better poet than Homer; but
it seemed undeniable that the scientist Newton had made a real
and undeniable advance in knowledge upon the scientist Aris-
totle.
Although the encyclopedists and others were quick to gener-
alize the notion of progress from science to political institutions
and morality, that generalization has appeared as dubious to the
twentieth century as it appeared evident to the nineteenth.
Auguste Comte built a philosophy, positivism, celebrating the
success of science. History, as Comte tells it, is a success story:
we start with primitive myths, these become refined and purified
until eventually the high religions appear, the high religions in
turn give way to the metaphysical theories of a Plato or a Kant,
and finally in our own day metaphysics itself has to give way to
'positive science'. Evidently, there is no doubt who the hero of
this success story is: the hero is Science. And if what impressed
the Few about science from the start was its stunning intellectual
success, there is no doubt that what has impressed the Many is
its overwhelming material and technological success. We are
impressed by this even when it threatens our very lives.
One reason, then, for doubting that value judgments have any
cognitive status is that they cannot be 'verified by the methods of
science', as it gets put over and over again. Then too there is the
fact, that we have already seen play a central role in Foucault's
discussion, that one cannot get universal, or even majority,
agreement on ethical matters. Whether abortion is right or
wrong or whether homosexuality is right or wrong do not seem
to be questions on which some answer can be demonstrated to
everyone's satisfaction; whereas it is widely believed that the
correctness of a scientific theory can be demonstrated to every-
one's satisfaction. The very rationality of science itself is some-
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
times thought to consist, or to consist in part, in the fact that at

least the predictions of science can be publicly demonstrated;
that everyone can be brought to agree that these results do
obtain, that the phenomena predicted by the theory do occur. Of
course there is a threat of circularity here: if we identify rational
procedures as those which lead to conclusions on which we can
get majority assent, then Weber's argument, even if correct, that
in ethical matters one cannot get the consent of all rational peo-
ple, would mean that one cannot get consent of all those who
use methods which are guaranteed to produce the assent of the
majority or the overwhelming majority. That is, the way in
which we determine that value judgments cannot be verified to
the satisfaction of all rational people is simply by observing that
they cannot be verified to the satisfaction of the overwhelming
majority of all people. It is not, after all, as if we had a test for
rationality; Weber's formulation suggests that somehow we first
take a headcount of those members of the population who are
rational and then see whether or not they can all be brought to
agree on whether or not some value judgment is true. But noth-
ing like this really goes on. All that Weber's examples (Chinese
Mandarins and so on) really show is that value judgments can-
not be verified to the satisfaction of all educated people or all
intelligent people (which is by no means the same thing as all
rational people). In a disguised form Weber's argument is a
Majoritarian argument; he is appealing to the fact that we can
get the agreement of educated people on 'positive science'
whereas we cannot get such an agreement on ethical values. It is
interesting to contrast this stance with Aristotle's: Aristotle said
that of course in ethics we should always try to get the agreement
of the Many, but very often we know that realistically we can-
not. Sometimes, elitist as it sounds to present day ears to say it,
we are only able to convince the wise; and of course we have to
rely on our own judgment to tell who are and who are not the
wise.
Of course it is not really true that one can get overwhelming
agreement on the truth of an arbitrary accepted scientific theory.
The fact is that most people are woefully ignorant of science and
many theories, especially in the exact sciences, require so much
mathematics for their comprehension that most people are not
even capable of understanding them. While this is of course con-
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
ceded, to most people it does not seem to affect the point. For,
according to the watered-down operationism which seems to
have become the working philosophy of most scientists, the con-
tent of the scientific theory consists in testable consequences, and
these can be expressed by statements of the form // we perform
such and such actions, then we will get such and such observable
results. Statements of this form, if true, can be demonstrated to
be true by repeating the appropriate experiment often enough. It
is true that there are many difficulties with this account: experi-
ments are much harder to design, perform, and evaluate than the
layman may think. But there is no doubt that as a matter of fact
it has been possible to achieve widespread agreement on the
experimental adequacy of certain theories in the exact sciences.
The layman's acceptance of these theories may be a matter of his
deference to experts, but at least the experts seem to be in agree-
ment.
Intellectually, of course, Instrumentalism does not simply in
and of itself constitute a tenable conception of rationality. No
doubt scientific results have enormous practical value; but, as we
have already said, no educated person thinks that science is val-
uable solely for the sake of its practical applications. And even if
science were valued solely for the sake of its applications, why
should rationality be valuable solely for the sake of applications?
To be sure it is of value to have an instrument that helps us select
efficient means for the attainment of our various ends; but it is
also valuable to know what ends we should choose. It is not
surprising that the truth of value judgments cannot be 'rationally
demonstrated' if 'rational verification' is by definition limited to
the establishment of means-ends connections. But why should
we have such a narrow conception of rationality in the first
place?3
3
Attributing just such a narrow conception of rationality to Weber, Apel
writes {loc. cit., p. 37):
This issue of Weber's methodology of 'understanding' was in

perfect accordance with his (more or less implicit) philosophy of
history. For in the context of his own reconstruction of the his-
tory of Western civilization, he started out from the heuristic
hypothesis that at least this part of history could be conceived of
as a continuous progress of 'rationalization' and at the same time,
as a process of disillusionment or, as he liked to say, 'disenchant-
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
Majoritarianism is also intenable. To be sure it is nice to

get agreement on what one takes to be true. And it is always nice
to avoid conflict with one's fellows. But people have lived for
centuries with the uncomfortable knowledge that on some mat-
ters one has to rely on one's judgment even when it differs from
the judgment of the majority. Many have gloried in relying on
their judgment when it differed from the judgment of the major-
ity. The idea that on some matters, ethical matters among them,
the considerations to be weighed are just so complex, and so
imprecise, that we cannot hope to rely on anything like scientific
proof or scientific definitions but have to rely on perception and
judgment is an old one. And it is plausible that one of the highest
manifestations of rationality should be the ability to judge cor-
rectly in precisely those cases where one cannot hope to 'prove'
things to the satisfaction of the majority. It seems strange indeed
that the fact that some things should be impossible to prove to
everyone's satisfaction should become an argument for the irra-
tionality of beliefs about those things.4
Even if these conceptions are intellectually weak however, it
seems to be the case that both Instrumentalism and Majoritari-
anism are powerfully appealing to the contemporary mind. The
contemporary mind likes demonstrable success; and the contem-
porary mind is uncomfortable with the very notions of judg-
ment and wisdom. I am not a sociologist, and I will not attempt
to explore the question why industrial society, in both its capi-
ment' ['Entzauberung']. By 'rationalization' he understood the

progress in putting into force means—end-rationality in all sectors
of the socio-cultural system, especially in the sphere of economics
and bureaucratic administration, under the constant influence of
the progress in science and technology. By the process of 'disillu-
sionment' or 'disenchantment,' on the other hand, Weber under-
stood, among other things, the dissolution of a commonly
accepted religious or philosophical value-order or world-view.
And he was prepared to draw practical consequences from this
development for his personal world-view in so far as he suggested
that a rigorous and sincere thinker had to accept the following
insight: Human progress in the sense of 'rationalization' has its
complement in giving up the idea of a rational assessment of last
values or norms in favor of taking recourse to ultimate pre-
rational decisions of conscience in face of a pluralism, or, as
Weber said, 'polytheism' of last norms or values.
4
Indeed, we saw in Chapter 5 that the consensus theory of rational
acceptability is self-refuting!
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
talist and its socialist versions, should be so taken with the

themes of instrumental success and majority consent. But doubt-
less the sociological fact has something to do with the ever
increasing prominence of the conception that rationality equals
scientific rationality, and the conception of scientific rationality
as itself based on the demonstration of instrumental connections
to the (potential) satisfaction of the overwhelming majority.
If the conception of rationality we have just described, the
conception of rationality as consisting in methods (whose nature
is usually left rather vague) which, whatever their nature, result
in the discovery of effective means/ends connections and the
establishment of these connections 'publicly', is not as it stands
intellectually tenable, philosophical attempts to make it respect-
able have not been lacking. One of these attempts grows out of
the older empiricism of Locke, Berkeley and Hume. By the time
of Mill this empiricism had solidified into what philosophers call
phenomenalism: that is the doctrine that all we can really talk
about are sensations. Even everyday objects, e.g. tables and
chairs, are really just sets of objective regularities in actual and
possible human sensations, on this conception. As Mill put it,
physical objects are 'permanent possibilities of sensation'.
Another way to put the same idea is to say that all talk that
appears to be about the physical world, or whatever, is really
just highly derived talk about sensations.
The virtue of this point of view, in the eyes of its holders, was
that it enabled them to say clearly what the content was, not
only of science, but of all cognitively meaningful talk whatso-
ever. Any scientific theory is really just an 'economical' way of
stating a number of facts of the form: if you perform such and
such actions, then you will have such and such experiences. The
holder of this view does not have to defend the untenable claim
that scientists are interested only in applications, or only in the
attainment of practical goals, and disinterested in knowledge for
its own sake. The phenomenalist does not have to deny that we
want to know the nature of black holes, that we want to know
whether or not there was a Big Bang, what the true origin of
homo-sapiens was, etc. We do want to know all these things and
not just because knowing them might enable us to build better
machines. But knowing these things is, although it takes refined
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
philosophical analysis to show it, just knowing a great many

facts of the form if you perform such and such actions, then you
will have such and such experiences. Whatever our reason for
being interested in them, all facts are ultimately instrumental. At
the same time, there seems to be no way of making out the claim
that to call something good is to make any prediction of the
form: if you perform such and such actions then you will have
such and such experiences. Thus statements about the goodness
or badness of anything have no cognitive meaning on this con-
ception; in the words of the twentieth-century Logical Empiri-
cists such statements are purely 'emotive'. Phenomenalism came
to grief on two points, however. In the first place the claim that
statements about material objects are translatable into state-
ments about actual and possible sensations seems as a matter of
fact to be false. Careful logical investigation of this claim, start-
ing with the work of Carnap and the Vienna Circle in the 1930s,
convinced the phenomenalists themselves that the claim was
unfounded. Scientific theories as a whole undoubtedly lead us to
expect that we will have certain experiences if we perform cer-
tain actions; but the idea that the statements of science are trans-
latable one by one into statements about what experiences we
will have if we perform certain actions has now been given up as
an unacceptable kind of reductionism. In the second place sen-
sations are necessarily private objects; although we may be able
in practice to decide whether or not someone had a sensation by
simply asking them, we are immediately in some kind of epis-
temological trouble if someone raises the question 'how do you
know that the person associates the same sensations with his
descriptions that you do?' If the content of science consists in
predictions about what sensations any rational being will have if
that rational being performs certain actions, then to know what
that content itself means we would have to be able to tell, for
example, whether extraterrestrials if we encounter any, have the
same sensations that we do or not, etc. For this reason, philoso-
phers like Rudolph Carnap and Sir Karl Popper insisted that the
observational predictions of science should be stated in the form
// anyone performs such and such actions, then such and such
publicly observable events will take place — where both the
actions to be performed and the observable events that should
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
be expected must be described in terms of 'public' objects, e.g.

