REASON, TRUTH AND HISTORY

Hilary Putnam

CAMBRIDGE
UNIVERSITY PRESS
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 Cambridge CB2 2RU, UK
Building, http: //[Link]
40 West 20th Street, York, NY 10011-4211, USA
New http: //www [Link]
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
Cambridge University Press.
the written permission of

First published 1981

Reprinted 1982, 1985, 1986, 1987, 1989, 1990, 1991, 1992, 1993, FO R RUTH ANNA
1994, 1995, 1997, 1998

Printed in the United States of America

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
Contents
 Note Preface

A reader who unused to technical philosophy, or who wishes

is
In the present work, the aim which I have in mind is to break the
to gain an overview of the argument of this book, might well which a number of dichotomies appear to have on
strangle hold
start by reading Chapter 5 to the end of the book, and only then
the thinking of both philosophers and laymen. Chief among
return to Chapters 1 to 4.
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 putforward 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
 Preface xi
Preface

and ideological fervor are sometimes necessary to bring moral scientificworld is in some way constructed out of 'sense data'
seriousness to an issue. But in time, both in philosophy and pol-
and the idea that terms in the laboratory sciences are 'operation-
itics, new ideas become old ideas; what was once challenging,
ally defined'. I shall not devote much space to criticizing opera-
tionalist and have been thor-
positivist views of science; these
becomes predictable and boring; and what once served to focus
oughly twenty-odd years. But the empiricist
criticized in the last
attention where it should be focussed, later keeps discussion
from considering new alternatives. This has now happened in idea that 'sense data' constitute some sort of objective 'ground
floor' for at least a part of our knowledge will be reexamined in
the debate between the correspondence views of truth and sub-
the light of what we have to say about truth and rationality (in
jectivist views. In the first three chapters of this book I shall try
which unites objective and sub-
to explain a conception of truth
Chapter 3).
In short, I shall advance a view in which the mind does not
jective components. This view, in spirit at least, goes back to
ideas of Immanuel Kant; and it holds that we can reject a naive
simply 'copy' a world which admits of description by One True

'copy' conception of truth without having to hold that

Theory. But my view is not a view in which the mind makes up
it's all a
matter of the Zeitgeist, or a matter of 'gestalt switches', or the world, either (or makes it up subject to constraints imposed
all a
matter of ideology. by 'methodological canons' and mind-independent 'sense-data').
If one must use metaphorical language, then let the metaphor be
The view which I shall defend holds, to put it very roughly,
this: the mind and the world jointly make up the mind and the
that there an extremely close connection between the notions
is

of truth and rationality; that, to put it even more crudely, tjje world. (Or, to make the metaphor even more Hegelian, the Uni-
only criterion for what verse makes up the Universe - with minds - collectively - play-
is a fact is what it is rational to accept.
mean ing ii special role in the making up.)
(I 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 A final feature of my account of rationality is this: I shall try

that the pictureis beautiful.) There can be value facts on this

to show that our notion of rationality is, at bottom, just one part
conception. But the relation between rational acceptability and of our conception of human flourishing, our idea of the good.
truth is a relation between two distinct notions. A statement can Truth is deeply dependent on what have been recendy called
be rationally acceptable at a time but not true; and this realist 'values' (Chapter 6). And what we said above about rationality

intuition will be preserved in my account. and history also applies to value and history; there is no given,
I do not believe, however, that rationality is defined by a set of
ahistorical, set of 'moral principles' which define once and for

unchanging 'canons' or 'principles'; methodological principles all what human flourishing consists in; but that doesn't mean
are connected with our view of the world, including our view of that it's all merely cultural and relative. Since the current state in

ourselves as part of the world, and change with time. the theory of truth - the current dichotomy between copy theo-
Thus I
agree with the subjectivist philosophers that there ries of truth and subjective accounts of truth - is at least partly
is no fixed,
ahistorical organon which defines what it is to be rational; but I
responsible, in my view, for the notorious 'fact/value' dichot-

don't conclude from the fact that our conceptions of reason omy, it is only by going to a very deep level and correcting our
evolve in history, that reason itself can be (or evolve into) any- accounts of truth and rationality themselves that we can get
thing, nor do I end up insome fancy mixture of cultural relativ- beyond the fact/value dichotomy. (A dichotomy which, as it is
ism and 'structuralism' like the French philosophers. The dichot-
conventionally understood, virtually commits one to some sort

omy: either ahistorical unchanging canons of rationality or of relativism.) The current views of truth are alienated views;
cultural relativism is a dichotomy that I regard as outdated. they cause one to lose one part or another of one's self and the
Another feature of the view is that rationality is not restricted world, to see the world as simply consisting of elementary par-
to laboratory science, nor different in a fundamental way in ticlesswerving in the void (the 'physicalist' view, which sees the
laboratory science and outside of The conception that
converging to the One True Theory), or
scientific description as
it. it is

seems to me a hangover from positivism; from the idea that the to see the world as simply consisting of 'actual and possible
 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 moit, alors tout est permis".
*
(reflexions . .

sur la philosophic du langage)', Critique, 1980. Brains in a vat

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

including Ned Block, David Helman, and Justin Leiber, and the line in the sand. By pure chance the line that it
traces curves and
like a rec-
students in my various lectures and seminars at Harvard. Several recrosses itself in such a way that it ends up looking
ognizable caricature of Winston Churchill. Has the ant
chapters were read as lectures in Lima in the spring of 1980 (a traced a

trip made possible by a grant from the Fulbright Commission), picture of Winston Churchill, a picture that depicts Churchill?
and Chapter 2 was actually finished during my Lima stay. I ben- Most people would say, on a little reflection, that it has not.
efited in this period from discussions with Leopoldo Chiappo, The ant, after all, has never seen Churchill, or even a picture of
It sim-
Alberto Cordero Lecca, Henriques Fernandez, Francisco Miro Churchill, and it had no intention of depicting Churchill.
line that
Quesada, and Jorge Secada. The entire book (in an earlier ver- ply traced a line (and even that was unintentional), a
sion) was read as lectures at the University of Frankfurt in the we can 'see as' a picture of Churchill.
summer of 1980, and I am grateful to my colleagues there (espe- We express this by saying that the line is not 'in itself' a
can
Similarity
cially Wilhelm Essler and Rainer Trapp), to my very stimulating representation 1 of anything rather than anything else.
group of students, and my other friends in Germany (especially (of a certain very complicated sort) to the features of Winston
Dieter Henrich, Manon Fassbinder, and Wolfgang Stegmuller) Churchill is not sufficient to make something represent or refer
for encouragement and stimulating discussions. Nor necessary: in our community the printed
to Churchill. is it

All of my colleagues in the Harvard Philosophy Department shape 'Winston Churchill', the spoken words 'Winston Chur-
deserve to be singled out for individual thanks. In recent years chill', and many other things are used to represent Churchill
Nelson Goodman and I have detected a convergence in our (though not pictorially), while not having the sort of similarity
views, and while the first draft of the present book was written
refer to a
1
In this bookthe terms 'representation* and 'reference' always
before had the opportunity to see his Ways of Worldmaking,
I
relation between a word (or other sort of sign, symbol, or
just an 'object
reading it and discussing these issues with him has been of great representation) and something that actually exists (i.e. not
to what does
value at a number of stages. of thought'). There is a sense of 'refer' in which I can 'refer'
not exist; thisnot the sense in which 'refer' is used here. An older
is
I am also grateful to Jeremy Mynott for encouragement and word for what I call 'representation' or 'reference' is denotation.
advice in his capacity as editor. Secondly, I follow the custom of modern logiciansand use 'exist' to
mean 'exist in the past, present, or future'. Thus Winston Churchill
*To study "The Appraisal of Scientific Theories: Historical versus Formal and we can 'refer to' or
'exists', 'represent' Winston Churchill, even
Methodological Approaches'; Agreement No. SOC78-04276. though he is no longer alive.
 2 Brains in a vat Brains in a vat

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

larity isnot necessary or sufficient to make something represent Magical theories of reference
something else, how can anything be necessary or sufficient for
We saw that the ant's 'picture' has no necessary connection with
this purpose? How on earth can one thing
represent (or 'stand Winston Churchill. The mere fact that the 'picture' bears a
for', etc.) a different thing?
'resemblance' to Churchill does not make it into a real picture,
The answer may seem easy. Suppose the ant had seen Winston
nor does it make it a representation of Churchill. Unless the ant
Churchill, and suppose that it had the intelligence and
skill to is an intelligent ant (which it isn't) and knows about Churchill
draw a picture of him. Suppose it produced the caricature inten-
(which it doesn't), the curve it traced is not a picture or even a
tionally. Then the line would have represented
Churchill. representation of anything. Some primitive people believe that
On the other hand, suppose the line had the shape WINSTON representations (in particular, names) have a necessary
some
CHURCHILL. And suppose this was just accident (ignoring the know the 'true name' of
connection with their bearers; that to
improbability involved). Then the 'printed shape' WINSTON someone or something gives one power over it. This power
CHURCHILL would not have represented Churchill, although
comes from the magical connection between the name and the
that printed shape does represent Churchill
when it occurs in bearer of the name; once one realizes that a name only has a
almost any book today.
contextual, contingent, conventional connection with its bearer,
So it may seem that what necessary for representation, or
is
it is hard to see why knowledge of the name should have any
what is mainly necessary for representation, is intention.
mystical significance.
But to have the intention that anything, even private lan-
What is important to realize what goes for physical
is that
guage (even the words 'Winston Churchill' spoken in my
mind pictures also goes for mental images,and for mental representa-
and not out loud), should represent Churchill, I must
have been tions in general; mental representations no more have a neces-
able to think about Churchill in the first place. If
lines in the sary connection with what they represent than physical represen-
sand, noises, etc., cannot 'in themselves' represent
anything, then tations do. The contrary supposition is a survival of magical
how is it that thought forms can 'in themselves' represent any-
thinking.
thing? Or can they? How can thought reach out and 'grasp' Perhaps the point is easiest to grasp in the case of mental
what is external?
images. (Perhaps the first philosopher to grasp the enormous sig-
Some philosophers have, in the past, leaped from this sort of
nificance of this point, even if he was not the first to actually
consideration to what they take to be a proof that the
mind is make it, was Wittgenstein.) Suppose there is a planet somewhere
essentially non-physical in nature. The argument is
simple; what on which human beings have evolved (or been deposited by alien
we said about the ant's curve applies to any physical object. No
spacemen, or what have you). Suppose these humans, although
physical object can, in itself, refer to one thing
rather than to otherwise like us, have never seen trees. Suppose they have never
another; nevertheless, thoughts in the mind obviously
do suc- imagined on their planet only
trees (perhaps vegetable life exists
ceed in referring to one thing rather than another. So
thoughts in the form of molds). Suppose one day a picture of a tree is
(and hence the mind) are of an essentially different nature
than accidentally dropped on their planet by a spaceship which passes
physical objects. Thoughts have the characteristic of
intention- on without having other contact with them. Imagine them puz-
ality- they can refer to something else; nothing
physical has zling over the picture. What in the world is this? All sorts of
'intentionality', save as that intentionality is
derivative from speculations occur to them: a building, a canopy, even an animal
some employment of that physical thing by a mind. Or so it is
of some kind. But suppose they never come close to the truth.
claimed. This too quick; just postulating mysterious powers of
is
For us the picture is a representation of a tree. For these
mind solves nothing. But the problem is very real. How is inten-
humans the picture only represents a strange object, nature and
tionality, reference, possible?
function unknown. Suppose one of them has a mental image
 Brains in a vat Brains in a vat 5

which is exactly like one of my mental images of a tree as a result ously has appropriate mental images, but neither understands
of having seen the picture. His mental image is not a represen- the words nor knows what a tree is. We can even imagine that
tation of a tree. It is only a representation of the strange object the mental images were caused by paint-spills (although the per-
(whatever it is) that the mysterious picture represents. son has been hypnotized to think that they are images of some-
Still, someone might argue that the mental image is in fact a thing appropriate to his thought - only, if he were asked, he
representation of a tree, if only because the picture which caused wouldn't be able to say of what). And we can imagine that the
this mental image was itself a representation of a tree to begin language the person is thinking in is one neither the hypnotist
with. There is a causal chain from actual trees to the mental nor the person hypnotized has ever heard of — perhaps it is just
image even if it is a very strange one. coincidence that these 'nonsense sentences', as the hypnotist sup-
But even this causal chain can be imagined absent. Suppose poses them to be, are a description of trees in Japanese. In short,
the 'picture of the tree' that the spaceship dropped was not really everything passing before the person's mind might be qualita-
a picture of a tree, but the accidental result of some spilled with what was passing through the mind of a
tively identical

[Link] if it looked exactly like a picture of a tree, it was, in Japanese speaker who was really thinking about trees — but
truth,no more a picture of a tree than the ant's 'caricature' of none of it would refer to trees.