meter readings, and not in terms of such private objects as sen-
sations.
To sum up, the older empiricism, or phenomenalism, seemed
to provide us with a tidy criterion of cognitive significance: a
statement is cognitively significant if it is translatable into a
statement about sensations. But it turned out that either the
notion of'translation' is hopelessly vague, or else that statements
of science itself failed to satisfy this criterion of cognitive signif-
icance. The trouble with drawing a sharp line between factual
statements and value judgments on the grounds that the former
but not the latter are 'translatable' into statements about sensa-
tions is that the alleged translatability of the first class of state-
ments has not been, and apparently cannot be, demonstrated.
Empiricist reductionism drew a sharp line between the factual
and the evaluational, but at the price of giving a wholly distorted
picture of the factual.
Our original purpose, however, was to consider answers to
the question 'why is it good to be rational?' The first answer we
considered, and rejected as too narrow, was that rationality
enables us to discover reliable means/ends connections. Phenom-
enalism came into the story because if phenomenalism were true
then the apparent conflict between being interested in a scientific
theory for the sake of its instrumental consequences, and being
interested in the theory in order to learn what it tells us about
natural processes would dissolve. The conflict between instru-
mental interests and purely theoretical interests could in a sense
be finessed. Of course there would still be some kind of differ-
ence between these interests; but even purely theoretical interests
would be interests in facts which, in the ultimate logical analysis,
would have been revealed to be of an instrumental nature. All
knowledge worthy of the name would have been shown to be
knowledge of means/ends connections; it is just that when we
are interested in the means/ends connection because we hope to
exploit it for the attainment of some goal, then we call our inter-
ests 'practical', and when we are interested in knowing the
means/ends connection out of pure curiosity then we call it a
theoretical interest. This attempt to reduce all the statements of
science to statements of the form // you perform action A then
you will get result B has, as we remarked, failed.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
The contention of Carnap and Popper, that the observation

statements of science are couched in physical thing language and
not in sensation language, is obviously correct if taken as a gen-
eralization about the practice of scientists. When it is erected
into an epistemological absolute, however, its import becomes
rather momentous. For one thing, if no observation statement at
all is allowed to talk about sensations, then introspection is ruled
out as a mode of scientific observation. Although many psychol-
ogists would agree that it should be ruled out, the fact is that
some psychologists, undeterred by both philosophical and psy-
chological dogma, have gone on performing experiments which
involve at least in part reliance on introspective reports. In fact,
Carnap would not have been as dogmatic in prohibiting this as
some behaviorist psychologists were; he would have permitted
the use of sensation reports, provided that they were construed
not as observation reports but as behavioral data, the 'behavior'
being the making of the verbal reports themselves. But what it
means to construe the acceptance of an introspective report as
an 'inference from verbal behavior' is not altogether clear. Even
in the case of reports which are not about sensations but about
physical objects, e.g. 'there is a table in front of me', we do not
normally accept the report unless we have some theory accord-
ing to which the person was in a position to observe the fact that
he reports. In this sense it is a part of our whole demand for
coherence in our world picture that observations should them-
selves be theoretically explainable; if someone claims to have
observed that there was a table in a certain place by clairvoyance
we do not accept this 'observation report' because it does not
cohere with our total body of theory. In this sense every obser-
vation report has some component which could be described as
'inferential'. On the other hand, when, say, a doctor accepts the
report of a patient that the patient feels pain, it is hard to know
on the basis of what 'scientific theory' the doctor infers that the
patient feels pain from the patient's verbal report; if the general
assumption that people are in a good position to tell whether or
not they have a pain counts as a theory, then the general assump-
tion that people are in a good position to tell whether or not
there is a table in front of them also counts as a theory; but it is
hard to see that there is any fundamental methodological differ-
ence between accepting someone's say so that there was a table
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
in front of them and accepting someone's say so that they had a

pain.
Popper and Carnap would reply that the methodological dif-
ference is that the former statement but not the latter is publicly
checkable; but they both exaggerate the extent to which obser-
vation reports about physical objects are always publicly check-
able. Many such reports are made with the aid of instruments
which it takes a good deal of training to use. (It is notorious that
learning to 'see' through a very high-powered microscope
requires a good deal of specialized training and skill and that not
everyone is capable of acquiring the skill.) What accepting this
epistemological dogma does is make it a part of the definition of
rationality that rational beliefs are capable of being publicly
checked. Making it part of the definition of rationality is very
convenient; it makes it unnecessary to provide any argument for
this contention. Perhaps the argument is, at bottom, that what-
ever is not publicly checkable may become a matter of disagree-
ment, and that wherever there is unsettleable disagreement there
is no being right or wrong. But this would assume what I have
called Majoritarianism, that is the idea that it is built into the
very notion of rationality that what is rationally verifiable is ver-
ifiable to the satisfaction of the overwhelming majority.
The fact that Logical Empiricism was fundamentally a sophis-
ticated expression of the broad cultural tendencies to instrumen-
talism and majoritarianism becomes evident, I think, in the later
history of this movement. Although the Logical Empiricists had
abandoned phenomenalism as early as 1936, for the next twenty
years, that is until the movement began to break up and disap-
pear as a recognizable philosophical tendency, Logical Empiri-
cist philosophers of science were fond of talking about 'the aim
of science' and fond of identifying the aim of science with pre-
diction (with some additions and qualifications which I shall dis-
cuss in a moment). The idea that the aim of science is prediction
was the fundamental idea of positivism from its beginnings in
the writings of Auguste Comte. As we saw, this idea had at least
some kind of serious philosophical rationale as long as phenom-
enalism was in vogue; for then one could argue that all cogni-
tively meaningful statements were predictions in disguise, or
infinite sets of predictions in disguise. The reappearance of this
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
doctrine, after the disappearance of phenomenalism, is like the

appearance of 'primitive' material in a patient's associations in
therapy, after the 'defenses' have been stripped away. To say
that the aim of science is successful prediction (or successful pre-
diction plus something simply described as 'simplicity'), seems
dangerously close to saying that science is pursued only for prac-
tical goals; and this is something that no philosopher has wanted
to be put in the position of maintaining. Indeed, the philosophers
who defended a purely instrumentalist conception of science did
so not because they were themselves worshippers of the practi-
cal, or narrow minded men who could not appreciate the beauty
of abstract scientific knowledge for its own sake, but they did so
rather because they felt that by identifying what is 'cognitively
significant' with what has value for the making of predictions
they could once and for all rule out all forms of obscurantism
and metaphysics. 'Metaphysics' was for these philosophers sim-
ply another name for various kinds of transcendental specula-
tion; it was religious and 'metaphysical' speculations (in their
sense of 'metaphysical') that they were afraid of.
I am suggesting that the appearance in the culture of a philo-
sophical tendency which was hypnotized by the success of sci-
ence to such an extent that it could not conceive of the possibility
of knowledge and reason outside of what we are pleased to call
the sciences is something that was to be expected given the enor-
mously high prestige that science has in the general culture, and
given the declining prestige of religion, absolute ethics, and tran-
scendental metaphysics. And I am suggesting that the high pres-
tige of science in the general culture is very much due to the
enormous instrumental success of science, together with the fact
that science seems free from the interminable and unsettleable
debates that we find in religion, ethics and metaphysics.
Since, however, the professional philosophers who, as it were,
rationalized the instrumentalist tendency in the culture were not
themselves vulgar-minded or purely practical persons, it is not
surprising that they themselves felt inclined to widen the descrip-
tion of 'the aim of science' somewhat, so as more explicitly to
leave room for aims other than just successful prediction. And
so we find other aims being listed by Logical Empiricist writers
in the 1940s and 1950s: the discovery of laws, retrodiction (i.e.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
the prediction of past as opposed to future events), and the dis-

covery of 'explanations', by which these writers meant simply
the deduction of predictions and retrodictions from laws.
What happened here is interesting. In order to make it explicit
that science is interested in discovering laws of nature for their
own sake, and not merely for the sake of the predictions to
which those laws lead, these writers replaced the simple formula
'the aim of science is successful prediction' with a list. The list
is in fact open-ended: laws of nature turn out to include not only
laws of nature in the strict sense, i.e. statements which it is phys-
ically impossible to falsify, but also the so-called 'laws' of evo-
lutionary theory, which are really descriptions of general ten-
dencies which may at some time, owing to the action of
intelligent life, cease to hold, and even statements about the
purely contingent dispositions of individual groups and even
individual organisms. To say that scientists are trying to discover
'laws of nature', including physically contingent generalizations
which hold for long periods of time and which have wide
explanatory significance, such as those upon which evolutionary
theory is based, or those upon which the science of economics is
based, and seek to discover significant truths concerning the dis-
positions of groups and individual organisms, and seek to organ-
ize all of these into a deductive (and inductive) structure, is, of
course, quite correct. But why this particular list?
The reason for the list is that it is thought to be embracing
enough to include all of the kinds of truths that scientists seek to
discover, certainly in physical science, and narrow enough not
to include any of the objectionable ('cognitively insignificant')
material. The old search for a 'criterion of cognitive signifi-
cance', such as 'a sentence is meaningful if and only if it is pos-
sible to verify or falsify it', has been replaced by a list of types of
statements, such that a statement is to be admitted if it is of one
of these types and otherwise to be rejected. But why was this at
all a plausible move for a philosopher to make? Even if it is true
that all of the statements in the disciplines that we call 'sciences'
are of these types (and it is not at all clear that this is the case —
is historical explanation really just subsumption of retrodictions
under 'laws'?), does it follow that the verification of these types
of statements and just these types of statements is the aim of
reason itself, and not just the aim of the special applications of
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
reason that we call the sciences? The answer, of course, is that

these philosophers did not seriously doubt that 'science'
exhausts reason. But why did they not doubt this? They did not
doubt this because for them the opposition was not between sci-
ence, in the sense of knowledge proceeding by essentially the
methods of the empirical and mathematical sciences, and infor-
mal reason, proceeding by methods which might be adapted to
interests different than those of the sciences, but no less capable
of having genuine standards. Rather the opposition was between
knowledge proceeding by the methods of the sciences, and
pseudo knowledge pretending to proceed by revelation, or some
kind of funny transcendental faculties. Reason had to be co-
extensive with science because What Else Could It Be? Never-
theless the claim of these philosophers that reason is co-extensive
with science landed them in some peculiar predicaments. Since
they did not wish to deny that there is such a thing as historical
knowledge, for example, they were committed to the position
that history is a science, and even to the position that what the
historian is really trying to do is to subsume individual state-
ments about the past under laws - a claim about history that
seems false on the face of it.
It is, perhaps, not surprising that the Logical Empiricist ten-
dency began to disintegrate by 1950. We have been looking at
this tendency solely from the point of view of one question; the
Logical Empiricists had a great many different philosophical
interests, and they made many valuable contributions. Neverthe-
less, from the point of view of the question we have been asking,
which is 'what good is rationality?', the Logical Empiricist
movement represented a reasoned philosophical defense of the
view that the answer, and the sole answer, to the question is that
rationality is good for the discovery of means/ends connections.
The philosophical doctrine of phenomenalism provided the Log-
ical Empiricists with an interesting philosophical defense of this
claim. When the phenomenalism was given up, and the philo-
sophical defense of the claim was replaced by the bare claim, and
even more when the bare claim was made more 'reasonable' by
allowing exceptions, modifications, etc., the whole cutting
power of the movement disappeared. The trouble with the posi-
tion that the aims of reason itself are the discovery of predic-
tions, retrodictions, laws of nature, and the systematization of
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
all of these, and that these are all the aims of reason, is that there
is simply no reason to believe it, I don't mean to say that there is
reason to believe that it is false; if the notion of a law of nature
is widened so that the discovery of laws of nature includes the
discovery of dispositional statements about individual orga-
nisms, and the notion of a disposition is so wide (or so vague)
that the statement that a certain scientist is envious of his col-
league's reputation counts as a statement of a 'disposition', and
the statement that that scientist told a certain joke because he
was jealous of his colleague's reputation is a 'subsumption of a
particular event under law', then it may be that everything one
says can be interpreted as either stating general laws or as sub-
suming descriptions under general laws. Perhaps even saying of
someone that he is morally good can be construed as ascribing a
'disposition' to that someone. No, the trouble with trying to
specify the aims of cognitive inquiry in general by means of a list
of this kind is that the list itself has to be construed: if the terms
in the list are construed in a more or less literal way, then the
kinds of statements in the list would not even include all of the
sorts of statements that scientists are interested in discovering,
certainly not if 'scientist' includes historian, psychiatrist, and
sociologist; while if the terms in the list are construed so leni-
ently that there is no difficulty in construing the statements made
by historians (and descriptive statements in the language of
everyday psychology) as belonging to types included in the list,
then the list becomes worthless. In any case, in the absence of
any epistemological explanation of why statements of these
kinds and only statements of these kinds should be capable of
rational verification such a list would only be a mere hypothesis
about the limits of rational inquiry. A mere hypothesis, whether
in the form of a list or in some other form, could not have the
exclusionary force that the Logical Empiricists wanted 'criteria
of cognitive significance' to have.
'Method' fetishism
Since the answer to the question 'Why is it good to be rational?'
cannot be simply that rationality enables one to attain practical
goals, and cannot be simply that rationality enables one to dis-
cover means/ends connections, we may consider another possi-
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
ble answer which has had a considerable amount of appeal at