Churchill was a picture of Churchill. We can even imagine that All of this is really impossible, of course, in the way that it is

the spaceship which dropped the 'picture' came from a planet really impossible that monkeys should by chance type out a copy
which knew nothing of trees. Then the humans would still have of Hamlet. That is to say that the probabilities against it are so
mental images qualitatively identical with my image of a tree, high as to mean it will never really happen (we think). But is is
but they would not be images which represented a tree any more not logically impossible, or even physically impossible. It could
than anything else. happen (compatibly with physical law and, perhaps, compatibly
The same thing is true of words. A discourse on paper might with actual conditions in the universe, if there are lots of intelli-

seem to be a perfect description of trees, but if it was produced gent beings on other planets). And if it did happen, it would be
by monkeys randomly hitting keys on a typewriter for millions a striking demonstration of an important conceptual truth; that
of years, then the words do not refer to anything. If there were even a large and complex system of representations, both verbal
a person who memorized those words and said them in his mind and visual, still does not have an intrinsic, built-in, magical con-
without understanding them, then they would not refer to any- nection with what it represents - a connection independent of
thing when thought in the mind, either. how it was caused and what the dispositions of the speaker or
Imagine the person who is saying those words in his mind has thinker are. And
this is true whether the system of representa-
been hypnotized. Suppose the words are in Japanese, and the per- tions (words and images, in the case of the example) is physically
son has been told that he understands Japanese. Suppose that as realized - the words are written or spoken, and the pictures are
he thinks those words he has a 'feeling of understanding'. physical pictures - or only realized in the mind. Thought words
(Although if someone broke into his train of thought and asked and mental pictures do not intrinsically represent what they are
him what the words he was thinking meant, he would discover about.
he couldn't say.) Perhaps the illusion would be so perfect that
the person could even fool a Japanese telepath! But if he
The case of the brains in a vat
couldn't use the words answer questions
in the right contexts,
about what he 'thought', etc., then he didn't understand them. Here is a science fiction possibility discussed by philosophers:
By combining these science fiction stories I have been telling, imagine that a human being (you can imagine this to be yourself)
we can contrive a case in which someone thinks words which are has been subjected to an operation by an evil scientist. The per-
in fact a description of trees in some language and simultane- son's brain (your brain) has been removed from the body and
 Brains in a vat Brains in a vat

placed in a vat of nutrients which keeps the brain alive. The reach your ears - for you don't have (real) ears, nor do I have a
nerve endings have been connected to a super-scientific com- real mouth and tongue. Rather, when I produce my words, what
puter which causes the person whose brain it is to have the illu- happens is that the efferent impulses travel from my brain to the
sion that everything is There seem to be peo-
perfectly normal. computer, which both causes me to 'hear' my own voice uttering
ple, objects, the sky, etc; but really all the person (you) is those words and 'feel' my tongue moving, etc., and causes you
experiencing is the result of electronic impulses travelling from to 'hear' my words, 'see' me speaking, etc. In this case, we are,
the computer to the nerve endings. The computer is so clever in a sense, actually in communication. I am not mistaken about
that if the person tries to raise his hand, the feedback from the your real existence (only about the existence of your body and
computer will cause him to 'see' and 'feel' the hand being raised. the 'external world', apart from brains). From a certain point of
Moreover, by varying the program, the evil scientist can cause view, it doesn't even matter that 'the whole world' is a collective
the victim to 'experience' (or hallucinate) any situation or envi- hallucination; for you do, after all, really hear my words when
ronment the evil scientist wishes. He can also obliterate the I speak to you, even ifthe mechanism isn't what we suppose it
memory of the brain operation, so that the victim will seem to to be. (Of course, if we were two lovers making love, rather than
himself to have always been in this environment. It can even justtwo people carrying on a conversation, then the suggestion
seem to the victim that he is sitting and reading these very words that it was just two brains in a vat might be disturbing.)
about the amusing but quite absurd supposition that there is an I want now to ask a question which will seem very silly and

evil scientist who removes people's brains from their bodies and obvious (at least to some people, including some very sophisti-
places them in a vat of nutrients which keep the brains alive. The cated philosophers), but which will take us to real philosophical
nerve endings are supposed to be connected to a super-scientific depths rather quickly. Suppose this whole story were actually
computer which causes the person whose brain it is to have the true. Could we, if we were brains in a vat in this way, say or
illusion that . . .
think that we were?
When this sort of possibility is mentioned in a lecture on the I am going to argue that the answer is 'No, we couldn't.' In
Theory of Knowledge, the purpose, of course, is to raise the clas- fact, I am going to argue that the supposition that we are
sical problem of scepticism with respect to the external world in actually brains in a vat, although it violates no physical law, and
a modern way. {How do you know you aren't in this predica- is perfectly consistent with everything we have experienced, can-
ment?) But this predicament is also a useful device for raising not possibly be true. It cannot possibly be true, because it is, in
issues about the mind/world relationship. a certain way, self-refuting.
Instead of having just one brain in a vat, we could imagine The argument I am going to present is an unusual one, and it
that all human beings (perhaps all sentient beings) are brains in took me several years to convince myself that it is really right.
a vat (or nervous systems in a vat in case some beings with just But it is a correct argument. What makes it seem so strange is
a minimal nervous system already count as 'sentient'). Of course, that it is connected with some of the very deepest issues in phi-
the evil scientist to be outside - or would he? Per-
would have losophy. (It first occurred to me when I was thinking about a
haps there is no perhaps (though this is absurd) the
evil scientist, theorem in modern logic, the 'Skolem-Lowenheim Theorem',
universe just happens to consist of automatic machinery tending and I suddenly saw a connection between this theorem and some
a vat full of brains and nervous systems. arguments in Wittgenstein's Philosophical Investigations,)
This time let us suppose that the automatic machinery is pro- A 'self-refuting supposition' is one whose truth implies its own
grammed to give us all a collective hallucination, rather than a [Link] example, consider the thesis that all general state-
number of separate unrelated hallucinations. Thus, when I seem ments are false. This is a general statement. So if it is true, then
to myself to be talking to you, you seem to yourself to be hearing it must be false. Hence, it is false. Sometimes a thesis is called
my words. Of course, it is not the case thatmy words actually 'self-refuting' if it is the supposition that the thesis is entertained
 Brains in a vat
8 Brains in a vat

or enunciated that implies its falsity. For example, 'I do not exist' not tell which is the computer and which is the human being,
s y then (assume the test to be repeated a sufficient number of times
is self-refuting if thought by me (forany me ). So one can be
with different interlocutors) the computer is conscious. In short,
certain that one oneself exists, if one thinks about it (as Des-
a computing machine is conscious if it can pass the 'Turing Test'.
cartes argued).
(The conversations are not to be carried on face to face, of
What I shall show is that the supposition that we are brains in
course, since the interlocutor is not to know the visual appear-
a vat has just this property. If we can consider whether it is true
show). Hence not true. ance of either of his two conversational partners. Nor is voice to
or false, then it is not true (I shall it is

why seems so be used, since the mechanical voice might simply sound different
Before I give the argument, let us consider it
from a human voice. Imagine, rather, that the conversations are
strange that such an argument can be given (at least to philoso-
who subscribe to a 'copy' conception of truth). We con- all carried on via electric typewriter. The interlocutor types in
phers
ceded that it is compatible with physical law that there should
his statements, and the two partners - the
questions, etc.,

As machine and the person - respond via the electric keyboard.

be a world in which all sentient beings are brains in a vat.
Also, the machine may lie - asked 'Are you a machine', it might
philosophers say, there is a 'possible world' in which all sentient
makes reply, 'No, I'm an assistant in the lab here.')
beings are brains in a vat. (This 'possible world' talk it
The idea that this test is really a definitive test of consciousness
sound as if there is a place where any absurd supposition is true,

which is why it can be very misleading in philosophy.) The

has been criticized by a number of authors (who are by no means

humans in that possible world have exactly the same experiences

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
that we do. They think the same thoughts we do (at least, the
general idea of the Turing test, the general idea of a dialogic test
same words, images, thought-forms, etc., go through their

minds). Yet, I am an argument we can give

claiming that there is
of competence, for a different purpose, the purpose of exploring
the notion of reference.
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 Imagine a situation in which the problem is not to determine

brains in a vat give it too?

if the partner is really a person or a machine, but is rather to

The answer is going to be (basically) this: although the people determine if the partner uses the words to refer as we do. The
world can think and any words we can obvious test is, again, to carry on a conversation, and, if no
in that possible 'say'
think and say, they cannot (I claim) refer to what we can refer problems arise, if the partner 'passes' in the sense of being indis-
tinguishable from someone who is certified in advance to be
to. In particular, they cannot think or say that they are brains in
speaking the same language, referring to the usual sorts of
a vat [even by thinking 'we are brains in a vaf).
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,
Turing's test
that is, to determine the existence of (shared) reference, I shall
Suppose someone succeeds in inventing a computer which can refer to the test as the Turing Test for Reference. And, just as
actually carry on an intelligent conversation with one (on as philosophers have discussed the question whether the original
many subjects as an intelligent person might). How can one Turing test is a definitive test for consciousness, i.e. the question
decide if the computer is 'conscious'? of whether a machine which 'passes' the test not just once but
The British logician Alan Turing proposed the following test: 2 regularly is necessarily conscious, so, in the same way, I wish to
let someone carry on a conversation with the computer and a discuss the question of whether the Turing Test for Reference
conversation with a person whom he does not know. If he can- just suggested is a definitive test for shared reference.
The answer will turn out to be 'No'. The Turing Test for Ref-
2
A. Turing, 'Computing Machinery and Intelligence', Mind (1950),
M.
reprinted in A. R. Anderson (ed.), Minds and Machines.
erence is not definitive. It is certainly an excellent test in practice;
10 Brains in a vat Brains in a vat 11

but it is not logically impossible (though it is certainly highly go on 'fooling* each other forever, even
if the rest of the world
improbable) that someone could pass the Turing Test for Refer- disappeared! There no more reason to regard the machine's
is
ence and not be referring to anything. It follows from this, as we talk of apples as referring to real world apples than there is to
shall see, that we can
extend our observation that words (and regard the ant's 'drawing' as referring to Winston Churchill.
whole texts and discourses) do not have a necessary connection What produces the illusion of reference, meaning, intelligence,
to their [Link] if we consider not words by themselves etc., here is the fact that there is a convention of representation
but rules deciding what words may appropriately be produced which we have under which the machine's discourse refers to
in certain contexts — even if we consider, in computer jargon, apples, steeples, New England, etc. Similarly, there is the illusion
programs for using words — unless those programs themselves that the ant has caricatured Churchill, for the same reason. But
refer to something extra-linguistic there is still no determinate we are able to perceive, handle, deal with apples and fields. Our
reference that those words possess. This will be a crucial step in talk of apples and fields is intimately connected with our non-
the process of reaching the conclusion that the Brain-in-a-Vat verbal transactions with apples and fields. There are 'language
Worlders cannot refer to anything external at all (and hence can- entry rules' which take us from experiences of apples to such
not say that they are Brain-in-a-Vat Worlders). utterances as 'I see an apple', and 'language exit rules' which
Suppose, for example, that I am in the Turing situation (play- take us from decisions expressed in linguistic form ('I am going
ing the 'Imitation Game', in Turing's terminology) and my part- to buy some apples') to actions other than speaking. Lacking
ner is actually a machine. Suppose this machine is able to win either language entry rules or language exit rules, there is no
the game ('passes' the test). Imagine the machine to be pro- reason to regard the conversation of the machine (or of the two
grammed to produce beautiful responses in English to state- machines, in the case we envisaged of two machines playing the
ments, questions, remarks, etc. in English, but that it has no Imitation Game with each other) as more than syntactic play.
sense organs (other than the hookup to my electric typewriter), Syntactic play that resembles intelligent discourse, to be sure;
and no motor organs (other than the electric typewriter). (As far but only as (and no more than) the ant's curve resembles a biting
as I can make out, Turing does not assume that the possession caricature.
of either sense organs or motor organs is necessary for con- In the case of the ant, we could have argued that the ant would
sciousness or intelligence.) Assume that not only does the have drawn the same curve even if Winston Churchill had never
machine lack electronic eyes and ears, etc., but that there are no existed. In the case of the machine, we cannot quite make the
provisions in the machine's program, the program for playing parallel argument; if apples, trees, steeples and fields had not
the Imitation Game, for incorporating inputs from such sense existed, then, presumably, the programmers would not have
organs, or for controlling a body. What should we say about produced that same program. Although the machine does not
such a machine? perceive apples, fields, or steeples, its creator-designers did.
To me, it seems evident that we cannot and should not attrib- There is some causal connection between the machine and the
ute reference to such a device. It is true that the machine can real world apples, etc., via the perceptual experience and knowl-
discourse beautifully about, say, the scenery in New England. edge of the creator-designers. But such a weak connection can
But it could not recognize an apple tree or an apple, a mountain hardly suffice for reference. Not only is it logically possible,
or a cow, a field or a steeple, if it were in front of one. though fantastically improbable, that the same machine could
What we have is a device for producing sentences in response have existed even if apples, fields, and steeples had not existed;
to sentences. But none of these sentences is at all connected to more important, the machine is utterly insensitive to the contin-
the real world. If one coupled two of these machines and let ued existence of apples, fields, steeples, etc. Even if all these
them play the Imitation Game with each other, then they would things ceased to exist, the machine would still discourse just as
12 Brains in a vat Brains in a vat 13

happily in the same way. That is why the machine cannot be our world; but we have already seen (the ant again!) that quali-
regarded as referring at all. tative similarity to something which represents an object (Win-
The point that is relevant for our discussion is that there is ston Churchill or a tree) does not make a thing a representation
nothing in Turing's Test to rule out a machine which is pro- all by itself. In short, the brains in a vat are not thinking about

grammed to do nothing but play the Imitation Game, and that a real trees when they think 'there is a tree in front of me' because
machine which can do nothing but play the Imitation Game is there is nothing by virtue of which their thought 'tree' represents
clearly not referring any more than a record player is. actual trees.
If this seems hasty, reflect on the following: we have seen that
the words do not necessarily refer to trees even
they are if
Brains in a vat (again)
arranged in a sequence which is identical with a discourse which
Let us compare the hypothetical 'brains in a vat' with the (were it to occur in one of our minds) would unquestionably be
machines just described. There are obviously important differ- about trees in the actual world. Nor does the 'program', in the
ences. The brains in a vat do not have sense organs, but they do sense of the rules, practices, dispositions of the brains to verbal
have provision for sense organs; that is, there are afferent nerve behavior, necessarily refer to trees or bring about reference to
endings, there are inputs from these afferent nerve endings, and treesthrough the connections it establishes between words and
these inputs figure in the 'program' of the brains in the vat just words, or linguistic cues and linguistic responses. If these brains
as they do in the program of our brains. The brains in a vat are think about, refer to, represent trees (real trees, outside the vat),
brains; moreover, they art functioning brains, and they function then must be because of the way the 'program' connects the
it

by the same rules as brains do in the actual world. For these system of language to non-verbal input and outputs. There are
reasons, it would seem absurd to deny consciousness or intelli- indeed such non-verbal inputs and outputs in the Brain-in-a-Vat
gence to them. But the fact that they are conscious and intelligent world (those efferent and afferent nerve endings again!), but we
does not mean that their words refer to what our words refer. also saw that the 'sense-data' produced by the automatic
The question we are interested in is this: do their verbalizations machinery do not represent trees (or anything external) even
containing, say, the word 'tree' actually refer to trees? More gen- when they resemble our tree-images exactly. Just as a splash of
erally:can they refer to external objects at all? (As opposed to, paint might resemble a tree picture without being a tree picture,
for example, objects in the image produced by the automatic so, we saw, a 'sense datum' might be qualitatively identical with
machinery.) an 'image of a tree' without being an image of a tree. How can
To fix our ideas, let us specify that the automatic machinery is the fact that, in the case of the brains in a vat, the language is

supposed to have come into existence by some kind of cosmic connected by the program with sensory inputs which do not
chance or coincidence (or, perhaps, to have always existed). In intrinsically or extrinsically represent trees (or anything exter-
this hypothetical world, the automatic machinery itself is sup- nal) possibly bring it about that the whole system of representa-
posed to have no intelligent creator-designers. In fact, as we said tions, the language-in-use, does refer to or represent trees or any-
at the beginning of this chapter, we may imagine that all sentient thing external?
beings (however minimal their sentience) are inside the vat. The answer is that it cannot. The whole system of sense-data,
This assumption does not help. For there is no connection motor signals to the efferent endings, and verbally or concep-
between the word used by these brains and actual trees.
'tree' as tually mediated thought connected by 'language entry rules' to
They would still use the word 'tree'
just as they do, think just the the sense-data (or whatever) as inputs and by 'language exit
thoughts they do, have just the images they have, even if there rules' to the motor signals as outputs, has no more connection to
were no actual trees. Their images, words, etc., are qualitatively trees than the ant's curve has to Winston Churchill. Once we see
identical with images, words, etc., which do represent trees in that the qualitative similarity (amounting, if you like, to quali-
14 Brains in a vat Brains in a vat 15

between the thoughts of the brains in a vat and

tative identity) the vat they are in; but this connection obtains between the use
the thoughts ofsomeone in the actual world by no means implies of every word in vat-English and that one particular vat; it is not
sameness of reference, it is not hard to see that there is no basis a special connection between the use of the particular word 'vat'
at all for regarding the brain in a vat as referring to external and vats). Similarly, 'nutrient fluid' refers to a liquid in the image
things. 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
The premisses of the argument
we now mean by 'we are brains in a vat' is that we are brains in
I have now given the argument promised to show that the brains a vat in the image or something of that kind (if we mean any-
in a vat cannot think or say that they are brains in a vat. It thing at all). But part of the hypothesis that we are brains in a
remains only to make it explicit and to examine its structure. vat is that we aren't brains in a vat in the image (i.e. what we are
By what was just said, when the brain in a vat (in the world 'hallucinating' isn't that we are brains in a vat). So, if we are
where every sentient being is and always was a brain in a vat) brains in a vat, then the sentence 'We are brains in a vat' says
thinks 'There is a tree in front of me', his thought does not refer something false (if it says anything). In short, if we are brains in
to actual trees. On some theories that we shall discuss it might a vat, then 'We are brains in a vat' is false. So it is (necessarily)
refer to trees in the image, or to the electronic impulses that false.

cause tree experiences, or to the features of the program that are The supposition that such a possibility makes sense arises
responsible for those electronic impulses. These theories are not from a combination of two errors: (1) taking physical possibility
ruled out by what was just said, for there is a close causal con- too seriously; and (2) unconsciously operating with a magical
nection between the use of the word 'tree' in vat-English and the theory of reference, a theory on which certain mental represen-
presence of trees in the image, the presence of electronic impulses tations necessarily refer to certain external things and kinds of
of a certain kind, and the presence of certain features in the things.
machine's program. On these theories the brain is right, not There is a 'physically possible world' in which we are brains
wrong in thinking 'There is a tree in front of me.' Given what in a vat - what
does mean except that there is a description
this
'tree' refers to in vat-English and what 'in front of refers of such a state of affairs which is compatible with the laws of
to, assuming one of these theories is correct, then the truth- physics? Just as there is a tendency in our culture (and has been
conditions for 'There is a tree in front of me' when it occurs in since the seventeenth century) to take physics as our metaphys-
vat-English are simply that a tree in the image be 'in front of the ics, that is, to view the exact sciences as the long-sought descrip-
'me' in question - in the image - or, perhaps, that the kind of tion of the 'true and ultimate furniture of the universe', so there
electronic impulse that normally produces this experience be is, immediate consequence, a tendency to take 'physical
as an
coming from the automatic machinery, or, perhaps, that the fea- possibility' as the very touchstone of what might really actually
ture of the machinery that is supposed to produce the 'tree in be the case. Truth is physical truth; possibility physical possibil-
front of one' experience be operating. And these truth- ity; and necessity physical necessity, on such a view. But we have

conditions are certainly fulfilled. just seen, if only in the case of a very contrived example so far,
By the same argument, 'vat' refers to vats in the image in vat- that this view wrong. The existence of a 'physically possible
is

English, or something related (electronic impulses or program world' in which we are brains in a vat (and always were and will
features), but certainly not to real vats, since the use of 'vat' in be) does not mean that we might really, actually, possibly be
vat-English has no causal connection to real vats (apart from the brains in a vat. What rules out this possibility is not physics but
connection that the brains in a vat wouldn't be able to use the philosophy.
word *vat\ if it were not for the presence of one particular vat - Some philosophers, eager both to assert and minimize the
16 Brains in a vat Brains in a vat 17

claims of their profession at the same time (the typical state of of which they can be described. But why should we accept these
mind of Anglo-American philosophy in the twentieth century), premisses? Since these constitute the broad framework within

would say: 'Sure. You have shown that some things that seem to which I am arguing, it is time to examine them more closely.
be physical possibilities are really conceptual impossibilities.
What's so surprising about that?' The reasons for denying necessary connections between
Well, to be sure, my argument can be described as a 'concep-
representations and their referents
tual' one. But to describe philosophical activity as the search for
about the I mentioned earlier that some philosophers (most
'conceptual' truths makes it all sound like inquiry famously,
what we have been Brentano) have ascribed to the mind a power, 'intentionality',
meaning of words. And that is not at all
which precisely enables it to refer. Evidently, I have rejected this
engaging in.

What we have been doing is considering the preconditions for asno solution. But what gives me this right? Have I, perhaps,
We have investi- been too hasty?
thinking about, representing, referring to, etc.
These philosophers did not claim that we can think about
gated these preconditions not by investigating the meaning of
external things or properties without using representations at all.
these words and phrases (as a linguist might, for example) but
by reasoning a priori. Not in the old 'absolute' sense (since we And the argument I gave above comparing visual sense data to
of reference are a priori the ant's 'picture' (the argument via the science fiction story
don't claim that magical theories
inquiring into what is reasonably about the 'picture' of a tree that came from a paint-splash and
wrong), but in the sense of
general premisses, or making certain that gave rise to sense data qualitatively similar to our 'visual
possible assuming certain
Such a procedure is neither images of but unaccompanied by any concept of a tree)
trees',
very broad theoretical assumptions.
'empirical' nor quite 'a priori', but has elements of both ways of
would be accepted as showing that images do not necessarily
If there are mental representations that necessarily refer
investigating. In spite of the fallibility of my procedure, and its
refer. (to
external things) they must be of the nature of concepts and not
dependence upon assumptions which might be described as

the assumption that the mind has no access to of the nature of images. But what are concepts?
'empirical' (e.g.
external things or properties apart from that provided by the
When we introspect we do not perceive 'concepts' flowing
senses), my procedure has a close relation to what Kant called a
through our minds as such. Stop the stream of thought when or
'transcendental' investigation; for an investigation, I repeat,
it is
where we will, what we catch are words, images, sensations,

of the preconditions of reference and hence of thought - precon-

feelings. When I speak my thoughts out loud I do not think them
ditions built in to the nature of our minds themselves, though
twice. I hear my words as you do. To be sure it feels different to
not (as Kant hoped) wholly independent of empirical assump- me when I utter words that I believe and when I utter words I

tions.
do not believe (but sometimes, when I am nervous, or in front of
One of the premisses of the argument is obvious: that magical a hostile audience, it feels as if I am lying when I know I am
theories of reference are wrong, wrong for mental representa-
telling the truth); and it feels different when I utter words I
tions and not only for physical ones. The other premiss is that
understand and when words I do not understand. But I
I utter

one cannot refer to certain kinds of things, e.g. trees, if one has can imagine without someone thinking just these
difficulty

no causal interaction at all with them, 3 or with things in terms words (in the sense of saying them in his mind) and having just
the feeling of understanding, asserting, etc., that I do, and real-
3 have causal connection with, say, trees in the izing a minute later (or on being awakened by a hypnotist) that
If the Brains in a Vat will
future, then perhaps they can now refer to trees by the description 'the he did not understand what had just passed through his mind at
things I will refer to as "trees" at such-and-such a future time*. But we all, that he did not even understand the language
these words are
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.
in. I don't claim that this is very likely; I simply mean that there
18 Brains in a vat Brains in a vat 19

isnothing at all unimaginable about this. And what this shows when used on Twin Earth and when
represents a different liquid

isnot that concepts are words (or images, sensations, etc.), but used on Earth, Contrary to a doctrine that has been with us
etc.)

that to attribute a 'concept' or a 'thought' to someone is quite since the seventeenth century, meanings just aren't in the head.

different from attributing any mental 'presentation', any intro- We have seen that possessing a concept is not a matter of pos-
spectible entity or event, to him. Concepts are not mental presen- - or even images, 'visual' or 'acous-
sessing images (say, of trees

tations that intrinsically refer to external objects for the very of sentences, or whole discourses, for that matter) since one
tic',

decisive reason that they are not mental presentations at all. could possess any system of images you please and not possess
Concepts are signs used in a certain way; the signs may be public the ability to use the sentences in situationally appropriate ways
or private, mental entities or physical entities, but even when the (considering both linguistic factors - what has been said
signs are 'mental' and 'private', the sign itself apart from its use before -and
non-linguistic factors as determining 'situational

is not the concept. And signs do not themselves intrinsically appropriateness'). A

man may have all the images you please,
refer.
and be completely at a loss when one says to him 'point to
still

We can see this by performing a very simple thought experi- a tree', even if a lot of trees are present. He may even have the

ment. Suppose you are like me and cannot tell an elm tree from image of what he is supposed to do, and still not know what he
a beech tree. We still say that the reference of 'elm' in my speech
is supposed to do. For the image, if not accompanied by the
is the same as the reference of 'elm' in anyone else's, viz. elm ability to act in a certainway, is just a picture, and acting in
trees, and that the set of all beech trees is the extension of 'beech' accordance with a picture is itself an ability that one may or may
(i.e. the set of things the word 'beech' is truly predicated of) both not have. (The man might picture himself pointing to a tree, but
just for the sake of contemplating something logically possible;
inyour speech and my speech. Is it really credible that the differ-
himself pointing to a tree after someone has produced the - to
ence between what 'elm' refers to and what 'beech' refers to is
brought about by a difference in our concepts? My concept of
him meaningless - sequence of sounds 'please point to a tree'.)
an elm tree is exactly the same as my concept of a beech tree (I
He would still not know that he was supposed to point to a tree,
blush to confess). (This shows that the determination of refer-
and he would still not understand 'point to a tree'.
I have considered the ability to use certain sentences to
ence is social and not individual, by the way; you and I both be the
defer to experts who can tell elms from beeches.) If someone criterion for possessing a full-blown concept, but this could eas-

heroically attempts to maintain that the difference between the ily be liberalized. Wecould allow symbolism consisting of ele-
reference of 'elm' and the reference of 'beech' in my speech is ments which are not words in a natural language, for example,
explained by a difference in my psychological state, then let him and we could allow such mental phenomena as images and other
imagine a Twin Earth where the words are switched. Twin Earth types of internal events. What is essential
is that these should

is very much like Earth; in fact, apart from the fact that 'elm'
have the same complexity, be combined with each
ability to
other, etc.,
and 'beech' are interchanged, the reader can suppose Twin Earth as sentences in a natural language. For, although a

is exactly like Earth. Suppose I have a Doppelganger on Twin

- say, a blue flash - might serve a partic-
particular presentation

Earth who is molecule for molecule identical with me (in the ular mathematician as the inner expression of the whole proof

sense in which two neckties can be 'identical'). If you are a dual- of the Prime Number Theorem,still there would be no tempta-

ist, then suppose my Doppelganger thinks the same verbalized

tion to say this (and would be false to say this) if that mathe-
it

thoughts I do, has the same sense data, the same dispositions, matician could not unpack his 'blue flash' into separate steps and
logical connections. But, no matter what sort of inner phenom-
etc. It is absurd to think his psychological state is one bit differ-

ent from mine: yet his word 'elm' represents beeches, and my
ena we allow as possible expressions of thought, arguments
word 'elm' represents elms. (Similarly, if the 'water' on Twin exactly similar to the foregoing will show that it is not the phe-