different times. Many philosophers have believed that science
proceeds by following a distinctive method; if there is in fact a
method with the property that by using that method one can
reliably discover truths, and if no other method has any real
chance of discovering truths, and if what explains the extraor-
dinary success of science, and the persistence of controversy in
fields other than science, is that science and science alone has
consistently employed this method, then perhaps rationality, to
the extent that there is such a thing, should be identified with the
possession and employment of this method. The answer to the
question 'Why is it good to be rational?' would then be that it is
good to be rational because if one is rational one can discover
truths (of whatever kind one is interested in), whereas if one is
not rational one has no real chance of discovering truths, save
by luck. Like the instrumentalists' view this view went through
a philosophical history of rise, stagnation, and decline. From the
publication of Mill's Logic in the 1840s until the publication of
Carnap's Logical Foundations of Probability, influential philos-
ophers of science continued to believe that something like a for-
mal method {'inductive logic') underlies empirical science, and
that continued work might result in an explicit statement of this
method, a formalization of inductive logic comparable to the
formalization of deductive logic that was achieved starting with
the work of Frege in 1879. If such a method had been discov-
ered, then even if this did not by itself prove that the method
exhausts rationality, still the burden of proof would have been
very much upon those who claimed that there were truths which
could be justified or shown to be rationally acceptable by any
other method.
According to the most influential school, the so-called 'Baye-
sian' school, the general character of this inductive method that
philosophers have been trying to formalize is as follows: we
assume or pretend that the language of science has been formal-
ized and that scientists have available a certain number of relia-
ble observations expressible by 'observation sentences' in this
formalized language. We also assume that the various
hypotheses under consideration are expressed by formulas of
this language. The problem of inductive logic is taken to be the
problem of defining a 'confirmation function', that is a probabil-
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
ity function which will determine the mathematical probability

of each one of the hypotheses relative to the observational evi-
dence or, in another terminology, the 'degree of support' the
evidence lends to each of the alternative hypotheses. Usually one
assumes that one knows the probability that the given evidence
would have been obtained if each of the alternative hypotheses
were true; this is the so-called 'forward probability', i.e. the
probability of the evidence given the hypothesis. What we wish
to calculate is the so-called 'inverse probability', that is the prob-
ability of the hypothesis given the evidence. Bayes' theorem gives
this 'inverse probability' as a simple function of the forward
probabilities and certain other probabilities, the so-called 'prior
probabilities' of the alternative hypotheses, i.e. the probabilities
or 'subjective degrees of belief assigned by scientists to those
alternative hypotheses prior to examining the observational evi-
dence.
The 'forward probabilities' are indeed easy to calculate in the
two most common cases: they are easy to calculate when (a) the
hypothesis actually implies the evidence (in this case the 'for-
ward probability' of the evidence given the hypothesis is one); or
when (b) the hypothesis is itself a statistical or stochastic hypoth-
esis part of whose content is that the particular evidence
obtained should occur with a certain probability r. The difficulty
in applying Bayes' theorem — a difficulty so serious that both
philosophers and statisticians are deeply divided over the impor-
tance and usefulness of Bayes' theorem in the case of the confir-
mation of theories — is the need for a prior probability metric, a
set of 'subjective degrees of belief, in the terminology of De
Finetti and Savage.
Let us confine ourselves for the moment to hypotheses which
are such that the 'forward probabilities' can really be computed.
For hypotheses of this kind the method just described is indeed
a purely formal method; that is we could program a computing
machine to compute the degrees of support of the various hypoth-
eses given the appropriate 'inputs'. But the inputs would have
to include not only the computable 'forward probabilities', but
also the prior probability metric in the given context. If we think
of this prior probability metric as representing the scientists'
antecedent beliefs about the world, as the term 'subjective prob-
ability function' suggests, then it looks as if one of the inputs to
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
the method itself is a set of substantive factual beliefs (or degrees

of belief) about the world. This is the way in which many phi-
losophers of science today view the matter; increasingly it is
coming to be believed it is not possible to draw a sharp line
between the content of science and the method of science; that
the method of science in fact changes constantly as the content
of science changes. Bayes' theorem, if it really does capture the
logic of theory confirmation, provides a way of formalizing this
dependence of the method of science upon the content of science,
through the need for a prior probability function.
To put the matter somewhat more abstractly, we might say
that the 'method' fetishist assumes that rationality is inseparable.
But Bayes' theorem indicates that this is not the case; that we
can separate rationality, even in the area of science, even in the
special area where we are dealing with theories for which the
forward probabilities are computable, into two parts: a formal
part, which can be schematized mathematically and pro-
grammed on a computer, and an informal part which cannot be
so schematized and which depends on the actual changing beliefs
of scientists. Now it would be nice, to put it mildly, if the formal
part of rationality sufficed to guarantee good results. If we could
say that provided scientists make their observations carefully,
gather sufficient observations, and calculate degrees of support
according to Bayes' theorem, then eventually they will come into
agreement even if they disagree at the beginning, owing to the
difference in their subjective degrees of belief, then all would be
well. But there are two things wrong with this happy picture.
The first thing wrong is that even if we could show that in the
long run the 'prior probability function' cancels out, or that sci-
entists with different prior probability functions eventually come
into agreement provided they continue to gather more evidence
and to use Bayes' theorem, it would still be necessary that this
convergence be reasonably rapid. If scientists with different prior
probability functions will not come into agreement until the phe-
nomenon to be predicted has already taken place, or until mil-
lions of years have passed, then, in the short run, the fact that
there is some mathematical guarantee of eventual convergence is
useless; the trouble with long-run justifications is that the long
run may be much too long. In the famous words of John May-
nard Keynes, 'in the long run we'll all be dead'. The second thing
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
wrong is that it does turn out, as a matter of fact, that differences

in the prior probability function can lead to violent differences
in the actual degrees of support assigned to theories, and that
these differences can amount to what would ordinarily be con-
sidered as gross irrationalities.
To put this last point in another way, a scientist will only
assign degrees of support to hypotheses that look 'reasonable' if
he starts out with a 'reasonable' prior probability function. If a
person only obeys the formal part of the description of ratio-
nality, if he is logically consistent and assigns degrees of support
in accordance with Bayes' theorem, but his prior probability
function is extremely 'unreasonable', then his judgments of the
extent to which various hypotheses are supported by the evi-
dence will be (as scientists and ordinary people actually judge
these matters) wildly 'irrational'. Formal rationality, commit-
ment to the formal part of the scientific method, does not guar-
antee real and actual rationality.
The extent to which this is true is in fact rather shocking.
Arthur Burks has in fact shown that there are even 'counter
inductive prior probability functions'. That is, there is a certain
logically possible prior probability metric such that if a scientist
had that metric then as more evidence came in for a hypothesis
(using the term more evidence on the basis of our normal induc-
tive judgments) then the scientist would assign lower and lower
weight to the hypothesis for a very long time.
One way out of the difficulty might be to try to supplement
the present formal account of scientific method by a further set
of formal rules which would determine which priors are reason-
able (henceforth, I shall refer to a prior probability function sim-
ply as a 'prior', in conformity with common statistical usage),
and which priors are unreasonable. But there does not seem to
be any good reason to think that there would be a set of rules
which could distinguish between reasonable and unreasonable
priors and which would be any simpler than a complete descrip-
tion of the total psychology of an ideally rational human being.
The hope for a formal method, capable of being isolated from
actual human judgments about the content of science (that is,
about the nature of the world), and from human values seems to
have evaporated. And even if we widen the notion of a method
so that a formalization of the psychology of an ideally rational
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
human scientist would count as a 'method', there is no reason to

think that a 'method' in this sense would be independent of judg-
ments about aesthetics, judgments about ethics, judgments
about whatever you please. The whole reason for believing that
the scientific method would not apply to or presuppose beliefs
about ethical, aesthetic, etc., matters was the belief that the sci-
entific method was a formal method, after all.
My discussion has depended on assuming the correctness of
one particular approach to formalizing the scientific method, the
so-called 'Bayesian' approach. But similar problems arise in each
of the other approaches that have been attempted. Even if one
tries to isolate some small part of the inductive method which
would not be as 'high-powered' as the confirmation of theories,
and which would be more in line with what Bacon understood
by 'induction', that is, even if one tries to isolate a method for
confirming simple generalizations by examining a sufficient
number of instances and 'projecting' the truth of the generaliza-
tion, similar problems arise. Nelson Goodman5 has shown that
no purely formal rule for inductive projection can even be free
from inconsistencies; before a formal rule can even hope to yield
consistent results one has to have in advance segregated the
predicates of the language into those one is willing to regard as
'projectable' and those which one will treat as 'non-projectable'.
The fact that even the most elementary part of induction turns
out to have a part (namely the division of the vocabulary into a
projectable and a non-projectable part) which is informal, again,
strongly supports the conclusion suggested by our discussion of
Bayes' theorem, that one cannot draw a sharp line between the
actual beliefs of scientists and the scientific method. What Good-
man did was to invent a predicate 'grue' which applies to things
just in case they are observed prior to the year 2000 and green
or not observed prior to the year 2000 and blue. Prior to the
year 2000, everything which is examined and seen to be green is
also examined and seen to be grue. Any formal rule of projection
which told us that when we've examined a certain number of
things, say emeralds, with a property P then we are allowed to
infer that 'all emeralds are P* would permit us to make the con-
5
See his Fact, Fiction and Forecast, 2nd ed., Hackett (1977), first published
in 1954.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
tradictory inferences that 'all emeralds are green' and 'all emer-
alds are grue\ And Goodman convincingly shows that all
attempts to rule out 'bizarre' predicates like 'grue' on purely for-
mal grounds cannot work.6
There is actually a close connection between Goodman's dif-
ficulty in the case of Baconian induction and the need for a prior
in connection with Bayes' theorem. Suppose the two hypotheses
the scientist has to choose between (at some time prior to the
year 2000) are 'all emeralds are green' and 'all emeralds are
grue'. Let us suppose that the relevant evidence is that a great
many emeralds have been examined and all found to be green
(and hence all found to be grue as well). If the scientist computes
the degree of support of the two hypotheses using Bayes' theo-
rem then it turns out that he can either find a much higher degree
of support for the normal hypothesis ('all emeralds are green') or
a much higher degree of support for the abnormal hypothesis
('all emeralds are grue') or an equal degree of support for both
hypotheses, depending on his prior. If one's subjective probabil-
ity metric assigns a much higher prior probability to 'all emer-
alds are green' than to 'all emeralds are grue', then one will, in
fact, behave as if one were projecting 'green' and not projecting
'grue'. From a Bayesian point of view the need for a decision as
to which predicates are projectable and which are not before one
can make an induction is just a special case of the need for a
prior.
Karl Popper has suggested that one should accept the most
falsifiable of the alternative hypotheses; but it turns out that his
formal measures of falsifiability will yield different results
depending on which predicates of the language one chooses to
take as primitive. Whether one thinks of the scientist, as Popper
does, as trying to find the most falsifiable hypothesis that has not
6
Goodman's own solution is to consider form plus the history of prior
projection of the predicates involved in the inference (along with certain
related matters, e.g. 'entrenchment' and 'over-riding'). On Goodman's
proposal it would follow that a culture which had always projected such
'crazy' predicates as his celebrated predicate 'grue' would now be
perfectly justified in doing so — their inferences would now be induct-
ively valid'!
While I agree with Goodman that fit with past practice is an impor-
tant principle in science, Goodman's version of this principle is too
simple and too relativistic.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
yet been ruled out, or thinks in the more conventional way that
one is trying to compute degrees of support for hypotheses, the
need for an informal element corresponding to a Goodmanian
decision that certain predicates are projectable and others are
not, or corresponding to the acceptance of a Bayesian prior, is
still necessary.
At this point the reader may wonder, if there is no such thing
as the scientific method, or if the method, in so far as it can be
formalized, depends on inputs which are not formalizable, then
how do we account for the success of science? It is undeniable
that science has been an astoundingly successful institution. We
tend to feel that the reason for its success must have something
to do with the differences between the ways in which scientists
proceed to gather knowledge and the way in which people tra-
ditionally proceeded to gather knowledge in the prescientific
ages. Is this wholly wrong? The answer is that it is not. The
alternatives that we have to choose between are not that science
succeeds because it follows some kind of rigorous formal algo-
rithm, on the one hand, and that science succeeds by pure luck.
Starting in the fifteenth century, and reaching a kind of peak in
the seventeenth century, scientists and philosophers began to put
forward a new set of methodological maxims. These maxims are
not rigorous formal rules; they do require informal rationality,
i.e. intelligence and common sense, to apply; but nevertheless
they did and do shape scientific inquiry. In short, there is a sci-
entific method; but it presupposes prior notions of rationality.7
It is not a method de novo which can serve as the be all and end
all, the very definition of rationality.
One of the most important methodologists of the seventeenth
century was the physicist Boyle. Prior to the seventeenth century,
physicists did not sharply distinguish between actually perform-
ing experiments and simply describing thought experiments
which would confirm theories that they believed on more or less
a priori grounds. Moreover, physicists did not see the need to
publish descriptions of experiments which failed. In short,
experiments were conceived of largely as illustrations for doc-
7
Mill himself concedes this (in a remarkably grudging tone of voice) when
he writes that we cannot expect the inductive method to work 'if we
suppose universal idiocy to be conjoined with it' {Utilitarianism, Chapter
2).
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
trines believed on deductive and a priori grounds; not as evi-

dence for and against theories. Boyle wrote manuals of experi-
mental procedure, he emphasized the need for a sharp
distinction between thought experiments and actual experi-
ments, and he emphasized the need to give a complete descrip-
tion of all the experiments one performed, especially including
the experiments that failed. Boyle was himself a disciple of the
philosopher Francis Bacon, and Boyle was undoubtedly led to
an appreciation of the importance of these rules by Bacon's
inductive outlook; in fact, however, the specific instructions
given by Boyle may have been more important or as important
in shaping the course of physical inquiry as the more abstract
and schematic defense of inductive procedure given by Bacon.
Turning away from trying to establish theories a priori
towards trying to test theories by deriving testable conclusions
from them and performing experiments certainly was a meth-
odological shift. As we have seen, however, we cannot simply
identify being rational with believing theories solely because they
are supported by carefully performed experiments. For one
thing, even in science it is not always possible to perform con-
trolled experiments. Sometimes one has to rely on passive obser-
vation rather than on the kind of active intervention which is
implied by the term 'experiment'. And, as we saw before, even
when one has carefully performed experiments for the purpose
of choosing among alternative theories, the estimation of the
degree to which the experimental results support the various
alternative theories is still a wholly informal matter.
Against what we have been urging, Karl Popper has repeatedly
argued that there is & distinctive scientific method, it can be
stated, and we should rely only on it for discovering the nature
of the world.
Popper does think, however, that there are notions of ration-
ality which are wider than scientific rationality and which do
apply to the making of ethical decisions.
In Popper's conception, set forth in his influential book, The
Logic of Scientific Inquiry (Logik der Forschung), and in subse-
quent publications, Popper has argued that the scientific method
consists in putting forward 'highly falsifiable' theories; theories
that imply risky predictions. We then proceed to test all of the
theories until only one survives. We then accept the surviving
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
hypothesis as the one to go on for the time being, and repeat the
entire procedure. Since the elimination of all the theories but one
is made on deductive grounds — a theory is eliminated when it
implies a prediction which is definitely falsified - no use of
Bayes' theorem is required, and no estimation of degrees of sup-
port is involved, Popper claims.
One problem with Popper's view is that it is not possible to
test all strongly falsifiable theories. For example, the theory that
if I put a flour sack on my head and rap the table 99 times a
demon will appear is strongly falsifiable, but I am certainly not
going to bother to test it. Even if I were willing to test it I could
think of 10100 similar theories, and a human lifetime, or even the
lifetime of the human species, would not suffice to test them all.
For logical reasons, then, it is necessary to select, on methodo-
logical grounds, a very small number of theories that we will
actually bother to test; and this means that something like a
prior selection is involved even in the Popperian method. As I
remarked above, even Popper's computations of degrees of fal-
sifiability are sensitive to which predicates one considers as
primitive in one's language, and in that sense even the notion of
falsifiability requires a prior decision analogous to Goodman's
decision that certain predicates are 'projectable' and others are
not. Let us waive these technical points, however, which are not
of interest to us in our present discussion, in any case. Even if
the Popperian method is incomplete, and requires to be supple-
mented by a more intuitive method which we are not able to
formalize at the present time, could it not be that it describes a
necessary condition, if not a sufficient condition, for scientific
rationality? Could it not be, in short, that a necessary condition
for the acceptability of a scientific theory be that it have survived
a Popperian test? The Popperian test itself may involve a prior
selection of theories to test which is itself informal and for which
we do not have an algorithm; the calculation of which theories
are most strongly falsifiable may involve informal decisions for
which we do not have an algorithm; but we could still insist that
no theory be accepted unless a set of theories has first been
selected all of which are intuitively 'highly falsifiable', and unless
all those theories except the one which we accept have been sub-
sequently refuted by carefully performed experiments. In short,
could it not be that the advice we ought to give the scientist is:
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
proceed as Popper advises you should proceed, and, where Pop-

per's methods are not capable of being formalized, rely on your
intuition as to how the Popperian maxims should be interpreted?
And might it not be the case that the Popperian method, vague
and informal as it is in part, exhausts not only the notion of
scientific rationality, but all of cognitive rationality, that is,
might it not be the case that a statement is warrantedly assert-
ible, or rationally acceptable, if and only if it is implied by a
theory which can be accepted on the basis of a Popperian test?
The answer is that such a conception of rationality is too narrow
even for science. For one thing it would rule out the acceptance
of one of the most successful and widely admired of all scientific
theories, Darwin's theory of evolution by natural selection. This
is a consequence which Popper himself is willing to accept with
equanimity, but certainly the scientific community is not. The
theory of natural selection is not highly falsifiable; it does not
imply definite predictions such that if they come out wrong then
the theory is refuted. We accept the theory of natural selection
not because it has survived a Popperian test, but because it pro-
vides a plausible explanation of an enormous amount of data,
because it has been fruitful in suggesting new theories and in
linking up with developments in genetics, molecular biology,
etc., and because the alternative theories actually suggested have
either been falsified or seem wholly implausible in terms of back-
ground knowledge. In short, we accept the Darwinian theory of
evolution by natural selection as what Peirce called an 'abduc-
tion', or what has recently been called an 'inference to the best
explanation'. This is exactly the kind of inference that Popper
wanted to drive out of science; but scientists are not going to be
persuaded by Popper that they should give up theories which are
not strongly falsifiable in cases where those theories provide
good explanations of vast quantities of data, and in cases where
no plausible alternative explanation is in the field. Indeed, as I
have pointed out in another publication,8 Popper exaggerates
the extent to which even the theories of classical physics are
themselves strongly falsifiable.
We weaken our description of the scientific method still fur-
ther, then, by allowing the 'inference to the best explanation' as
8
See 'The Corroboration of Theories', in my Mathematics, Matter and
Method.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
a legitimate form of scientific inference-drawing, even when the

'best explanation' inferred to is not strongly falsifiable in Pop-
per's sense. The scientific method has now become a tremen-
dously vague thing;9 but this we expected to happen, anyway, in
view of the formal results in inductive logic we described above.
Could it be that the 'scientific method', described this vaguely is
now exhaustive? And could it be that given even such a vague
description, it is clearly the case that no value judgment is capa-
ble of being verified or confirmed by this method? The answer is
that if the scientific method is described simply as 'make experi-
ments and observations as carefully as you can, and then make
inferences to the best explanation and eliminate theories which
can be falsified by crucial experiments', then it is impossible to
see what cannot be verified by a method so vaguely described.
Suppose, for example, I want to verify the statement 'John is a
bad man.' I might argue as follows: 'The following are observed
facts, that John is inconsiderate, that John is extremely selfish,
and that John is a very cruel person. Someone who is incon-
siderate, selfish and cruel is prima facie a bad person; therefore
John is a bad man.' There are two points at which a defender of
the view that 'value judgments' cannot be 'scientifically verified'
might object to this argument. He might object to the last step;
that is the step from John is cruel, John is inconsiderate, John is
extremely selfish, to John is a bad man. Admittedly this is a con-
ceptual step; the claim is that there is a conceptual link between
being cruel, inconsiderate, and selfish and being morally bad.10
9
Alternatively, we could restrict the term 'scientific method' to refer to the
conscious application of maxims of experimental procedure, as I
recommended in Meaning and the Moral Sciences, and just stop trying
to make it so elastic that it can cover everything we call 'knowledge'.
10
The fact that a truth or an inference is of the sort we call 'conceptual'
does not mean that it must be purely linguistic in character (i.e. true by
virtue of arbitrary linguistic conventions). Philosophers of many
different tendencies have seen that concepts, fact, and observations are
interdependent. As we remarked in Chapter 6, concepts are shaped by
what we observe or intuit and in turn shape what we are able to observe
and intuit. In these respects, the inference in the text involving 'good' is
exactly analogous to the following inference involving 'conscious'. 'John
is speaking intelligently, acting appropriately, and responding to what
goes on; therefore John is conscious.' The conceptual link here is that
'speaking intelligently', 'acting appropriately\ 'responding to what goes
on' are prima facie reasons for attributing consciousness, in just the way
that inconsiderateness, selfishness, and cruelty are prima facie reasons
for attributing moral badness.
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
Of course, if there are no conceptual links among the moral

predicates then this is an invalid step; but why should one
believe that there are no conceptual links among the moral pred-
icates? It might be argued that the use of steps described as 'con-
ceptual' in an argument is itself unscientific; but it surely cannot
be maintained that there are no such steps in science itself. For
example, if I make the inference from Newton's description of
the solar system to the statement that 'it is the gravitational
attraction of the moon that causes the tides' then I am employing
my informal knowledge that there is a conceptual link between
statements about forces and statements of the form A caused B.
The word 'cause' does not even appear in Newton's description
of the solar system and of the tides; but I know that the gravi-
tational force that A exerts on B can be described as caused by
(the mass of) A simply by virtue of understanding Newton's the-
ory.
Of course, if we describe the scientific method as consisting in
the drawing of 'inferences to the best explanation', or whatever,
from 'observational statements' which are themselves in value-
neutral language, then we can rule out 'John is inconsiderate'
and 'John is selfish' as 'observation statements' (although, in
particular cases, it might be easier to get agreement on these than
on, say, whether an object is mauve). But such statements occur
constantly in the writings of, for example, historians. That his-
tory, clinical psychology and ordinary language description can
really avoid words like 'considerate' and 'selfish' altogether is
doubtful (and where to draw the line would be an immense
problem: is 'stubborn' value-neutral? is 'angry' value-neu-
tral? for that matter, is 'twisted her arm savagely' value-neu-
tral?). But, in any case, to identify rationality with scientific
rationality so described would be to beg the question of the cog-
nitive status of value judgments; it would be to say these judg-
ments are not rationally confirmable because they are value
judgments, for rationality has been defined as consisting exclu-
sively of raw and neutral observation and the drawing of infer-
ences from value-neutral premisses. But why should one accept
such a definition?
Cambridge
https://doi.org/10.1017/CBO9780511625398.010
Values, facts and cognition
I argued in Chapter 6 that 'every fact is value loaded and every