Earth is a different liquid - say, XYZ and not H 2 - then 'water' nomena themselves that constitute understanding, but rather the
20 Brains in a vat Brains in a vat 21

ability of the thinker to employ these phenomena, to produce identical with mental objects of any kind. For, assuming that
the right phenomena in the right circumstances. by a mental object we mean something introspectible, we have
The foregoing is a very abbreviated version of Wittgenstein's just seen that whatever it is, it may be absent in a man who

argument in Philosophical Investigations, If it is correct, then the does understand the appropriate word (and hence has the full
attempt to understand thought by what is called 'phenomeno- blown concept), and present in a man who does not have the
logicaP investigation fundamentally misguided; for what the
is concept at all.
phenomenologists fail to see is that what they are describing is Coming back now to our criticism of magical theories of ref-
the inner expression of thought, but that the understanding of erence (a topic which also concerned Wittgenstein), we see that,
that expression - one's understanding of one's own thoughts - on the one hand, those 'mental objects' we can introspectively
isnot an occurrence but an ability. Our example of a man pre- detect - words, images, feelings, etc. - do not intrinsically refer

tending to think in Japanese (and deceiving a Japanese telepath) any more than the ant's picture does (and for the same reasons),
already shows the futility of a phenomenological approach to while the attempts to postulate special mental objects, 'con-
the problem oi understanding. For even if there is some intros- cepts', which do have a necessary connection with their refer-

pectible quality which is present when and only when one really ents, and which only trained phenomenologists can detect, com-

understands (this seems false on introspection, in fact), still that mit a logical blunder; for concepts are (at least in part) abilities
quality only correlated with understanding, and it is still pos-
is and not occurrences. The doctrine that there are mental presen-
sible that the man fooling the Japanese telepath have that quality tations which necessarily is not only bad
refer to external things

too and still not understand a word of Japanese. natural science; it is also bad phenomenology and conceptual
On the other hand, consider the perfectly possible man who confusion.
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
 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 0\ Suppose, now,
that the rivers and lakes on Twin Earth are filled with a liquid
that superficially resembles water, but which is not H 2 0. Then
A problem about reference 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-
Why is it surprising that the Brain in a Vat hypothesis turns out ation, does not fix the reference.
to be incoherent? The reason is that we are inclined to think that
Some philosophers have objected to this example, however.
what goes on inside our heads must determine what we mean These philosophers suggest that one should say, if such a planet
and what our words refer to. But it is not hard to see that this is is ever discovered, that 'There are two kinds of water', and not
wrong. Ordinary indexical words, such as I, this, here, now, are that our word 'water' does not refer to the Twin Earth liquid. If
a counterexample of a trivial sort. I may be in the same mental we ever find lakes and rivers full of a liquid other than H 2 that
state as Henry when I think 'I am late to work' (imagine, if you superficially resembles water, then we will have falsified the
like, that Henry and I are identical twins) and yet the token
of
statement that all water is H
2 0, according to these critics.
the word T that occurs in my thought refers to me and the token
It is easy to modify the example so as to avoid this argument.
of the word T that occurs in Henry's thought refers to Henry, I
First of all, the liquid on Twin Earth need not be that similar to
may be in the same mental state
1
when I think 'I am late to work'
water. Suppose it is actually a mixture of 20% grain alcohol and
on Tuesday and when I think 'I am late to work' on Wednesday; 80% water, but the body chemistry of the Twin Earth people is
but the time to which my tensed verb 'am' refers is different in such that they do not get intoxicated or even taste the difference
the two cases. The case of natural kind terms is a more subtle H
between such a mixture and 2 0. Such a liquid would be differ-
example of the same point. ent from water in many ways; yet a typical speaker might be
Suppose, to spell out the case mentioned in the previous chap- unacquainted with these differences, and thus be in exactly the
ter, that there are English speakers on Twin Earth (by a kind
of
same mental state as a typical speaker in 1750 on Earth. Of
1
At least I may be in the 'same mental state' in the sense that the course Twin Earth 'water' tastes different from Earth water to
parameters involved in the psychological process that results in my us; but it does not taste different to them. And it behaves differ-
thinking the thought may have the same values. My global mental state ently when you boil it; but must an English speaker have noticed
is Tuesday'
is, to be sure, different since on Tuesday I believe 'this
exactly when water boils and exactly what takes place in order
and on Wednesday I don't; but a theory that says the meaning of the
words changes whenever my global mental state changes would to associate a fairly standard conceptual content with the word
not allow any words to ever have the same meaning, and would thus 'water'?
amount to an abandonment of the very notion of word meaning.
Moreover, we could construct a Twin Earth story in which 1 and my 2 See 'The Meaning of "Meaning" in my Mind, Language, and Reality
'

Doppelganger are in the same global mental state, and the reference {Philosophical Papers, vol. 1), Cambridge University Press, 1975, for an
of T and 'now' is still different (the calendar on Twin Earth is not extended discussion of this point.
synchronized with ours).
24 A problem about reference A problem about reference IS

may be objected that there might well be experts on Twin

It
or 'water is a mixture of two liquids', or does not change its
ordinary meaning and reference (of course it may develop more
Earth who do know things about 'water' (for instance, that it is
technical uses as a result of such discoveries),
a mixture of two liquids) that we do not know about water (did and that 'water' in

not believe about water in 1750, because they aren't true of itsordinary Earth meaning and reference does not include mix-
tures of alcohol and water, then we must say that expert knowl-
water), and hence that the collective mental state of Twin Earth
English speakers is different from the collective mental state of edge is not what accounts for the difference in the meaning of
Earth English speakers (in 1750). One might concede that the the word 'water' on Earth and on Twin Earth. Nor does it

reference of a person's term isn't fixed by his individual mental account for the reference: for we could consider yet another
state, but insist that the total mental state of all the members of
Twin Earth where water was a different mixture and the expert
the language community fixes the reference of the term.
knowledge was the same (rather scanty) expert knowledge as on
the first Twin Earth. Or, as just indicated, we could simply imag-
One difficulty with this is that it might not have been the case
ine that experts on Earth and on Twin Earth did not yet exist.
that people on Earth or Twin Earth had developed that much
chemistry in 1750. If the term had the same meaning and refer-
The word 'water' would still refer to different stuff even if the
does collective mental state in the two communities were the same.
ence oh Earth prior to the development of chemistry that it

today ordinary use), and if the term had the same meaning
(in
What goes on inside people's heads does not fix the reference of
and reference on Twin Earth prior to the development of the their terms. In a phrase due to Mill, 'the substance itself com-
corresponding knowledge that it does today, then we can go pletes the job of fixing the extension of the term.

back to this earlier time when the collective mental states of the Once we see that mental state (in either the individualistic or

two communities were same in all respects relevant to fixing

the the collective sense) does not fix reference, thenwe should not
the extension of 'water', and argue that the extension was differ- be surprised that the Brains in a Vat could not succeed in refer-
ent then (as it is now) and so the collective mental state does not ring to external objects (even though they have the same mental
fix the extension. Should we then say the reference changed states we have), and hence could not say or think that they are
when chemistry was developed? That the term used to refer to Brains in a Vat.

both kinds of water (in spite of the difference in taste to us!), and
only refers to different kinds after chemistry is developed? Intentions, extensions, and 'notional worlds'
If we say that the reference of their terms or of our terms

changed when they or we developed chemistry (to the extent of In order to look at the problem of how the reference of our terms
distill liquids, tell that water plus alcohol is a mix- is fixed, given that not fixed simply by our mental states, it
it is
being able to
ture, etc.) then we will have to say that almost every scientific will be convenient to have available some technical terms. In

discovery changes the reference of our terms. We did not dis- logic the set of things a term is true of is called the extension of

cover that water (in the pre-scientific sense) was H 2 on such a the term. Thus the extension of the term 'cat' is the set of cats. If

view; rather we stipulated it. To me

seems clearly wrong.
this a term has more than one sense, then we pretend the word car-
What we meant by water all along was whatever had the same ries invisible subscripts (so that there are really two words and

nature as the local stuff picked out by that term; and we discov- not one), e.g. 'rabbit i - extension: the set of rabbits - 'rabbit 2 ' -
ered that water in that sense was H 2 0; what the people on Twin extension: the set of cowards. (Strictly speaking, the extension

Earth meant by 'water' all along was the stuff in their environ- of terms in a natural language is always somewhat fuzzy: but we
ment picked out by that term, and their experts discovered that shall pretend for simplicity that the borderline cases have been
'water' in that sense is a mixture of two liquids. somehow legislated.)
If agree that 'water' does not change meaning (in either
we A word like T which refers to different people on different

language) when experts make such discoveries as 'water is 2

0' H occasions will have not an extension but an extension- function:
26 A problem about reference A problem about reference 27

that is a function which determines an extension in each context

What the intension does is to specify how the extension
of use. In the case of the word T
the extension-function is rather depends on the possible world. It thus represents what we are
interested in, the extension associated with a term, in a very
simple; it is simply the function f(x) whose value for any speaker
x is the set consisting of just x. The argument which ranges V complete way, since it says what that extension would have been
in any possible world.
over the relevant parameter used to describe the context (in this
case, the speaker) is referred to in semantics as an index. Indices
The reason 'intension' (in this sense) cannot be identified with
are needed for times, for things demonstratively referred to, and
meaning is that any two terms which are logically equivalent
for yet other features of context in a full semantic treatment (but
have the same extension in every possible world, and hence the
we shall ignore the details).
same intension, but a theory which cannot distinguish between
The set of things which makes up the extension of 'cat' is dif- terms with the same meaning and terms which are only equiva-
lent in logic and mathematics is inadequate as a theory of mean-
ferent in different possible situations or 'possible worlds'. In a
possible world M
which there are no cats, the extension of
in
ing. 'Cube' and 'regular polyhedron with six square faces' are
logically equivalent predicates. So the intension of these two
'cat' is the empty my cat Elsa had had offspring, then the
set. If

extension of 'cat' would have at least one member it does not

terms the same, namely the function whose value in any pos-
is

sible world is the set of cubes in that world; but there is a differ-
have in the actual world. (We can express this by saying that in
each possible world M
in which Elsa had offspring, the extension
ence in meaning which would be lost if we simply identified the

of 'cat' includes members it does not include in the actual meaning with this function.
Let me emphasize that possible worlds, sets, and functions are
world.)
to be thought of as abstract extra-mental entities in this theory,
Wecan indicate the way in which the extension of a term
varies with the possible world M
in exactly the way in which we and not to be confused with representations or descriptions of
indicate how the extension of the word varies with the T these entities.
Frege thought that the meaning (Sinn) of an expression was
speaker: by using a function. We assume a set of abstract objects
an extra-mental entity or concept which could somehow be
called 'possible worlds' to represent the various states of affairs
we associate with the term 'cat' 'grasped' by the mind. Such a theory cannot help us with inten-
or possible world histories, and
a function f(M) whose value on each possible world is the set M 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
of possible objects which are cats in the world M. This function,
understanding of a term cannot consist only in associating it
following Montague and Carnap, I shall refer to as the inten-
the intension of the two-place with an intension. More important, if we assume that we have
sion 3 of the word 'cat'. Similarly,

predicate 'touches' is the function f(M) whose value on any pos- no which enables us to directly perceive extra-men-
'sixth sense'

sible world M is the set of ordered pairs of possible objects which tal entities, or to do something analogous to perceiving them
('intuiting' them, perhaps), then 'grasping' an intension, or any
touch each other in the world M; the intension of the three-place
extra-mental entity, must be mediated by representations in
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 some way. (This also seems clear introspectively, to me at least.)
But the whole problem we are investigating is how representa-
x is between y and z, and so on. The intension of a word like T,
tions can enable us to refer to what outside the mind. To
whose extension in any world is context-dependent, will be a is

more complicated function having as arguments both the possi- assume the notion of 'grasping' an X which is external to the
ble world and the indices representing the context.
mind would be to beg the whole question.
If I say of someone that he 'believes there is a glass of water
3
Montague, R., Formal Philosophy, Yale University, 1974. This use of on the then
table', I normally attribute to him the capacity to
'intension' is not the traditional one which I discussed in 'The
Meaning of "Meaning" \
refer to water. But, as we have seen, being able to refer to water
28 A problem about reference A problem about reference 29

requires being directly or indirectly linked to actual water ifyou like; it is just that none of their terms had any external
(H 2 0); the statement 'John believes there is a glass of water in world reference at all. The traditional theory of meaning
front of him' is not just a statement about what goes on in John's assumed that a thinker's notional world determines the inten-
head, but is in part a statement about John's environment, and sions of his terms (and these, together with the fact that a partic-