one of our values loads some fact'. The argument in a nutshell
was that fact (or truth) and rationality are interdependent
notions. A fact is something that it is rational to believe, or, more
precisely, the notion of a fact (or a true statement) is an ideali-
zation of the notion of a statement that it is rational to believe.
'Rationally acceptable' and 'true' are notions that take in each
other's wash. And I argued that being rational involves having
criteria of relevance as well as criteria of rational acceptability,
and that all of our values are involved in our criteria of rele-
vance. The decision that a picture of the world is true (or true by
our present lights, or 'as true as anything is') and answers the
relevant questions (as well as we are able to answer them) rests
on and reveals our total system of value commitments. A being
with no values would have no facts either.
The way in which criteria of relevance involve values, at least
indirectly, may be seen by examining the simplest statement.
Take the sentence 'the cat is on the mat'. If someone actually
makes this judgment in a particular context, then he employs
conceptual resources - the notions 'cat', 'on', and 'mat' - which
are provided by a particular culture, and whose presence and
ubiquity reveal something about the interests and values of that
culture, and of almost every culture. We have the category 'cat'
because we regard the division of the world into animals and
non-animals as significant, and we are further interested in what
species a given animal belongs to. It is relevant that there is a cat
on that mat and not just a thing. We have the category 'mat'
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
202 Values, facts and cognition
because we regard the division of inanimate things into artifacts

and non-artifacts as significant, and we are further interested in
the purpose and nature a particular artifact has. It is relevant
that it is a mat that the cat is on and not just a something. We
have the category 'on' because we are in te rested inspatial relations.
Notice what we have: we took the most banal statement im-
aginable, 'the cat is on the mat', and we found that the presup-
positions which make this statement a relevant one in certain
contexts include the significance of the categories animatel
inanimate, purpose, and space. To a mind with no disposition
to regard these as relevant categories, 'the cat is on the mat' would
be as irrational a remark as 'the number of hexagonal objects
in this room is 76' would be, uttered in the middle of a tete-a-tete
between young lovers.
Not only do very general facts about our value system show
themselves in our categories (artifacts, species name, term for a
spatial relation) but, as we saw in Chapter 6, our more specific
values (for example, sensitivity and compassion) also show up in
the use we make of specific classificatory words ('considerate',
'selfish'). To repeat, our criteria of relevance rest on and reveal
our whole system of values.
The relevance of this discussion of relevance to the question
raised in the preceding chapter ('What is the value of rational-
ity?') is immediate. If 'rationality' is an ability (or better, an inte-
grated system of abilities) which enables the possessor to deter-
mine what questions art relevant questions to ask and what
answers it is warranted to accept, then its value is on its sleeve.
But it needs no argument that such a conception of rationality is
as value loaded as the notion of relevance itself.
It may be objected that I have lumped together factors that
belong apart, however, and thereby masked a sort of sleight of
hand. The very fact that I have spoken of two factors, rational
acceptability and relevance, testifies, it may be claimed, to the
persistence and permanence of something like the fact/value
dichotomy. A rational person, on the conception the objector
has in mind, would be one who could tell what was and what
was not warrantedly assertible; what a person chose to regard as
interesting, or important, or relevant might have bearing in eval-
uating his character or even his mental health, but not his cog-
nitive rationality, the objector would say.
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
Values, facts and cognition 203
Acceptability and relevance are interdependent in any real

context, however. Using any word - whether the word be
'good', or 'conscious', or 'red', or 'magnetic' - involves one in a
history, a tradition of observation, generalization, practice and
theory.
It also involves one in the activity of interpreting that tradi-
tion, and of adapting it to new contexts, extending and criticiz-
ing it. One can interpret traditions variously, but one cannot
apply a word at all if one places oneself entirely outside of the
tradition to which it belongs. And standing inside a tradition
certainly affects what one counts as 'rational acceptability'. If
there were one method one could use to verify any statement at
all, no matter what concepts it contained, then the proposed sep-
aration of the ability to verify statements from the mastery of a
relevant set of concepts might be tenable; but we have already
seen that there is no reason to accept the myth of the one
Method.
The two-components theory

Our present intuitions about rationality seem to be in conflict;
certainly no one philosophical theory seems to reconcile them
all. On the one hand, it is simply not true that we never judge
ends as rational or irrational; on the other hand, when we are
confronted with a case like that of the hypothetical 'rational
Nazi' we do not see how to justify criticizing such an intelligently
elaborated and considered system of ends as irrational even if we
find it morally repellant.
One way in which it has been suggested we might resolve these
problems is the following: assuming a sharp fact/value dichot-
omy, we can justify condemning the man who is only interested
in knowing the number of hairs on people's heads as irrational
on the ground that he has an inadequate perception of facts
(what 'adequate' means is a problem, of course). The Nazi only
disagrees with us about values, which is why be is not irrational.
In between cases can be handled, perhaps, along the lines sug-
gested by Bernard Williams. In particular, then, our argument
against the method fetishist, that we cannot, without circularity,
rule out of the 'observational evidence' such descriptive judg-
ments as 'John is considerate', can be met by advancing the the-
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
ory that the ordinary language moral—descriptive vocabulary has

two 'meaning components' simultaneously. One component is a
factual component; there are certain generally accepted stan-
dards for considerateness, and 'John is considerate' conveys the
information that John meets those standards.1 But there is also
an emotive meaning component: 'John is considerate' conveys a
'pro-attitude' towards a certain aspect of John's conduct. What
can be rational, it is claimed, is the acceptance of the factual
component of the statement 'John is considerate'; acceptance of
the emotive meaning component, sharing the 'pro-attitude', is
what is neither rational nor irrational.
The notion of the 'factual' here comes from the individual phi-
losopher's preferred view of the Furniture of the World. For a
materialist philosopher, the 'factual' component in the meaning
of any statement has to consist in a statement expressible in the
vocabulary of physical science. But a difficulty comparable to
the difficulty that plagued phenomenalism at once arises.
Phenomenalism, we recall, was the doctrine that all meaning-
ful statements are translatable without residue into statements
about sensations. 'Physicalism' (as the type of Materialism we
are discussing came to be called) was the doctrine that the whole
'factual' meaning of any statement can be translated without res-
idue into the language of physics. And, once again, the doctrine
seems to be false.
To see why, consider not a moral-descriptive statement but a
psychological statement, say 'X is thinking about Vienna.' It is
obvious that even if there are necessary and sufficient conditions
expressible in terms of brain states or whatever for an arbitrary
organism to be thinking about Vienna, it would take an
unimaginably perfected neurological (or, perhaps, functionalist-
psychological) theory to say what these are. The conditions for
What 'standards' means here is a problem however. If the claim is

that 'John is considerate' is descriptively true (i.e. the 'factual com-
ponent' is true) if and only if most speakers would agree that John is
considerate, then it would follow from this analysis of the 'factual
component' that there cannot be a person whom a sensitive judge would
correctly classify as 'considerate' even though most speakers disagree.
Such an account of the 'factual component' would simply amount to the
claim that all truths (at least about 'standards') must be 'public', but
why should one believe this unless one is a Majoritarian? (As we saw in
Chapter 5, the claim that all truth is public is self-refuting.)
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
the truth of this statement are context dependent, interest rela-

tive, and vague. There is no reason to think that even 'in princi-
ple' there exists a finite expression in the language of physical
theory which (in any physically possible world) is true of an X if
and only if that X is thinking about Vienna. Not only might it
be false that a finite equivalent in physicalist language exists for
the ordinary language statement fX is thinking about Vienna';
even if such an equivalent does exist, the equivalence would be
equivalence on the basis of an empirical theory or group of the-
ories which are themselves not known (perhaps they are so com-
plicated human beings will never know them), and which are
certainly not part of the meaning of fX is thinking about
Vienna'. In short, it is just false that 'X is thinking about Vienna'
means 'X is in such-and-such a (physically or functionally speci-
fied) brain state.'
What holds for 'X is thinking about Vienna' will hold for any
ordinary language predicate whose conditions of application do
not mesh well with those which govern physical concepts. (X is
considerate' - even (X is brown', (X is an earthquake', and (X is
a person' - are also not translatable into the language of 'physi-
cal theory'. What this means is that, if there are two components
to the meaning of (X is considerate', then the only description
we can give of the 'factual meaning' of the statement is that it is
true if and only if X is considerate. And this trivializes the notion
of a 'factual component'.
To say that the 'two components' theory collapses is not to
deny that (X is considerate' normally has a certain emotive force.
But it does not always have it. As we pointed out in Chapter 6,
we can use the statement (X is considerate' for many purposes:
to evaluate, to describe, to explain, to predict, and so on. Distin-
guishing the uses to which the statement can be put does not
require us to deny the existence of such a statement as 'X is
considerate.'
Moore and the 'Naturalistic Fallacy'

Weber's claim that 'value judgments' cannot be rationally con-
firmed was the source of the present fact/value dichotomy; but
the dichotomy was reinforced by G. E. Moore (contrary to his
own intentions). Writing at a time when Bertrand Russell and
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
John Maynard Keynes, along with other future members of the

Bloomsbury group, were still young students, Moore argued for
the thesis that Good was a 'non-natural' property, i.e. one totally
outside the physicalistic ontology of natural science. His defense
of non-naturalism backfired; his students may have been con-
vinced by Moore that there were such things as 'non-natural
properties' (although Russell, at least, was to lose the faith) but
later philosophers of a naturalistic kind tended to feel that
Moore had provided a reductio ad absurdum of the idea that
there are such things as value properties. In the 1930s Charles
Stevenson and the Logical Positivists were to advance the 'emo-
tive theory of ethics', that is the theory that 'X is good' means 'I
approve of X, do so as well!', or something of that kind. Value
properties began to be rejected on epistemological grounds, but
even more on ontological grounds; as John Mackie2 has recently
expressed it, it is compatible with natural science that there
should be such things as value attitudes but it is not compatible
with natural science that there should be such things as value
properties. Value properties, Mackie claims, are 'ontologically
queer' — i.e. they are funny mysterious properties in whose exis-
tence scientifically enlightened people should not continue to
believe.
Moore's argument that Good cannot be a physicalistic prop-
erty (a 'natural' property) was that if'Good' is the same property
as 'conducive to maximizing total utility' (or whatever natural
property, physical or functional, you care to substitute), then
(1) 'this action is not good even though it is conducive to
maximizing total utility'
is a self-contradictory statement (not just a false one).
But even a Utilitarian would not claim (1) is self-contradic-
tory. And this shows, Moore claims, that although being Good
and being conducive to maximizing total utility might be corre-
lated properties, they could not be the same identical property.
Moore's argument turns on assumptions that I and many
other philosophers of language would reject today, however.
First of all, he implicitly denied that there could be such a thing
as synthetic identity of properties. But, as I pointed out in Chap-
2
Ethics, Inventing Right and Wrong, Penguin, 1977.
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
ter 4, this would rule out such accepted scientific discoveries as