John's relationship to that environment. If it turns out that John ular possible world M
is the actual one, determine the extensions

is a Twin Earth person, then what John believes when

he says of the terms and the truth- values of all the sentences). We have
'there is a glass of water on the table' is that there is a glass seen that the traditional theory of meaning is wrong; and this is
containing a liquid which in fact consists of water and grain why the literature today contains many different concepts (e.g.,

alcohol on the table. 'intension' and 'notional world') and not a single unitary con-

Husserl introduced a device which is useful when we wish to cept of 'meaning'. 'Meaning' has fallen to pieces. But we are left
talk of what goes on in someone's head without any assumptions with the task of picking up the pieces. If intension and extension
about the existence or nature of actual things referred to by the are not directly fixed by notional world, then how are they
4
thoughts: the device of bracketing. If we 'bracket' the belief that fixed?
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-

The received view of interpretation
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- The most common view of how interpretations of our language
eted' ordinary sense). Thus, if John on Twin Earth cannot taste are fixed by us, collectively if not individually, is associated with
the difference between water and water-cum-grain-alcohol, he the notions of an operational constraint and a theoretical con-
may be in the same mental state as an actual or possible speaker straint. Operational constraints were originally conceived of
of Earth English when he says 'there is a glass of water on the rather naively; we simply stipulate (conventionally, as it were)
table', notwithstanding the fact that what he refers to as water that a certain sentence (say, 'Electricity is flowing through this
would make a reasonable highball. We will say that he has the wire') is to be true if and only if a certain test result is observed
bracketed belief that [there is a glass of water on the table]. In (the voltmeter needle being deflected, or, in a phenomenalistic
effect, the device of bracketing subtracts entailments from the version, my having the visual impression of seeing the voltmeter
ordinary belief locution (all the entailments that refer to the needle being deflected). This sort of crude operationalism no
external world, or to what is external to the thinker's mind). longer has any defenders because it has been appreciated that (1)
Daniel Dennett has recently used the locution 'notional world' the links between theory and experience are probabilistic and
5
in a way related to the way Husserl used bracketing. The total- cannot be correctly formalized as perfect correlations (even if
ity of a thinker's bracketed beliefs constitute the description of there is current flowing through the wire, there are always low
the thinker's notional world, in Dennett's sense. Thus, people on probability events or background conditions which could pre-
Twin Earth have roughly the same notional world (and even the vent the voltmeter needle from being deflected); and (2) even
same notional water) that we do; it is just that they live on a these probabilistic links are not simple semantic correlations, but
different real planet (and refer to different actual stuff as depend on empirical theory which is subject to revision. On a
'water'). And the Brains in a Vat of the previous chapter could naive operationist account every time a new way of testing
have had the same notional world we do down to the last detail, whether a substance is really gold is discovered, the meaning and
reference of 'gold' undergoes a change. (In fact, we shouldn't
4 Husserl, Ideas; General Introduction to Pure Phenomenology, Allen and speak of a new test for gold being discovered,) On an operation-
Unwin, 1969. (Originally appeared in 1913).
s
Dennett, D. 'Beyond Belief', in Thought and Object, Andrew Woodfield
ist picture, theories are tested sentence by sentence (the stipu-
(ed.), Oxford University Press, forthcoming. lated operational meanings of the individual sentences tell you
30 A problem about reference A problem about reference 31

how to go about testing the theory); on the more recent picture, ory. For example, 'an admissible interpretation is such that it

theories 'meet the test of experience as a corporate body', as turns out to be true that different effects always have different

Quine puts it.

causes'. Kant held that such a 'theoretical constraint' was part of

It is possible, however, to relax the notion of an operational rationality itself: we impose the principle of determinism on the

constraint so as to overcome or bypass

of these objections.
all world rather than discovering it. In this form, the constraint is
Thus one can restrict the class of interpretations (assignments of certainly too strong: the price of preserving determinism might
intensions to the predicates of one's language) that will be be too great a complication of our system of knowledge as a
accepted as admissible by constraints of the form: 'an admissible whole. But this sort of constraint can be relaxed just as opera-
interpretation is such that most of the time the sentence S is true tional constraints have been relaxed. (For example, one can

when the experiential condition £ is fulfilled' (respectively, 'such require that determinism be preserved whenever the 'cost' in
that most of the time the sentence S is false when £ is fulfilled'). terms of complications in other parts of the theory is not too
Such constraints model the idea that there are probabilistic rela- great; in this form, the constraint seems to be one we accept.)

tions between truth or falsity of sentences in the language and Theoretical constraints are often stated as constraints on the-
experience. And, secondly, one can take the view that these con- ory acceptance rather than as constraints on theory interpreta-
straints are revisable as theory develops. Rather than thinking of tion, but they can easily be reinterpreted to play the latter role.

them as meaning stipulations, as crude operationism did, one Thus, if an author states the constraint of 'conservativism' or
can think of them as tentative restrictions on the class of admis- 'preservationism' as a constraint on theory acceptance ('do not
and with Peirce (who wrote 50 years before
sible interpretations; accept a theory which requires giving up a great many previously
Bridgeman announced 'operationism'!) one can take the view accepted beliefs if an - otherwise equally 'simple' - theory is
that the ideal set of operational constraints is itself something available which preserves those beliefs and agrees with observa-
that we successively approximate in the course of empirical tion'), then we can reformulate the constraint as a constraint on
inquiry, and not something we just stipulate. In short, one can interpretation thus: 'an admissible interpretation is such that it

take the view that it is the operational constraints that rational renders true sentences which have been accepted for a long time,

inquirers would impose, if they observed and experimented and except where this would require undue complication in the the-
reasoned as well as is possible, the constraints that they would ory consisting of the set of sentences true under the interpreta-
adopt in the state of 'reflective equilibrium', that singles out the tion, or too great a revision in the operational constraints'.
interpretation of our terms; the constraints we actually accept at Again, has been widely held that no inductive logic is possible
it

any given time have the status of a rational estimate or approx- unless we impose some a priori ordering (called a 'simplicity

imation. ordering' or 'plausibility ordering') on the hypotheses which

Such a view is compatible with Quine's insistence that the may be accepted given particular observational data (although
theory-experience links are just as much subject to revision as the ordering may itself be different in different experimental or

any other aspect of our corporate body of knowledge. And it observational contexts); the constraint that 'the set of sentences
does not see each such revision as a 'meaning change': such revi- which is true under an admissible interpretation must not be
sions can be and often are simply better efforts to specify what lower down in the simplicity ordering than any other set with
we have already been talking about; that which earlier the-
it is
the same observational or experiential consequences would cor-
ory-cum-operational-constraints captured only inadequately. respond to the constraint in inductive logic that one is to accept
In addition to restricting the class of admissible interpretations the most simple (or most 'plausible') of the hypotheses compat-
by means of operational constraints (or successive approxima- ible with one's observations.

tions to a Peircian ideal set of operational constraints), one can Theoretical constraints of many other kinds have been pro-
also have constraints which refer to formal properties of the the- posed in the literature of the philosophy of science. There are
 pr
32 A problem about reference A problem about reference 33

constraints which, like 'simplicity', refer to properties of the set conditions for whole sentences. The idea, as we just saw, is that
of accepted sentences, and constraints which refer to the history the operational and theoretical constraints (the ones rational
of the inquiry by which that set came to be accepted. But the inquirers would accept in some sort of ideal limit of inquiry)
details need not concern us. The reasons for being attracted to determine which sentences in the language are true. Even if this
the idea that the admissible interpretations of our language (in is right, however, such constraints cannot determine what our
the sense of admissible intension-assignments to the terms of the terms refer to. For there is nothing in the notion of an opera-
language) are fixed by operational and theoretical constraints tional or theoretical constraint to do this directly. And doing it
are obvious: whether or not one is having an experience of a indirectly, by putting down constraints which pick out the set of
certain kind something the mind is able to judge (philosophical
is true sentences, and then hoping that by determining the truth-
problems about 'experience' notwithstanding). So if a theory values of whole sentences we can somehow fix what the terms
implies or contains a sentence which is associated with an expe- occurring in those sentences refer to, won't work.
rience E
by an operational constraint of some kind, probabilistic That it won't work has been shown by Quine. 6 I shall extend
or whatever, then the thinker can know if the theory works, or previous 'indeterminacy' results in a very strong way. I shall
if there is some awkwardness of fit, at least in this case, by seeing argue that even if we have constraints of whatever nature which
whether or not he has the experience E. Since the constraints determine the truth-value of every sentence in a language in
that we use to test the theory also fix the extensions of the terms, every possible world, still the reference of individual terms

the thinker's estimate of the 'theory's 'working' isat the same remains indeterminate. In fact, it is possible to interpret the
time an estimate of its truth. Since the speaker's knowledge of entire language in violently different ways, each of them com-
these constraints is knowledge of the intensions of the terms, patible with the requirement that the truth-value of each sen-
grasping a correct semantics would tell one, for any proposed tence in each possible world be the one specified. In short, not
theory of the world, what our world would have to be like for only does the received view not work; no view which only fixes
the theory to be true. the truth-values of whole sentences can fix reference, even if it
Furthermore, if we
by supposing thinkers to have
idealize specifies truth-values for sentences in every possible world.
what economists call 'perfect information' about each other, The detailed proof is technical, and I think it appropriate to
each thinker knows the formal structure of the accepted theory give an Appendix. What I shall give here is an illustration of
it in
T and the past history of the research program to which it the method of the proof only, and not the detailed proof.
belongs, the previous beliefs that it does or does not preserve, Consider the sentence
etc. So each thinker is know if the theoretical
in a position to
constraints are met or we do not wish to idealize by
not. (If (1) A cat is on a mat. (Here and in the sequel 'is on' is

assuming perfect information, then we can still say that the col- tenseless, i.e. it means 'is, was, or will be on'.)

lective body of thinkers is in a position to know this.)

In short, if the received view is correct, then we would have an
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
elegant account of how intensions and extensions are fixed (in
principle - of course the details are too complicated to fill in at
some time, past, present, or future. Moreover, 'cat' refers to cats
the present stage of methodological knowledge). But, unfortu-
and 'mat' refers to mats. I shall show that sentence (1) can be
reinterpreted so that in the actual world 'cat' refers to cherries
nately, the received view does not work!
and 'mat' refers to trees without affecting the truth-value of sen-
tence (1) in any possible world. ('Is on' will keep its original
Why the received view doesn't work interpretation.)

The difficulty with the received view is that it tries to fix the 6 See his 'Ontological Relativity*, in Ontological Relativity and Other
intensions and extensions of individual terms by fixing the truth- Essays, Columbia University Press, 1969.
34 A problem about reference A problem about reference 35

The idea is that sentence (1) will receive a new interpretation word 'cat' by assigning to it the intension we just assigned to
in which what it will come to mean is: 'cat*' and simultaneously reinterpreting the word 'mat' by
assigning to it the intension we just assigned to 'mat*' would
(a) A cat* is on a mat*.
only have the effect of making 'A cat is on a mat' mean what 'A
The definition of the property of being a cat* (respectively, a cat* is on a mat*' was defined to mean; and this would be per-
mat*) is given by cases, the three cases being: fectly compatible with the way truth-values are assigned to 'A

(a) Some cat is on some mat, and some cherry is on some cat is on a mat' in every possible world.

tree.
In the Appendix, I show that a more complicated reinterpre-

Some on some mat, and no cherry tation of this kind can be carried out for all the sentences of a
(b) cat is is on any
tree.
whole language. It follows that there are always infinitely many
(c) Neither of the foregoing. different interpretations of the predicates of a language which
assign the 'correct' truth-values to the sentences in all possible
Here is the definition of the two properties: worlds, no matter bow these 'correct
9
truth-values are singled