the discovery that the magnitude temperature is the same mag-
nitude as the magnitude mean molecular kinetic energy. (One
could use Moore's 'proof to show that temperature must be a
'non-natural property', in fact. For one is not contradicting one-
self when one says 'x has temperature T but x does not have
mean molecular kinetic energy E', where £ is the value of the
kinetic energy that corresponds to temperature T, even if the
statement is always false as a matter of empirical fact. So, Moore
would have to conclude, Temperature is only correlated with
Mean Molecular Kinetic Energy; the two properties cannot lit-
erally be identical.) In fact, Moore conflated properties and con-
cepts. There is a notion of property in which the fact that two
concepts are different (say 'temperature' and 'mean molecular
kinetic energy') does not at all settle the question whether the
corresponding properties are different. (And discovering how
many fundamental physical properties there are is not discover-
ing something about concepts, but something about the world.)
The concept 'good' may not be synonymous with any physical-
istic concept (after all, moral-descriptive language and physical-
istic language are extremely different 'versions', in Goodman's
sense), but it does not follow that being good is not the same
property as being P, for some suitable physicalistic (or, better,
functionalistic) P. In general, an ostensively learned term for a
property (e.g. 'has high temperature') is not synonymous with a
theoretical definition of that property; it takes empirical and the-
oretical research, not linguistic analysis, to find out what tem-
perature is (and, some philosopher might suggest, what good-
ness is), not just reflection on meanings.
An idea which came into philosophy of language a few years
after I introduced the 'synthetic identity of properties', and
which enlarges and illuminates the point I have been making, is
Saul Kripke's idea of 'metaphysically necessary' truths which
have to be learned empirically, 'epistemically contingent neces-
sary truths'.3 Kripke's observation, applied to the
temperature/kinetic energy case, is that, if someone describes a
logically possible world in which people have sensations of hot
3
Kripke's Naming and Necessity, Harvard, 1980 (lectures originally given
in Princeton in 1970).
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
and cold, there are objects that feel hot and objects that feel cold,
and in which these sensations of hot and cold are explained by
a different mechanism than mean molecular kinetic energy, then
we do not say that he has described a possible world in which
temperature is not mean molecular energy. Rather we say that
he has described a world in which some mechanism other than
temperature makes certain objects feel hot and cold. Once we
have accepted the 'synthetic identity statement' that temperature
is mean molecular kinetic energy (in the actual world), nothing
counts as a possible world in which temperature is not mean
molecular kinetic energy.
A statement which is true in every possible world is tradition-
ally called 'necessary'. A property which something has in every
possible world is traditionally called 'essential'. In this tradi-
tional terminology, Kripke is saying that 'temperature is mean
molecular kinetic energy' is a necessary truth even though we
cant know it a priori. The statement is empirical but necessary.
Or, to say the same thing in different words, being mean molec-
ular kinetic energy is an essential property of temperature. We
have discovered the essence of temperature by empirical investi-
gation. These ideas of Kripke's have had widespread impact on
philosophy of language, metaphysics, and philosophy of mathe-
matics; applied to Moore's argument they are devastating.
Moore argued from the fact that (1) can only be false contin-
gently, that being P (for some suitable natural property P) could
not be an essential property of goodness; this is just what the
new theory of necessity blocks. All that one can validly infer
from the fact that (1) is not self-contradictory is that 'good' is not
synonymous with 'conducive to maximizing utility' (not
synonymous with P, for any term P in the physicalistic version of
the world). From this nonsynonymy of words nothing follows
about non-identity of properties. Nothing follows about the es-
sence of goodness.
Ruth Anna Putnam has pointed out that another common
argument that goodness cannot be a natural property does not
work.4 This is the argument that (X is good' has 'emotive force',
'expresses a pro-attitude', and so forth.
4
'Remarks on Wittgenstein's Lecture on Ethics', in Haller et al. (eds.),
Language, Logic, and Philosophy, Proceedings of the 4th Intern.
Wittgenstein Symposium, Vienna, 1980.
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
What is wrong with the argument, Ruth Anna Putnam points

out, is that many descriptive predicates naturally acquire an
emotive force. In our culture, 'slobbers his food all over his shirt'
has strong negative emotive force although the phrase is literally
a description. Any word that stands for something people in a
culture value (or disvalue) will tend to acquire emotive force.
The word 'good', in its moral sense, is applied to many things.
Some of these - good states of mind, for example - may be nat-
urally valued: it may be part of the content of that very state of
mind that one values being in it. I don't mean to suggest the
converse: that any state of mind that is naturally valued in this
sense is good; that would be clearly false. Suppose 'good' were
defined so that things which are naturally valued and are such
that there is no good reason to disvalue them (as there is reason
to disvalue certain drug-induced states which are naturally val-
ued) count as 'good'.
Then one would expect the statement that something is good
to have positive 'emotional force' because of the nature of the
property. Even if one is not a Consequentialist (i.e. one who
thinks everything with sufficiently good consequences is good),
there is no doubt that the most common reason for calling an
action good is that it has good consequences, among which
might be that it ultimately promotes states or situations which
are naturally valued; again, the very nature of the property
explains why the description comes to have 'pro' emotive force.
Mackie defends his claim that goodness is ontologically
'queer' by introducing as a premiss the assumption that one can-
not know that something is good without having a 'pro' attitude
towards that something. This amounts to assuming emotivism
in order to prove emotivism. The devils in hell are frequently
depicted as using 'good' with a negative emotive force ('He has
a deplorable tendency to moral goodness' one of them might
say); contrary to Mackie, I do not find such uses to be linguisti-
cally improper or to involve any contradiction. And do we not
hear people say, 'I know that's a bad thing to do but I don't
care'? As Philippa Foot has pointed out, one can even rebuff
appeals to morality by saying, 'I'm not out to be a good man.'
What the possibility of these utterances shows is that while there
is indeed a difference between the describing use of language and
the prescribing use or the commending use, this difference in
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
uses is not a simple function of vocabulary. 'Descriptive' words

can be used to praise or blame ('He slobbers food all over his
tie') and 'evaluative' words can be used to describe and explain.
(Consider the following dialogue: 'John must have been an
exceptionally good man to do such a thing.' 'No, he had never
been a moral paragon, if anything the contrary; but he must
have had a capacity for self-sacrifice we never suspected.' Here
moral language is being used in an explanatory function.) To
repeat Ruth Anna Putnam's point again, nothing about good-
ness not being a property follows from the fact that 'is good' is
used to commend.
Professor Putnam points out that there is, nevertheless, some-
thing right about Mackie's argument. Some moral expressions
undeniably do have a built-in orientation towards action.
'Should', 'ought', 'right', and 'must' are prime examples of such
'action guiding' words. The 'is/ought' problem is not the same as
the 'fact/value' problem, as she points out. 'I am not out to do
what I should' sounds much odder than 'I am not out to be a
good man' (and 'I am not out to do what I must' sounds crazy).
Mackie points out that no physical property has a built-in
connection to action (or to approval of an action), and concludes
that 'being the right thing to do', etc., are 'ontologically queer'.
But (besides depending on the assumption that the physicalist
version of the world is the One True Theory), this argument
proves too much. For some epistemic predicates (e.g. 'rationally
acceptable', 'justified belief) are also action-guiding (taking
'action' in a wide sense, so that accepting a statement counts as
an action). One can say, 'X is a good thing to do' and 'There is
a good deal of evidence that Y and not be committed to doing
(or prescribing) X or to accepting Y; but if one says 'X is the
right action to perform in this situation', or 'Believing that Y is
completely justified', then one is oriented to doing (or prescrib-
ing) X and to accepting Y. 'Justified' (in the case of beliefs) has
the characteristic of being action-guiding as much as 'right' in
the moral sense does.
If we now mimic Mackie's argument and conclude that there
is no such property as being justified, but only 'justification atti-
tudes', then we land ourselves in total relativism. Before going
so catastrophically far, we should pause to see why action-
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
guiding predicates seem 'ontologically queer' to a committed

physicalist.
In Chapter 2, I argued that reference itself should seem 'onto-
logically queer' to a committed physicalist. If only physicalistic
properties and relations really exist, then reference, to exist,
must be a physicalistic relation - but then the problem, as we
saw, is an overabundance of 'candidates'. There are an infinite
number of admissible reference relations (and all are physicalis-
tic, or at least naturalistic, if we count set theory as part of the
naturalistic version of the world). If one of these were the rela-
tion of reference, then that fact would itself be an ultimate meta-
physical fact of a very strange kind.
What would make such a fact strange is that we have built a
certain neutrality, a certain mindlessness, into our very notion of
Nature. Nature is supposed to have no interests, intentions, or
point of view. Given that this is right, then how could one admis-
sible reference-relation be metaphysically singled-out?
It is this same mindlessness of Nature that makes the action-
guiding predicates 'is right' and 'is a justified belief seem 'queer'.
If one physicalistic property P were identical with moral right-
ness or with epistemological justification, that would be
'queer' - queer for precisely the same reason that it would be
'queer' if reference were a physicalistic relation. It would be as if
Nature itself had values, in the moral case, or referential inten-
tions, in the semantical case.
For this reason, I think that Moore was right (even if his argu-
ments are not acceptable) in holding that 'good', 'right' (and also
'justified belief, 'refers', and 'true') are not identical with phys-
icalistic properties and relations. What this shows is not that
goodness, rightness, epistemic justification, reference, and truth
do not exist, but that monistic naturalism (or 'physicalism') is an
inadequate philosophy.
The 'rational Nazi' again

What troubled us earlier was that we did not see how to argue
that the hypothetical 'perfectly rational Nazi' had irrational
ends. Perhaps the problem is this: that we identified too simply
the question of the rationality of the Nazi (as someone who has
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
a world view or views) with the rationality of the Nazi's ends. If

there is no end 'in' the Nazi to which we can appeal, then it does
seem odd to diagnose the situation by saying 'Karl has irrational
goals.' Even if this is part of what we conclude in the end, surely
the first thing we want to say is that Karl has monstrous goals,
not that he has irrational ones.
But the question to look at, if we are going to discuss Karl's
rationality at all, is the irrationality of his beliefs and arguments,
not his goals.
Suppose, first, that Karl claims Nazi goals are morally right
and good (as Nazis, in fact if not in philosophers' examples, gen-
erally did). Then, in fact, he will talk rubbish. He will assert all
kinds of false 'factual' propositions, e.g. that the democracies are
run by a 'Jewish conspiracy'; and he will advance moral propo-
sitions (e.g. that, if one is an 'Aryan', one has a duty to subjugate
non-Aryan races to the 'master race') for which he has no good
arguments. The notion of a 'good argument' I am appealing to
is internal to ordinary moral discourse; but that is the appropri-
ate notion, if the Nazi tries to justify himself within ordinary
moral discourse.
Suppose, on the other hand, that the Nazi repudiates ordinary
moral notions altogether (as our hypothetical Super-Benthamite
did). I argued that a culture which repudiated ordinary moral
notions, or substituted notions derived from a different ideology
and moral outlook for them, would lose the ability to describe
ordinary interpersonal relations, social events and political
events adequately and perspicuously by our present lights. Of
course, if the different ideology and moral outlook are superior
to our present moral system then this substitution may be good
and wise; but if the different ideology and moral outlook are
bad, especially if they are warped and monstrous, then the result
will be simply an inadequate, unperspicuous, repulsive represen-
tation of interpersonal and social facts. Of course, 'inadequate,
unperspicuous, repulsive' reflect value judgments; but I have
argued that the choice of a conceptual scheme necessarily reflects
value judgments, and the choice of a conceptual scheme is what
cognitive rationality is all about.
Even if the individual Nazi does not lose the ability to use our
present moral descriptive vocabulary, even if he retains the old
notions somewhere in his head (as some scholars, perhaps, still
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
are familiar with and able to use the medieval notion of 'chiv-
alry'), still these (our present moral descriptive notions such as
'considerate', 'compassionate', 'just', 'fair') will not be notions
that he employs in living his life: they will not really figure in his
construction of the world.
Again, I wish to emphasize that I am not saying that what is
bad about being a Nazi is that it leads one to have warped and
irrational beliefs. What is bad about being a Nazi is what it leads
you to do. The Nazi is evil and he also has an irrational view of
the world. These two facts about the Nazi are connected and
interrelated; but that does not mean the Nazi is evil primarily
because he has an irrational view of the world in the sense that
the irrationality of his world view constitutes the evil. Neverthe-
less, there is a sense in which we may speak of goals being
rational or irrational here, it seems to me: goals which are such
that, if one accepts them and pursues them then one will either
be led to offer crazy and false arguments for them (if one accepts
the task of justifying them within our normal conceptual
scheme), or else one will be led to adopt an alternative scheme
for representing ordinary moral-descriptive facts (e.g. that
someone is compassionate) which is irrational, have a right to be
called 'irrational goals'. There is a connection, after all, between
employing a rational conceptual scheme in describing and per-
ceiving morally relevant facts and having certain general types
of goals as opposed to others.
'But what if the Nazi gives no reason for being a Nazi except
"that's how I feel like acting"?' This is a natural question, but
here surely the natural answer is also the right one: in such a
case the Nazi's conduct, besides being evil, would also be com-
pletely arbitrary. Notice that 'arbitrary' is one of the words I
have been calling 'moral—descriptive', i.e. a word which can be
used, without change of denotation, to evaluate (in this case to
blame), to describe ('John quite arbitrarily decided to change
jobs'), to explain (or to indicate that no explanation of a certain
kind can be given), etc. Indeed, when I just said that Karl's deci-
sion to be a Nazi (in the case described) would be completely
arbitrary, I was primarily describing, not evaluating. Many
things I do are, quite literally, arbitrary — e.g. choosing one path
across the campus rather than another; but this does not mean
there is anything wrong about these actions. (The matters are
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
simply too trivial.) Even if I do something important 'arbitrar-