DEFINITION OF 'CAT*' out. Quine argued for a similar conclusion in Word and Object;
x a cat* and only case holds and x
in Quine's example (as applied to English) 'There is a rabbit over
is if if (a) is a cherry;
there' was interpreted to mean 'There is a rabbit-slice over there'
or case (b) holds and x is a cat; or case (c) holds and x is
(where a 'rabbit-slice' is a three-dimensional spatial cross-
a cherry.
section of the whole four-dimensional space-time rabbit), or,
DEFINITION OF 'MAT*'
alternatively again, to mean, 'There is rabbithood being exem-
x is a mat* and only if case (a) holds and x is a tree;
if
plified again.' (This last reinterpretation also reinterprets the
or case (b) holds and x is a mat; or case (c) holds and
syntactic form of the sentence, or at least its logical grammar.)
x is a quark.
Quine makes the point I just made, that truth -conditions for
Now, in possible worlds falling under case (a), 'A cat is on a whole sentences underdetermine reference. Since 'rabbit-slices',
mat' is true, and 'A cat* is on a mat*' is also true (because a 'rabbithood', and 'undetached rabbit-parts' all have a close con-
cherry is on a tree, and all cherries are cats* and all trees are nection to rabbits, one might come away from Word and Object
mats* in worlds of this kind). Since in the actual world some with the impression that all reinterpretations that leave a sen-
cherry ison some tree, the actual world is a world of this kind, tence's truth-value unchanged are at least closely connected with
and in the actual world 'cat*' refers to cherries and 'mat*' refers the standard interpretation (in the way that rabbit-parts and rab-
to trees. bithood are connected with rabbits). The argument spelled out
is on a mat' is
In possible worlds falling under case (b), 'A cat in the Appendix and illustrated in this chapter shows that the
true, and 'A cat* on a mat*' is also true (because in worlds
is truth conditions for 'A cat is on a mat' don't even exclude the
falling under case (b), 'cat' and 'cat*' are coextensive terms and possibility that 'cat' refers to cherries.
so are 'mat' and 'mat*'). (Note that although cats are cats* in
some worlds - the ones falling under case (b) - they are not
'Intrinsic' and 'extrinsic'
cats* in the actual world.)
In possible worlds falling under case (c), 'A cat is on a mat' is Perhaps thefirst idea that comes to mind when one is confronted

false and 'A cat* is on a mat*' is also false (because a cherry by non-standard interpretations, such as the one that interprets
can't be on a quark). 'cat' as cat * and 'mat' as mat* is to dismiss them as presenting
Summarizing, we see that in every possible world a cat is on a us with an unimportant paradox. But genuine paradoxes are
mat if and only if a cat* is on a mat*. Thus, reinterpreting the never unimportant; they always show something is wrong with
36 A problem about reference A problem about reference 37

the way we have been thinking. Perhaps the second reaction is to The point is that the fact that one can build a machine to
protest that cat* and mat* are 'queer' properties; surely our inspect things and tell if they are cats differentiates cats from
terms correspond to 'sensible' properties (such as being a cat or cats* if one can be sure 'inspect' and 'tell' refer to inspecting and
being a mat) and not to such 'funny' properties as these. One telling, and it is no easier to say how the reference of these words
might explicate the way in which cat* (or, rather, cathood* or is fixed than to say how the reference of 'cat' is fixed. One might
cat* hood) a funny property by pointing out that one can
is say that when I look at something and think that it is a cat, my
'build a machine' to 'inspect' things and 'tell' whether or not 'mental representations', the visual images or tactile images, the
they are cats (a human being is such a 'machine'), but one cannot verbalized thought 'cat', and so on, refer to cathood and to var-
build a machine that will any world which resembles ours
tell (in ious other physical or biological properties (being a certain
in its laws and general conditions) whether or not something is shape, being a certain color, belonging to a certain species) and
a cat*. If the machine (or a person) looks at something and sees not to their counterparts; this may be true, but it just repeats
it is neither a cat nor a cherry, then they can tell it is not a cat*; that the reference one way rather than the other. This is
is fixed
but if the thing is either a cat or a cherry then the device or the what we want to explain and not the explanation sought.
person needs to be informed of the truth- values of 'A cat is on a 'But,' one might protest, 'the definitions oV'cat*" and"mat*"
mat' and 'A cherry is on a tree' to decide if it is inspecting or given above refer to things other than the object in question
seeing a cat*, and these truth- values go beyond what it can learn (cherries on trees and cats on mats), and thus signify extrinsic
by just examining the object presented to it for inspection. properties of the objects that have these properties. In the actual
Unfortunately, one can reinterpret 'sees' (say, as sees*) so that world, every cherry is a cat*; but it would not be a cat*, even
the two sentences (3) John (or whoever) sees a cat; and (4) John though its intrinsic properties would be exactly the same, if no
sees* a cat*, will have the same truth-value in every possible cherry were on any tree. In contrast, whether or not something is
world (by the method given in the Appendix). So whenever a a cat depends only upon its intrinsic properties.' Is the distinc-
person sees a cat, he is seeing* a cat*; the experience we typically tion here referred to, the distinction between intrinsic and extrin-
have when we see a cat is the experience we typically have when sic properties, one that will enable us to characterize and rule
we see* a cat*, and so on. Similarly, we can reinterpret 'inspects' out 'queer' interpretations?
and 'tells' so that, when a machine inspects a cat, it is inspecting* The trouble with this suggestion is a certain symmetry in the
a cat*, and when it 'tells' something is a cat, it is telling* that it relation of 'cat' and 'mat' to 'cat*' and 'mat*'. Thus, suppose we
is a cat*. define 'cherry*' and 'tree*' so that in possible worlds falling
To use an illustration (suggested by Nozick), suppose half of under case (a) cherries* are cats and trees* are mats; in possible
us (the females perhaps) use 'cat' to mean 'cat*', 'mat' to mean worlds falling under case (b) cherries* are cherries and trees* are
'mat*', 'look' to mean 'look*', 'tells' to mean 'tells*', and so on. and in possible worlds falling under case (c) cherries* are
trees;
Suppose the other half (the males) use 'cat' to denote cats, 'mat' catsand trees* are photons. Then we can define 'cat' and 'mat'
to denote mats, 'look' to denote looking, and so on. How could by means of the *-terms as follows: Cases:
we ever know? 7 (Ifyou ask a male what 'cat' refers to, he will
(a*) Some cat* is on some mat*, and some cherry* is on
answer 'to cats, of course' and so will a female, whatever 'cat'
some tree*.
refers to.)
(b*) Some cat* is on some mat*, and no cherry* is on
any tree*.
7
A female might answer that the supposition that she is referring to
cats* when she says 'cat' is incoherent {because within her language (c*) Neither of the foregoing.
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
Strangely enough, these cases are just our old (a), (b), (c) under
call a cat*, and vice versa; and this is Nozick's point. a new description. Now we define:
38 A problem about reference A problem about reference 39

DEFINITION OF 'CAT which 'Hamlet' only refers to a person in a play); the success of
x is a cat if case (a*) holds and x is a cherry*; or case science is explained by trial-and-error, not by any correspon-
(b * ) holds and x is a cat * ; or case (c * ) holds and x is a dence between its objects and real things, Kuhn says. Bas van
cherry*. (Note that in all three cases cats come out being Fraassen, in a new book, argues that a successful theory need
cats.) not be true but only 'observationally adequate', i.e. correctly
DEFINITION OF 'MAT predict observation. He too explains the success (or 'observa-
x is a mat if and only if case (a*) holds and x is a tree*; tional adequacy') of science as the product of trial-and-error.
or case (b*) holds andx is a mat*; or case (c*) holds If these philosophers are right, then the whole idea of using
and* is a quark*. (Supposing quark* to be defined so evolution to justify belief in an objective relation of reference is
that in cases of type (c *
) quarks * are mats, in all three undercut. Evolution, on such instrumentalist views, only estab-
cases mats come out being mats.) lishes a correspondence between some terms (the observation
terms) and 'permanent possibilities of sensation'. Such a corre-
The upshot is that viewed from the perspective of a language
which takes 'cat*', 'mat*', etc., as primitive properties, it is 'cat'
spondence is not reference, unless we are willing to abandon the
idea that external things (the observable ones) are more than
and 'mat' that refer to 'extrinsic' properties, properties whose
constructs out of sensations.
definitions mention objects other than x; while relative to 'nor-
I believe that the other philosophers are right, however (the
mal' language, language which takes 'cat' and 'mat' to refer to
ones who say we would not survive if sufficiently many of our
cathood and mathood {you know which properties I mean, dear
beliefs were not true).
reader!), it is 'cat*' and 'mat*' that refer to 'extrinsic' properties.
The reason I believe this is that trial-and-error does not
Better put, being 'intrinsic' or 'extrinsic' are relative to a choice
of which properties one takes as basic; no property
explain why our theories are 'observationally adequate'; that
is intrinsic or
can only be explained by referring to characteristics of the envi-
extrinsic in itself.
ronment-human interaction which explain why trial-and-error
is successful. (Trial-and-error does not succeed in all enterprises,

'Survival' and evolution after all!) To posit that the interaction produces in our minds
false theories which just happen to have successful predictions
The suggestion popular nowadays that the evolutionary pro-
is
as consequences is to posit a totally inexplicable series of coin-
cess itself has somehow produced a correspondence between our cidences. But how does the fact that our beliefs are (approxi-
words and mental representations and external things; people mately) true explain our survival?
say that we would not have survived if there had not been such Some of our connected with action. If I
beliefs are intimately
a correspondence, and that this correspondence is, at least in a believe the sentence something I value very much if I
'I will get
primitive way, the relation of reference. push that button' (assume I understand this sentence in a normal
But what do 'correspondence' and 'reference' have to do with way, or at least associate the normal 'bracketed' or 'notional'
what does truth have to do with sur-
survival? For that matter, belief with it), then I will reach out my hand and push the but-
vival? ton. Call beliefs of the form 'If I do x, I will get .', where the . .

Here opinions differ. Some philosophers believe that we blank describes a goal the agent has, directive beliefs. If too
would not survive if (sufficiently many of) our beliefs were not many of our directive beliefs are false, we will perform too many
true. Other philosophers claim that even our best established sci- unsuccessful actions; so truth of (sufficiently many of) our direc-
entific beliefs aren't true, or at least that we have no reason to tive beliefs is necessary for survival.
think they are. Thomas Kuhn has suggested that our beliefs only Now, our directive beliefs are themselves derived from many
'refer' to objects within those beliefs (somewhat in the way in
other beliefs: beliefs about the characteristics and causal powers
40 A problem about reference A problem about reference 41

of external things, and beliefs about our own characteristics and

have certain truth conditions (and certain action conditions, or
'language exit rules'). But the truth-conditions for whole sen-
powers. If these beliefs were mainly false, would it not be a mere
they nonetheless led to true prediction of experi- tences were just shown not to determine the reference of sen-
coincidence if
tence parts (nor does adding the 'language exit rules' help, for
ence and to true directive beliefs? So, since (sufficiently many of)
these are preserved under/). It follows that it is simply a mistake
our directive beliefs are true, and the best explanation of this fact
ones constituting our 'the- to think that evolution determines a unique correspondence (or
is that many of our other beliefs (the

ory of the everyday world') are at least approximately true, we even a reasonably narrow range of correspondences) between
are justified in believing that our theory of the everyday world is referring expressions and external objects.
at least approximately true, and that we would not have sur-
vived if this were not the case.
Intentions: pure and impure
Imagine, now, that some of us are actually referring to the,
things that are assigned to our terms by the non-standard inter- We have seen that nature does not single out any one correspon-
pretation / (described in the Appendix). This interpretation dence between our terms and external things. Nature gets us to
agrees with the standard interpretation on terms referring to our
process words and thought signs in such a way that sufficiently

notional world, our sensations, our volitions, etc. So 'I seem to many of our directive beliefs will be true, and so that sufficiently
myself to push the button', when understood in the 'bracketed many of our actions will contribute to our 'inclusive genetic fit-

sense' (as meaning that I have a certain subjective experience of

ness'; but this leaves reference largely indeterminate. W. V.

voluntarily pushing a button) has not just the same truth condi- Quine has urged that that is what reference in fact is - indeter-
tions but thesame interpretation under/ and under the 'normal' minate! It is just an illusion, he thinks, that the terms
in our

interpretation I, and so does 'I seem to myself to get the satisfac- language have determinate well-defined counterparts. As he puts
it,
tion I expected.'
Now, if sufficiently many of our directive beliefs are true
For, consider again our standard regimented nota-
under the non-standard interpretation /, then we will certainly
tion, with a lexicon of interpreted predicates and some
be successful, and we will certainly survive (since if we weren't
fixed range of values for the variables of quantifica-
alive we wouldn't be attaining these goals) and have offspring tion. The sentences of this language that are true remain
(since if they weren't alive they wouldn't be attaining these true under countless reinterpretations of the predicates
goals). In short, /-truth of (sufficiently many) directive beliefs is
and revisions of the range of values of the variables.
as good for 'evolutionary success' as J-truth. In fact it is /-truth, Indeed any range of the same size can be made to serve by
since the truth conditions for every sentence (not just directive a suitable reinterpretation of the predicates. If the range
beliefs) are the same under I and under /. My directive beliefs
of values is infinite, any infinite range can be made to
are not only associated with the same subjective experience un- serve; this is the Skolem-Lowenheim theorem. The true
der the interpretation I and under the interpretation/; they have sentences stay true under all such changes.
the same truth conditions. From the point of view (or non-point
Perhaps then our primary concern belongs with the
of view) of 'evolution', all that is necessary is that sufficiently
truth of sentencesand with their truth conditions, rather
many of my be true under any interpretation that con-
beliefs
than with the reference of terms.
nects those beliefs with the relevant actions. Evolution may pro-
In the next chapter will explore the alternative here sug-
duce in me a tendency to have true beliefs (of certain kinds); but I

this only means that evolution affects linguistically mediated or

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
conceptually mediated survival via its tendency to produce in us
relation to (discourse-independent) things and sets of things. It
representation systems whose sentences or sentence-analogues
42 A problem about reference A problem about reference 43

may seem, however, that there is a much simpler way out: why state, although they could be in the corresponding 'bracketed'
not just say that it is our intentions, implicit or explicit, that fix state.)

the reference of our terms? What goes for belief goes for intention as well. Pure mental
At the beginning of the discussion in the previous chapter, I states of intending - e.g. intending that the term 'water' refer to
rejected this as not constituting an informative answer on the water in one's notional world - do not fix real world reference
ground that having intentions (of the relevant kind) presupposes at all. Impure mental states of intending - e.g. intending that the
the ability to refer. It may be good at this stage to expand upon term 'water' refer to actual water - presuppose the ability to

this brief remark. refer to (real) water.

The problem is that the notions 'intention' and 'mental state' Some philosophers have suggested that belief can be defined
have a certain ambiguity. Let us call a mental state a pure mental in terms of the state I called 'bracketed belief and reference,

state if its presence or absence depends only on what goes on thus:

'inside' the speaker. Thus whether or not I have a pain depends = John
John believes that snow is white believes that
only on what goes on 'inside' me, but whether or not I know [snow is white]
that snow is white depends not only on whether or not some- (i.e. snow is white in John's notional world)
thing goes on 'inside' me (believing or being confident that snow and the words 'snow' and 'white' in John's thought (or
is on whether or not snow is white, and thus is
white), but also whatever words he uses to express this belief) refer
something 'outside' my body and mind. Thus pain is a pure men- to snow and to the property white, respectively.
tal state but knowledge is an impure mental state. There is a

(pure) mental state component to knowledge, but there is also a Without accepting this as a correct and complete analysis of
component which is not mental in any sense: this is the compo- what it is to believe that snow is white, we can accept this
nent that corresponds to the condition that what a man believes account as making a point which is certainly correct: that believ-
is not knowledge unless the belief is true. I am not in the 'state' ing presupposes the ability to refer. And in exactly the same way,
of knowing that snow is white if I am not in a suitable pure intending presupposes the ability to refer! Intentions are not
mental state;but being in a suitable pure mental state is never mental events that cause words to refer: intentions (in the ordi-
sufficient for knowing that snow is white; the world has to coop- nary 'impure' sense) have reference as an integral component.
erate as well. To explain reference in terms of (impure) intention would be
What about belief? We have defined bracketed belief circular. And the problem of how pure mental states of intend-

('notional world') so that having a bracketed belief that [there is ing, believing, etc., can (in the proper causal setting) constitute

water on the table] or having a notional world which includes or cause reference is just what we have found so puzzling.

there being water on the table is a pure mental state. But, in

accordance with what was said before, believing that there is The origin of the puzzle
water on the table (without any 'bracketing') presupposes that
one's word 'water' actually refers to water, and this depends on At first blush, nothing seems more obvious than that our words
the actual nature of certain 'paradigms', one's direct or indirect and mental representations refer. When I think or say 'the cat
causal relations to those paradigms, and so on. When I have the just went out', the thought is usually about our cat Mitty; the

belief that there is water on the table, my Doppleganger on Twin word 'cat' in the sentence I think or say refers to a set of entities
Earth has the same bracketed belief but not the same belief of which Mitty is a member. Yet we have just seen that the
because his word 'water' refers to water-with-grain alcohol and nature of this relation of 'aboutness' or reference is puzzling.

not to water. In short, believing that there water on the table

is The between real world and notional world (and
distinction

is an impure mental state. (Brains in a Vat could not be in this the correlative distinction between beliefs and bracketed beliefs,
44 A problem about reference A problem about reference 45

or intentions and bracketed intentions) explains part of the

itself
by operational and theoretical constraints), then two terms
puzzle. The reason that it is surprising and troubling to discover which refer to disjoint sets in each admissible interpretation can

that there are unintended 'admissible interpretations' of our lan-

have the same potential referents when the totality of all admis-
guage (where by an admissible interpretation I mean simply an sible interpretations is considered. From the fact that notional

interpretation that satisfies the appropriate operational and the- cats are as different as can be from notional cherries it does not
in part, that no such 'indeterminacy' rises follow that there are determinate disjoint sets of cats-in-
oretical constraints) is,

my notional world, cats themselves and cherries-in-themselves.

in the 'notional world' of the speaker. In
and cats* are quite distinct (in fact, in my notional world cats* What makes this so distressing is that operational plus theo-

are cherries). 'There is a cat on a mat' and 'there is a cat* on a retical constraints are the natural way in which to allow the
actual empirical context to determine the admissible interpreta-
mat*' may be logically equivalent, but they contain terms with
tion (or interpretations) of one's representational system. Such
quite different notional referents; thus it seems strange indeed
that there should be any confusion between the real world refer-
constraints can to some extent determine which sentences in

and the world referents of the other. one's language are true and which false; it is the slack between
ents of the one belief real

But if the number of cats happens to be equal to the number truth-conditions and reference that remains.

of cherries, then it follows from theorems in the theory of models Quine, as we remarked, would be willing to put up with the
(as Quine remarks in the passage quoted above) that there is a
slack and simply acknowledge that reference is indeterminate. A
reinterpretation of the entire language that leaves all sentences young philosopher, Hartry Field, 8 has recently suggested a dif-
ferent view. In Field's view reference is a 'physicalistic relation',
unchanged in truth value while permuting the extensions of 'cat'
and 'cherry'. By the techniques just mentioned, such reinterpre- i.e. a complex causal relation between words or mental represen-
tations can be constructed so as to preserve all operational and
tations and objects or sets of objects. up to empirical science
It is

theoretical constraints (and by the techniques we illustrated with to discover what that physicalistic relation is, Field suggests.

the 'cat/cat*' example, they can be extended so as to provide There is, however, a problem with this suggestion too. Sup-
'intensions', or functions which determine an extension in each pose there is a possible naturalistic or physicalistic definition of
possible world, and not just extensions in the actual world). This reference, as Field contends. Suppose

does not contradict the statements just made about our 'notional x refers to and only x bears R
(1) y if if to y
world', or subjective belief system, for the following reason: the
fact that in our belief system or 'notional world' no cat is a
is true, where R is a relation definable in natural science vocab-

cherry means that in each admissible interpretation of that belief ulary without using any semantical notions (i.e. without using

system (each assignment of external world referents to the terms, or any other words which would make the definition
'refers'

images, and other representations we employ in thought) the immediately circular). If (1) is true and empirically verifiable,
referents of 'cat' and the referents of 'cherry' must be disjoint
then (1) is a sentence which is itself true even on the theory that

sets. But the disjointness of these sets is comparable with the

reference is fixed as far as (and only as far as) is determined by
(remarkable) fact that what is the set of 'cats' in one admissible operational plus theoretical constraints. (1) is a sentence which
interpretation may be the set of 'cherries' in a different (but would be part of our 'reflective equilibrium' or 'ideal limit' the-

equally admissible) interpretation. From the fact that notional ory of the world.
cats are wholly different from notional cherries it only follows If reference is only determined by operational and theoretical
(

that real cats are wholly different from real cherries if the num- constraints, however, then the reference of x bears R to y' is

ber of admissible interpretations is exactly one. If there is more 8

Field, R, 'Tarski's Theory of Truth', The Journal of Philosophy, vol. 69.
than one admissible interpretation of the whole language (as Field's view is discussed in my Meaning and the Moral Sciences,
there will be if the admissible interpretations are singled out only Routledge and Kegan Paul, 1978.
46 A problem about reference A problem about reference 47

itself indeterminate, and so knowing that (1) is true will not help. refer to just those things that have the same lawful behavior and

Each admissible model of our object language will correspond the same ultimate composition as various standard samples of
to a model of our meta-language in which (1) holds; the inter- actual water (i.e. speakers have such intentions even when talk-
(

pretation of
(
x bears R to y* will fix the interpretation of x refers ing about hypothetical cases or 'possible worlds'), it follows that
to y\ But this will only be a relation in each admissible model; it (2) must also be true in every possible world; for to describe a
will not serve to cut down the number of admissible models at hypothetical liquid which is not 2 H
but which has some simi-
all. larities to water is only to describe a hypothetical liquid which
This is, of course, not at all what Field intends. What Field is resembles water, and not to describe a possible world in which
claiming is that (a) there is a determinate unique relation water isn't H 2 0. It is 'metaphysically necessary' (true in all pos-
between words and things or sets of things; and (b) this relation sible worlds) that water is H 2 6; but this 'metaphysical necessity'
is the one to be used as the reference relation in assigning
a truth is explained by mundane chemistry and mundane facts about
value to (1) itself. But this is not necessarily expressed by just speakers' intentions to refer.
saying (1), as we have just seen; and it is a puzzle how we could If there is a determinate physicalistic relation R (whether it be
learn to express what Field wants to say. definable in the language of natural science in finitely many
Putting this last puzzle aside, let us consider the view that (1), words or not) which just is reference (independently of how or
understood as Field wants us to understand it (as describing the whether we describe that relation), this fact cannot itself be the
determinate, unique relation between words and their referents), consequence of our intentions to refer; rather, as we have repeat-
it true? Given
is true. If (1) is true, so understood, what makes edly noted, it enters into determining what our very intentions
that there are many 'correspondences' between words and to refer signify. Kripke's view, that 'water is H 2
0' is true in all

things, even many that satisfy our constraints, what singles out possible worlds, could be right even if reference in the actual
one particular correspondence R? Not the empirical correctness world is fixed only by operational and theoretical constraints;
of (1); for that is a matter of our operational and theoretical the view presupposes the notion of reference, it does not tell us
constraints. Not, as we have seen, our intentions (rather R enters whether reference is determinate or what reference is.
into determining what our intentions signify). It seems as if the To me, believing that some correspondence intrinsically just is
fact that R is reference must be a metaphysically unexplainable reference (not as a result of our operational and theoretical con-
fact, a kind of primitive, surd, metaphysical truth. straints, or our intentions, but as an ultimate metaphysical fact)

This kind of primitive, surd, metaphysical truth, if such there amounts to a magical theory of reference. Reference itself
be, must not be confused with the sort of 'metaphysically neces- becomes what Locke called a 'substantial form' (an entity which
9
sary' truth recently introduced by Saul Kripke. intrinsically belongs with a certain name) on such a view. Even

Kripke's point, which is closely related to points made above if one is willing to contemplate such unexplainable metaphysical

about the reference of natural kind terms (terms for animal, veg- facts, the epistemological problems that accompany such a meta-

etable and mineral species, for example), was that given that, as physical view seem insuperable. For, assuming a world of mind-
a matter of fact, independent, discourse-independent entities (this is the presup-
position of the view we are discussing), there are, as we have
(2) Water is H 2 seen, many different 'correspondences' which represent possible
or candidate reference relations (infinitely many, in fact, if there
(i.e. given that (2) is true in the actual world), and given that
are infinitely many things in the universe). Even requiring that
(Kripke points out) speakers intend that the term 'water' shall
(1) be true under whichever notion of truth corresponds to the
metaphysically singled-out 'real' relation of reference does not
9
See his Naming and Necessity, Harvard University Press, 1980.
(Originally given as lectures in 1970.) exclude any of these candidates, if (1) is itself empirically accept-
48 A problem about reference

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- Two philosophical perspectives
tions) then even conceivable that the operational-p/as-
it is

theoretical-constraints-view is metaphysically correct after all!

For why could not be a surd metaphysical fact that reference
it

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
The problems we have been discussing naturally give rise to two
philosophical points of view (or two philosophical tempera-
of the world', and the same optimal methodology for discover-
ments, as I called them in the Introduction). It is with these
ing what is true!
points of view, and with their consequences for just about every
issue in philosophy that be concerned: the question of
I shall
'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 world consists of some fixed
this perspective, the
mind-independent objects. There is exactly one true
totality of
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 hold further that there is more than one 'true'
all,

theory or description of the world. 'Truth', in an internalist

view, is some sort of (idealized) rational acceptability - some
50 Two philosophical perspectives Two philosophical perspectives 51

sort of ideal coherence of our beliefs with each other and with thought does not have reference conditions that would make it
our experiences as those experiences are themselves represented true. So it is not possible after all that we are Brains in a Vat.
in our belief system - and not correspondence with mind-inde- Suppose we assume a 'magical theory of reference'. For exam-
pendent or discourse-independent 'states of affairs'. There is no ple, we might assume that some occult rays - call them 'noetic

God's Eye point of view that we can know or usefully imagine; rays' 1
-connect words and thought-signs to their referents.
there are only the various points of view of actual persons Then there is no problem. The Brain in a Vat can think the
reflecting various interests and purposes that their descriptions words, 'I am a brain in a vat', and when he does the word 'vat'
and theories subserve. ('Coherence theory of truth'; 'Non-real- corresponds (with the aid of the noetic rays) to real external vats
ism'; 'Verificationism'; 'Pluralism'; 'Pragmatism'; are all terms and the word 'in' corresponds (with the aid of the noetic rays)
that have been applied to the internalist perspective; but every to the relation of real spatial containment. But such a view is

one of these terms has connotations that are unacceptable obviously untenable. No present day philosopher would
because of their other historic applications.) espouse such a view. It is because the modern realist wishes to
Internalist philosophers dismiss the 'Brain in a Vat' hypothe- have a correspondence theory of truth without believing in
sis. For us, the 'Brain in a Vat World' is only a story, a mere
'noetic rays' (or, believing in Self-Identifying Objects 2 - objects

linguistic construction, and not a possible world at all. The idea that intrinsically correspond to one word or thought-sign rather

that this story might be true in some universe, some Parallel than another) that the Brain in a Vat case is a puzzler for him.
Reality, assumes a God's Eye point of view from the start, as is As we have seen, the problem is this: there are these objects
easily seen. For from whose point of view is the story being told? out there. Here is the mind/brain, carrying on its

Evidently not from the point of view of any of the sentient crea- thinking/computing. How do the thinker's symbols (or those of
tures in the world. Nor from the point of view of any observer his mind/brain) get into a unique correspondence with objects
in another world who interacts with this world; for a 'world' by and sets of objects out there?
definition includes everything that interacts in any way with the The reply popular among externalists today is that while
things it contains. If you, for example, were the one observer indeed no sign necessarily corresponds to one set of things rather
who was not a Brain in a Vat, spying on the Brains in a Vat, then than another, contextual connections between signs and external
the world would not be one in which all sentient beings were things (in particular, causal connections) will enable one to

Brains in a Vat. So the supposition that there could be a world explicate the nature of reference. But this doesn't work. For

in which all sentient beings are Brains in a Vat presupposes from example, the dominant cause of my beliefs about electrons is
the outset a God's Eye view of truth, or, more accurately, a No probably various textbooks. But the occurrences of the word
Eye view of truth - truth as independent of observers altogether. 'electron' I produce, though having in this sense a strong connec-