ily' - say, change jobs - if I don't have family responsibilities,
etc., this may simply be my right. But if the action is one that
requires justification, then performing it arbitrarily and with no
justification will expose one to legitimate blame. Making a deci-
sion which adversely affects the lives of others (and perhaps
one's own life) to a great extent with no justification, just as an
arbitrary and willful (another of those moral descriptive words!)
act, is a paradigmatic example of irrationality, and not just irra-
tionality but perverseness.
We started our discussion in Chapter 7 by looking at Ben-

tham's claim that 'prejudice aside' the game of pushpin (an
ancient children's game similar to tiddlywinks) is just as good as
'the arts and sciences of music and of poetry'. In Bentham's view
the only reason poetry is better than pushpin, ultimately, is the
brute fact that poetry gives greater satisfaction than pushpin (or
gives satisfaction to more people, or both). There are, basically,
two things wrong with this view: one thing wrong is that 'satis-
faction' (or 'self-interest') itself cannot be an aim of any being
who does not have other aims. If I had no aim other than 'my
welfare', then my 'welfare' would be a meaningless notion, a
point which goes back to Bishop Butler. More important, some
satisfactions are better and 'nobler' than others, and one can give
reasons why. Poetry and music give solace, they enlarge our sen-
sibilities, they provide important modes of self-expression to
many people, including many of the most gifted people the
human race has produced. Calling these reasons for valuing cer-
tain satisfactions above others 'prejudice' is actually closely con-
nected with both the 'two components' theory and the idea that
value properties are 'ontologically queer'. Bentham is operating
with the model of 'neutral facts' and arbitrary 'prejudices'.
Indeed, calling the preference for poetry a 'prejudice' is just Ben-
tham's way of suggesting that the fact that poetry gives greater
satisfaction than pushpin is the only consideration that is not
'arbitrary' in comparing the two; any preference for one kind of
satisfaction over another (it is suggested) is arbitrary. But this is
simply false, given the actual place in our conceptual scheme of
the notion of an 'arbitrary' preference, and meaningless if 'arbi-
trary' is wrenched out of the scheme to which it belongs. (Simi-
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
larly, the statement that preferring poetry to pushpin is a preju-

dice is literally false.) It is being suggested that it is somehow
ontologically legitimate to admit that there are such things as
satisfactions, but not ontologically legitimate to admit such
things as enlarged sensibilities, enlarged repertoires of meaning
and metaphor, modes of expression and self-realization, and so
on. The idea that values are not part of the Furniture of the
World and the idea that 'value judgments' are expressions of
'prejudice' are two sides of the same coin.
We have investigated the question whether 'value judgments'
can be rationally supported. We have seen that various negative
answers rest on dubious philosophical assumptions: that ration-
ality itself is only good for 'prediction', or only good for getting
'consensus', or that there is only one Method for arriving at
truth (where, sometimes, the only criteria for 'truth' are said to
be prediction and consensus), or that value judgments have 'two
meaning components', or that value properties are 'ontologically
queer'. The position I have defended is that any choice of a con-
ceptual scheme presupposes values, and the choice of a scheme
for describing ordinary interpersonal relations and social facts,
not to mention thinking about one's own life plan, involves,
among other things, one's moral values. One cannot choose a
scheme which simply 'copies' the facts, because no conceptual
scheme is a mere 'copy' of the world. The notion of truth itself
depends for its content on our standards of rational acceptabil-
ity, and these in turn rest on and presuppose our values. Put
schematically and too briefly, I am saying that theory of truth
presupposes theory of rationality which in turn presupposes our
theory of the good.
'Theory of the good', however, is not only programmatic, but
is itself dependent upon assumptions about human nature, about
society, about the universe (including theological and metaphys-
ical assumptions). We have had to revise our theory of the good
(such as it is) again and again as our knowledge has increased
and our world-view has changed.
It has become clear that in the conception I am defending there
is no such thing as a 'foundation'. And at this point people
become worried: are we not close to the view that there is no
difference between 'justified' and 'justified by our lights' (relativ-
ism) or even 'justified by my lights' (a species of solipsism)?
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
The position of the solipsist is indeed the one we will land in

if we try to stand outside the conceptual system to which the
concept of rationality belongs and simultaneously pretend to
offer a more 'rational' notion of rationality! (Many thinkers
have fallen into Nietzsche's error of telling us they had a 'better'
morality than the entire tradition; in each case they only pro-
duced a monstrosity, for all they could do was arbitrarily wrench
certain values out of their context while ignoring others.) We
can only hope to produce a more rational conception of ratio-
nality or a better conception of morality if we operate from
within our tradition (with its echoes of the Greek agora, of New-
ton, and so on, in the case of rationality, and with its echoes of
scripture, of the philosophers, of the democratic revolutions, and
so on, in the case of morality); but this is not at all to say that all
is entirely reasonable and well with the conceptions we now
have. We are not trapped in individual solipsistic hells, but
invited to engage in a truly human dialogue; one which com-
bines collectivity with individual responsibility.
Does this dialogue have an ideal terminus? Is there a true con-
ception of rationality, a true morality, even if all we ever have
are our conceptions of these? Here philosophers divide, like
everyone else. Richard Rorty, in his Presidential Address5 to the
American Philosophical Association, opted strongly for the view
that there is only the dialogue; no ideal end can be posited or
should be needed. But how does the assertion that 'there is only
the dialogue' differ from the self-refuting relativism we discussed
in Chapter 5? The very fact that we speak of our different con-
ceptions as different conceptions of rationality posits a Grenz-
begriff, a limit-concept of the ideal truth.
5
'Pragmatism, Relativism and Irrationalism', Proceedings and Addresses
of the American Philosophical Association, August 1980. See also
Rorty's Philosophy and the Mirror of Nature, Princeton University Press,
1979.
Cambridge
https://doi.org/10.1017/CBO9780511625398.011
Appendix
Here is the Theorem referred to in Chapter 2.

Theorem Let L be a language with predicates FUF2, . . . yFk
(not necessarily monadic). Let / be an interpretation, in the sense
of an assignment of an intension to every predicate of L. Then if
I is non-trivial in the sense that at least one predicate has an
extension which is neither empty nor universal in at least one
possible world, there exists a second interpretation / which dis-
agrees with I, but which makes the same sentences true in every
possible world as / does.
Proof Let W1,W2, . . . , be all the possible worlds, in some
well-ordering, and let Ut be the set of possible individuals which
exist in the world Wt. Let Ry be the set which is the extension of
the predicate ¥t in the possible world Wj according to / (if Ftj is
non-monadic, then R^ will be a set of « r tuples, where nx is the
number of argument places of JF*). The structure (Uj\R\j (i = 1,2,
. . . ,&)) is the 'intended model' of L in the world Wj relative to
/ (i.e. Uj is the universe of discourse of L in the world Wj, and
(for / = 1,2, . . . Jz) Rfj is the extension of the predicate F, in Wj.
If at least one predicate, say, Fu, has an extension Ruj which is

neither empty nor all of Uj, select a permutation Pj of Uj such
that Pj(Ruj)=?RUj. Otherwise, let Pj be the identity. Since Pj is
a permutation, the structure (Uj;Pj{Rij) (z = l,2, . . . ,&)) is iso-
morphic to (Uj;Rij (i = 1,2, . . . ,&)) and so is a model for the
same sentences of L (i.e. for the sentences of L which are true
under / in Wj).
Let/ be the interpretation of L which assigns to the predicate
F( (/ = 1,2, . . . ,&) the following intension: the function fi(W)
Cambridge
https://doi.org/10.1017/CBO9780511625398.012
218 Appendix
whose value at any possible world Wj is Pj{Rij). In other words,

the extension of Ft in each W5 under the interpretation / is
defined to be ?,(£„). Since < 1/^(1? y ) (i = l,2, . . . , k)) is a
model for the same set of sentences as (UjjRy (/ = 1,2, . . . , & ) )
(by the isomorphism), the same sentences are true in each possi-
ble world under/ as under I, a n d / differs from / in every world
in which at least one predicate has a non-trivial extension, q.e.d.
Comment: If, in a given world Wjf there are two disjoint sets
which are extensions of predicates of L in Wj under / - say, the
set of cats and the set of dogs - then, if there are more dogs than
cats (respectively, at least as many cats as dogs) we can take any
set of dogs the same size as the set of cats (respectively, any set
of cats the same size as the set of dogs) and choose a Pj which
maps the selected set of dogs onto the set of cats (respectively,
the selected set of cats onto the set of dogs) and vice versa; this
will ensure that under/ the extension of the first predicate — the
one whose extension under / is the set of cats — is a set of dogs
under / in Wj, or the extension of the second predicate - the
one whose extension under / is the set of dogs - is a set of cats
under / in Wj.
Second Comment: If there are objects — say, 'sensations' —

which one wishes not to be permuted, because one regards pred-
icates of those objects as 'absolute' in some sense, one simply
stipulates that the permutations Pj are to be the identity on those
objects. This will have the effect of making the restriction of any
predicate of L to those privileged objects the same under / and
under / in each world.
Third Comment: Since sentences receive logically equivalent

truth conditions under / and under /, it follows that on the stan-
dard 'possible worlds semantics', counter factual conditionals are
also preserved.
Cambridge
https://doi.org/10.1017/CBO9780511625398.012
Index
Althusser, L., 158-60 consciousness, 85-102; see mind-body

anthropology (relativism in), 161-2 problem
Apel, K. O., 175, 178 correlation (mind-body), 80-1; see
a priori (and identity theory), 82-4 mind-body problem
Aristotle, 57-8, 135, 148, 177 correspondence, 38-41; theory of
truth, 51, 56-69, 72-4
Bacon, Francis, 96
Baker, J., 105 Darwin, 109, 198
bats (sensations of), 92-3 Davidson, D., 116, 124
Bayes' theorem, 190-4 Da Vinci, Leonardo, 76
Bayesian school, 189-95 De Finetti, B., 190
Bentham, 151-2, 168-73, 214, see Dennett, D., 28, 89, 91
super-Benthamites Descartes, 57, 75-7
Benthamite psychology, 168-73 Dewey, J., 162, 167-8
Berkeley, 58-60, 64, 180 Diderot, 78
Block, N., 78, 92 Divine Right of Kings, 156-7
Boyd, R., 166 Dummett, M., 56-7
Boyle, 195-6
brains in a vat, 1-21, 130-5 Eccles, J., 91
Burks, A., 192 Einstein, 124
empiricism, 64-71, 124, 180-8
Carnap, R., 26, 88, 106, 112, 125, emotive force of ethical utterances,
136, 163, 181-4, 189 208-11
causal chain of the appropriate type ethics, 'inverted pyramid' picture,
(theory of reference), 14, 70, 51-4, 141-2; and projection, 141-7; and
65-6 metaphysical realism, 143-7; and
causal realism, 60 authoritarianism, 147-9; relativities
Cavell, S., 110 in, 148; non-cognitivism, 149
Chomskian linguistics, 126 Eudaemonia, 134, 148
Churchill, 1, 4, 13 evolution, 38-41, 198; Wittgenstein's
Comte, 176, 184 attitude towards, 109
concepts, 18-21 Existentialist-Positivist model, 154
Cambridge
https://doi.org/10.1017/CBO9780511625398
220 Index
extension, 18, 25-9; received view of Kant, x, 31, 56, 57, 60-4, 74, 83,116,
how fixed, 32ff; and non-standard 121, 128
interpretations, 29-35, 217-18; see Keynes, J. M., 191, 205
reference Kohler, W., 152
externalist view, 49ff Kolers, P., 68
Kripke, S., 46-7, 207-8
fact—value dichotomy, 127-49, Kuhn, T., ix, 38, 113-19, 126, 150
201-16; not to be drawn on basis of
vocabulary, 138-9; and subjectivism Leibniz, 75
about goodness, 141-7 Lenin, 124
falsifiability, 194-8 Lewis, C. S., 147
Feyerabend, P., ix, 113-19, 126 Locke, 47, 57, 61, 180
Field, H., 45-6 logical positivism, see Carnap, Steven-
Foot, P., 209 son, empiricism
Foucault, M., ix, 121, 126, 155-62
Frege, G., 27, 124-5 Mach, 124
Freud, 157 Mackie, J., 206-11
functionalism, 78-82; see mind-body majoritarianism, 177-8
problem Malament, D., 90
Marx, 157
Garfinkel, A., 119-20 Marxism (of Althusser), 158-60
Glymour, C., 90 meaning, 29; see extension, index, in-
Goodman, N., 68-9, 74, 79, 98, 123, tension, interpretation, intentions,
125, 146, 193-4 intentionality, notional world,
Grice, P., 105 reference, truth, two-components
Griffin, D., 92 theory
'The Meaning of "Meaning" ', 22-5
Harre, R., 109 (summary of the theory)
Hegel, xi, 158 mentalism, 79
history, x, 155-8 metaphysically necessary truths, 46-7,
Hume, 107, 124, 180 207-8
Husserl, E., 28 metaphysically unexplainable facts (on
hydra-headed robot, 96-7 physicalist theory of reference), 46-8
metaphysical realism, 134, 143-7; see
identity theory, 77-9; functionalism, correspondence theory of truth, ex-
78-82; and synthetic identity of ternalist view, internalist view,
properties, 84-5; and split brains, non-realist semantics, reference
85-92; and a priori, 82-4; and con- qualia
sciousness, 85-102 method fetishism, 188-200, 203
incommensurability, 113-19 Mill, John Stuart, 180, 189
index (in semantics), 26 mind-body problem, 75-102; and
inductive logic, 125-6, 189-94 parallelism, 76-7; and interac-
'instrumentalism', 178-80 tionism, 76-7; and identity theory,
intension, 25-9; and meaning, 27; 77-9; and mentalism, 79; role of
and Sinn, 27; and non-standard in- physics in, 75-6; functionalism,
terpretations, 29-35, 217-18 78-82; correlation, 80-1; and
intentionality, 2, 17ff synthetic identity of properties,
intentions, 41-3 84-5; and split brains, 85-92; and
interactionism, 76-7 consciousness, 85-102; and subjec-
internalist view, 49ff; and Kant, 60ff tive color, 79-81, 86-94; and
internal realism, see internalist view realism about qualia, 85, 88-9,
interpretation, 29-35, 217-18 94-6, 99-102; and a priori, 82-4
intrinsic properties, 36-8 Monod, J., 109
Cambridge
https://doi.org/10.1017/CBO9780511625398
Index 221
Montague, R., 26 and perspicuousness, 137; and

Moore, G. E., 205-11 perception, 137-8; and interper-
Moses, 158 sonal situations, 138-9; and value
Murdoch, I., 139, 154, 167 terms, 138-41; see rationality
rationality, 103-5, 163-4, 174-200;
Nagel, T., 92 and reasonableness, 107; criterial
naturalistic fallacy, 205-11 conceptions of, 105-13; and
natural kind terms, 22-5, 46 philosophy, 113; ordinary language
Neurath, O., 106 philosophers' view of, 110; scien-
Newman, John (Cardinal), 136, 163 tism and, 124-6; modern instru-
Newton, 58, 73, 75, 76, 200, 216 mental notion, 168-9; of pig-men,
Nietzsche, 157, 216 171-2; modern versus ancient views,
non-realist semantics, 56; see inter- 173; and technological success,
nalist view 175-6; and 'instrumentalism',
Nozick, R., 36, 122, 164-6 178-80; and empiricism, 180-8; and
tradition, 203; and majority agree-
Ockham's razor, 133 ment, 177-8; and method fetishism,
188-200, 203; and inductive logic,
onto logical queerness, 206-11 189-94; and falsifiability, 194-8;
organic unity, 152 and theory of the good, 215-16; and
solipsism, 215-16; as Grenzbegriff
parallelism (mind-body), 76-7 (limit conception), 216; see rational
Peirce, C. S., 30, 74, 198 acceptability
phantasm, 57 rational Nazi, 168-71, 211-14
phenomenalism, 180-5, 187 realism, see correspondence theory of
philosophical discussion (compared truth, externalist view, internalist
with political), 164-6 view, Dummett, reference, qualia,
Plato, 75, 120-1, 124, 155 metaphysical realism, causal
Platonism, 69 realism, causal chain of the appro-
Popper, K., 166, 181-4, 195-9 priate type
primary qualities, 57-61 reason, see rational acceptability, ra-
Private Language Argument, 64-72; tionality
further interpreted, 121-4 reductionism, 56-7
projection, 141-7 reference, magical theories of, 3-5, 16,
properties (synthetic identity of), 84-5, 51; and use, 8-12, 17ff; Turing Test
207 for reference, 9; causal theories of,
Putnam, R. A., 208-10 14, 45-8, 51ff; not in the head,
22-5; operational and theoretical
qualia, 75-102; subjective color, constraints on, 29-32; of natural
79-81, 86-94; realism about, 85, kind terms, 22-5; and Self-
88-9, 94-6, 99-102 Identifying Objects, 51, 53-4; inter-
Quine, W., 30, 33, 35, 44-5, 82-3, nalist view of, 52; causal chain of
113, 116, 124 the appropriate type and, 14, 51-4,
65-6, 70; similitude, 70-1, 56ff; ex-
rational acceptability, 103-26; logical ternalist view of, 49ff; see extension,
positivist conception of, 105-13; truth
'anarchist' conception (Feyerabend), relativism, 54, 119-24, 151-62; in
113-19; and relativism, 119-24; and anthropology, 161-62; false rel-
inductive logic, 125; and scientism, ativism, 166; objective relativism,
126; and science, 134; and empiri- 167-8; B. A. O. Williams' discus-
cal world, 134; and optimal sion of, 168-73
speculative intelligence, 134; and relativity of perception, 59-60
'real world', 135; role of adequacy relevance, 201-3
Cambridge
https://doi.org/10.1017/CBO9780511625398
222 Index
robots (whether they could have sen- Tarski, A., 128-9

sations), 96-7 taste, 152-5
rocks (whether they could have sensa- tradition, 203
tions), 94-6 truth, 49-50, 55-6; idealization theory
Rorty, R., 216 of, 56; correspondence theory of,
Russell, B., 99, 205, 206 51, 56-69, 72-4, 130; Tarski's
theory, 128-30; equivalence princi-
Savage, L. J., 190 ple (Tarski's Convention T), 128-9;
scepticism, 162-3 see reference, externalist view, in-
science, 128-37, 174-200; Boyle's ternalist view
contribution to methodology of, Turing, A., test for consciousness (Im-
195-6; and value judgments, 198- itation Game), 8-12; Test for Refer-
200; see rationality, ence, 9
majoritarianism, method fetishism, Twin Earth, 18-19, 22-5
inductive logic, falsifiability two-components theory of meaning,
secondary qualities, 57-61; 'all prop- 203-5
erties are secondary', 61-4
Self-Identifying Objects, 51, 53-4 utilitarianism, 151; super-
sensations, 54; empiricist attitude to- Benthamites, 139-41; 'Benthamite
wards, 64-72; possibility of always psychology', 168-73; see Bentham
being wrong about, 70-2
Sextus Empiricus, 147 values, in science, 132-5; truth (purely
similarity, 'of the same kind as', 53; formal value), 129, coherence,
similitude theory of reference, 56ff; 132-3, comprehensiveness, 133,
and Kant, 60ff functional simplicity, 133, instru-
similitude (not the mechanism of mental efficacy, 134-5, and total
reference to sensations), 70-1 human flourishing (Eudaemonia),
Skolem-Lowenheim theorem, 7, 67 134; ethical, 139-47; relativities in,
Smart, J. J. C, 78, 115 148; ethical values and
Spinoza, 75, 78 Eudaemonia, 148
split brains, 85-92
Stevenson, C, 206 Weber, M., 174-9
subjectivism (about goodness), 141-7, Wiggins, D., 51, 147-9
149; see relativism Williams, B. A. O., 169-73, 203
substantial forms, 57 Wittgenstein, L., 3, 7, 20-1, 62, 66-71,
super-Benthamites, 139-41 74, 107-9, 113, 121-4, 128
synthetic identity of properties, 84-5,
207 Zemach, E., 51
Cambridge
https://doi.org/10.1017/CBO9780511625398

Reason, Truth and History

Uploaded by

Copyright:

Available Formats

Reason, Truth and History

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reason, Truth and History

Uploaded by

Copyright:

Available Formats

REASON, TRUTH AND HISTORY

CAMBRIDGE UNIVERSITY PRESS

© Cambridge University Press 1981

This book is in copyright. Subject to statutory exception

First published 1981

Library of Congress Cataloguing-in-Publication Data is available

ISBN 0-521-23035-7 hardback

Transferred to digital printing 2004

A reader who is unused to technical philosophy, or who wishes

and ideological fervor are sometimes necessary to bring moral

scientific world is in some way constructed out of 'sense data'

sense-data' (the older empiricist view), or to deny that there is a

An ant is crawling on a patch of sand. As it crawls, it traces a

to Churchill that a picture - even a line drawing- has. If simi-

Magical theories of reference

which is exactly like one of my mental images of a tree as a result

ously has appropriate mental images, but neither understands

The case of the brains in a vat

placed in a vat of nutrients which keeps the brain alive. The

but it is not logically impossible (though it is certainly highly

go on 'fooling' each other forever, even if the rest of the world

happily in the same way. That is why the machine cannot be

Brains in a vat (again)

tative identity) between the thoughts of the brains in a vat and

The premisses of the argument

claims of their profession at the same time (the typical state of

of which they can be described. But why should we accept these

The reasons for denying necessary connections between

is nothing at all unimaginable about this. And what this shows

represents a different liquid when used on Twin Earth and when

ability of the thinker to employ these phenomena, to produce

identical with mental objects of any kind. For, assuming that

Why is it surprising that the Brain in a Vat hypothesis turns out

miraculous accident they just evolved resembling us and speak-

It may be objected that there might well be experts on Twin

or 'water is a mixture of two liquids', or does not change its

Intentions, extensions, and 'notional worlds'

that is a function which determines an extension in each context

What the intension does is to specify how the extension

requires being directly or indirectly linked to actual water

The received view of interpretation

how to go about testing the theory); on the more recent picture,

ory. For example, 'an admissible interpretation is such that it

constraints which, like 'simplicity', refer to properties of the set

Why the received view doesn't work

conditions for whole sentences. The idea, as we just saw, is that

The idea is that sentence (1) will receive a new interpretation

Now, in possible worlds falling under case (a), 'A cat is on a

word 'cat' by assigning to it the intension we just assigned to

'Intrinsic' and 'extrinsic'

the way we have been thinking. Perhaps the second reaction is to

'Survival' and evolution

which 'Hamlet' only refers to a person in a play); the success of

of external things, and beliefs about our own characteristics and

have certain truth conditions (and certain action conditions, or

Intentions: pure and impure

state, although they could be in the corresponding 'bracketed'

The origin of the puzzle

or intentions and bracketed intentions) itself explains part of the

by operational and theoretical constraints), then two terms