For the externalist philosopher, on the other hand, the tion to textbooks, do not refer to textbooks. The objects which
hypothesis that we are all Brains in a Vat cannot be dismissed so are the dominant cause of my beliefs containing a certain sign
simply. For the truth of a theory does not consist in its fitting the may not be the referents of that sign.
world as the world presents itself to some observer or observers The now reply that the word 'electron' is not
externalist will
(truth is not 'relational' in this sense), but in its corresponding to connected to textbooks by a causal chain of the appropriate
the world as it is in itself. And the problem that I posed for the type. (But how can we have intentions which determine which

externalist philosopher is that the very relation of correspon- causal chains are 'of the appropriate type' unless we are already
dence on which truth and reference depend (on his view) cannot able to refer f)
logically be available to him if he is a Brain in a Vat. So, if we 1
'Noetic rays' was suggested to me by Zemach.
are Brains in a Vat, we cannot think that we are, except in the 2
The term 'Self Identifying Object* is from Substance and Sameness by
bracketed sense [we are Brains in a Vat]; and this bracketed David Wiggins (Blackwell, 1980).
52 Two philosophical perspectives Two philosophical perspectives 53

real connection to. Using the basic terms in complex conbina-

For an internalist like myself, the situation is quite different.
In an internalist view also, signs do not intrinsically
correspond tions we can then, he says, build up descriptive expressions
objects, independently of how those signs are employed and which refer to kinds of things we have no real connection to, and
to
by whom. But a sign that is actually employed in a
particular that may not even exist (e.g. extraterrestrials).

correspond to par- In fact, already with a simple word like 'horse' or 'rabbit' he
way by a particular community of users can
ticular objects within the conceptual scheme of those users. might have observed that the extension includes many things we
'Objects' do not exist independently of conceptual schemes. We have not causally interacted with (e.g. future horses and rabbits,
another or horses and rabbits that never interacted with any human
cut up the world into objects when we introduce one or
scheme of description. Since the objects and the signs are alike being). When we use the word 'horse' we refer not only to the
internal to the scheme of description, it is possible to say what horses we have a real connection to, but also to all other things
matches what. of the same kind.
Indeed, it is trivial to say what any word refers to within the At this point, however, we must observe that 'of the same
language the word belongs to, by using the word itself. What kind' makes no sense apart from a categorial system which says

does 'rabbit' refer to? Why, to rabbits, of course! What does what properties do and what properties do not count as similar-
'extraterrestrial' refer to? To extraterrestrials (if there are any). ities. In some ways, after all, anything is 'of the same kind' as

Of course the externalist agrees that the extension of 'rabbit' anything else. This whole complicated story about how we refer
the set of rabbits and the extension of 'extraterrestrial' is
the to some things by virtue of the fact that they are connected with
is
not regard such statements us by 'causal chains of the appropriate kind', and to yet other
set of extraterrestrials. But he does
as telling us what reference is. For him finding out what refer- things by virtue of the fact that they are 'of the same kind' as

ence is, i.e. what the nature of the 'correspondence' between things connected with us by causal chains of the appropriate

words and things is, is a pressing problem. (How pressing, we kind, and to still other things 'by description', is not so much

saw in the previous chapter.) For me there is little to say about false as [Link] makes horses with which I have not inter-
what reference is within a conceptual system other than these acted 'of the same kind' as horses with which I have interacted

tautologies. The idea that causal connection is necessary is is that fact that the former as well as the latter are horses. The
refuted by the fact that 'extraterrestrial' certainly refers to extra- metaphysical formulation of the problem once again
realist

terrestrials whether we have ever causally interacted

with any makes it seem as if there are to begin with all these objects in
extraterrestrials or not! themselves, and then I get some kind of a lassoo over a few of

The externalist philosopher however, that we

would reply, these objects (the horses with which Ihave a 'real' connection,
can refer to extraterrestrials even though we have never inter- via a 'causal chain of the appropriate kind'), and then I have the
acted with any (as far as we know) because we have interacted problem of getting my word ('horse') to cover not only the ones
of the rela- I have 'lassooed' but also the ones I can't
with terrestrials and we have experienced instances lassoo, because they
from the same planet and instances of the property
as' are too far away in space and time, or whatever. And the 'solu-
tion 'not
'intelligent being'. And we can define an extraterrestrial
as an tion' to this pseudo-problem, as I consider it to be- the meta-
intelligent being that is not from the same planet as terrestrials. physical realist 'solution' - is to say that the word automatically
Also, 'not from the same planet as' can be analyzed in terms of covers not just the objects I lassooed, but also the objects which

'not from the same place as' and 'planet' (which can be
further are of the same kind - of the same kind in themselves. But then

analyzed). Thus the externalist gives up the requirement that we the world is, after all, being claimed to contain Self-Identifying
have some 'real' connection (e.g. causal connection) with every- Objects, for this is just what it means to say that the world, and
thing we are able to refer to, and requires only that the basic not thinkers, sorts things into kinds.
terms refer to kinds of things (and relations) that we have some In a sense, I would say, the world does consist of 'Self-Identi-
 Two philosophical perspectives 55
54 Two philosophical perspectives
part, its coherence and fit; coherence of 'theoretical' or less
fying Objects' -
but not a sense available to an externalist. If, as experiential beliefs with one another and with more experiential
I maintain, 'objects' themselves are as much made as discovered, beliefs, and also coherence of experiential beliefs with theoretical
as much products of our conceptual invention as of the 'objec- beliefs. Our conceptions of coherence and acceptability are, on
tive' factor in experience, the factor independent of our will, the view I shall develop, deeply interwoven with our psychology.
then of course objects intrinsically belong under certain labels; They depend upon our biology and our culture; they are by no
because those labels are the tools we used to construct a version means 'value free*. But they are our conceptions, and they are
of the world with such objects in the first place. But this kind of
conceptions of something real. They define a kind of objectivity,
'Self-Identifying Object' is not mind-independent; and the exter-
objectivity for us, even if it is not the metaphysical objectivity of
nalist wants to think of the world as consisting of objects that the God's Eye view. Objectivity and rationality humanly speak-
are at one and the same time mind-independent and Self-Identi- ing are what we have; they are better than nothing.
fying. This is what one cannot do.
To reject the idea that there is a coherent 'external' perspec-
tive, a theory which 'in itself, apart from all pos-
is simply true
Internalism and relativism sible observers, is not to identify truth with rational acceptabil-
ity. Truth cannot simply be rational acceptability for one
Internalism is not a 'Anything goes'.
facile relativism that says, fundamental reason; truth is supposed to be a property of a
Denying that it makes sense to ask whether our concepts 'match'
statement that cannot be lost, whereas justification can be lost.
uncontaminated by conceptualization is one
something totally The statement The earth is flat' was, very likely, rationally
hold that every conceptual system is therefore just
thing; but to acceptable 3,000 years ago; but it is not rationally acceptable
as good would be something else. If anyone really
as every other today. Yet it would be wrong to say that 'the earth is flat' was
believed that, they were foolish enough to pick a concep-
and if
true 3,000 years ago; for that would mean that the earth has
tual system that told them they could fly and to act
upon it by
changed its shape. In fact, rational acceptability is both tensed
jumping out of a window, they would, if they were lucky enough
and relative to a person. In addition, rational acceptability is a
weakness of the latter view at once. Internal-
to survive, see the matter of degree; truth is sometimes spoken of as a matter of
ism does not deny that there are experiential inputs to knowl- degree (e.g., we sometimes say, 'the earth is a sphere' is approx-
edge; knowledge is not a story with no constraints except inter- imately true); but the 'degree' here is the accuracy of the state-
nal coherence; but it does deny that there are anf inputs which ment, and not its degree of acceptability or justification.
are not themselves to some extent shaped by our concepts, by What this shows, in my opinion, not that the externalist
is
the vocabulary we use to report and describe them, or any inputs view is right after all, but that truth an idealization of rational
is
which admit of only one description, independent of all concep- acceptability. We speak as if there were such things as epistemi-
tual choices. Even our description of our own sensations,
so dear
cally ideal conditions, and we call a statement 'true' if it would
knowledge to generations of epistemolo-
as a starting point for be justified under such conditions. 'Epistemically ideal condi-
heavily affected (as are the sensations themselves, for
gists, is tions', of course, are like 'frictionless planes': we cannot really
that matter) by a host of conceptual choices. The very
inputs
attain epistemically ideal conditions, or even be absolutely cer-
upon which our knowledge is based are conceptually contami-
tain that we have come sufficiently close to them. But frictionless
nated; but contaminated inputs are better than none. If contam- planes cannot really be attained either, and yet talk of friction-
inated inputs are all we have, still all we have has proved to
be
less planes has 'cash value' because we can approximate them
quite a bit. to a very high degree of approximation.
What makes a statement, or a whole system of statements - a Perhaps it will seem that explaining truth in terms of justifi-
theory or conceptual scheme - rationally acceptable is, in large
56 Two philosophical perspectives Two philosophical perspectives 57

cation under ideal conditions is explaining a clear notion in tences of the reducing class, then that view is metaphysical realist

terms of a vague one. But 'true' is not so clear when we move at base. A truly non-realist view is non-realist all the way down.
away from such stock examples as 'Snow is white.' And in any The error is often made of regarding reductionist philosophers
case, I am not trying to give a formal definition of truth, but an as non-realists, but Dummett is surely right; their disagreement
informal elucidation of the notion. with other philosophers is over what there really is, and not over
The simile of frictionless planes aside, the two key ideas of the the conception of truth. If we avoid this error, then the claim I

idealization theory of truth are (1) that truth is independent of just made, that it is impossible to find a philosopher before Kant

justification here and now, but not independent of all justifica- who was not a metaphysical realist, at least about what he took
tion. To claim a statement is true is to claim it could be justified. to be basic or unreducible assertions, will seem much more
(2) truth is expected to be stable or 'convergent'; if both a state- plausible.

ment and its negation could be 'justified', even if conditions were The oldest form of the correspondence theory of truth, and
as ideal as one could hope to make them, there is no sense
in one which endured for approximately 2,000 years, is one that
thinking of the statement as having sl truth-value. 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
The 'similitude' theory
and the external objects that they refer to is literally a similarity.
The theory that truth is correspondence is certainly the natural The theory, like modern employed the idea of a men-
theories,
one. Before Kant it is perhaps impossible to find any philosopher tal representation. This presentation, the mind's image of the

who did not have a correspondence theory of truth. external thing, was called sl phantasm by Aristotle. The relation
Michael Dummett has recently 3 drawn a distinction between between the phantasm and the external object by virtue of which
non-realist (i.e. what I am calling 'internalist') views and reduc- the phantasm represents the external object to the mind is
tionist views in order to point out that reductionists can be (according to Aristotle) that the phantasm shares a form with
metaphysical realists, i.e. subscribers to the correspondence the- the external object. Since the phantasm and the external object
ory of truth. Reductionism, with respect to a class of assertions are similar (share the form), the mind, in having available the
(e.g. assertions about mental events) is the view that
assertions phantasm, also has directly available the very form of the exter-
in that class are 'made true' by facts which are outside of that nal object. 4
class. For example, facts about behavior are what 'make true' Aristotle himself says that the phantasm does not share with
assertions about mental events, according to one kind of reduc- the object such properties as redness (i.e. the redness in our
tionism. For another example, the view of Bishop Berkeley that minds is not same property as the redness of the
literally the

all there 'really is' is minds and their sensations is

reductionist, object), which can be perceived by one sense, but does share such
for it holds that sentences about tables and chairs and other properties as length or shape which can be perceived by more
ordinary 'material objects' are actually made true by facts about than one sense (which are 'common sensibles' as opposed to 'sin-
sensations. gle sensibles').
If a view reductionist with respect to assertions of one kind,
is In the seventeenth century the similitude theory began to be
but only to insist on the correspondence theory of truth for sen- restricted, much as it had been by Aristotle. Thus Locke and
Descartes held that in the case of a 'secondary' quality, such as
3 Dummett' s views are set out in 'What is a theory of Meaning I, II' in a color or a texture, it would be absurd to suppose that the prop-
Truth and Other Enigmas (Harvard, 1980). His forthcoming
(eventually) William James Lectures (given at Harvard in 1976) develop
4
them in much more detail. See De Anima, Book III, Ch. 7 and 8.
58 Two philosophical perspectives Two philosophical perspectives 59

erty of the mental image is literally the same property as the their sensations). It is generally unappreciated that the premiss
property of the physical thing. Locke was a Corpuscularian, that from which Berkeley worked - the similitude theory - was not
is, an advocate of the atomic theory of matter, and like a modern
something he merely learned from Locke (or read into Locke)
physicist he conceived that what answers to the sensuous pre- but was the accepted theory of reference before his time and,
sented redness of my image of a red piece of cloth is not a simple indeed, for a hundred years afterwards; but we have just
property of the cloth, but a very complex dispositional property remarked how venerable this theory actually was.
or 'Power', namely the Power to give rise to sensations of this Berkeley's argument is very simple. The usual philosophical
particular kind (sensations which exhibit 'subjective red', in the argument against the similitude theory in the case of secondary
language of psychophysics). This power in turn has an explana- qualities is correct (the argument from the relativity of percep-
tion, which we did not know in Locke's day, in the particular tion), but it goes just as well in the case of primary qualities. The
micro-structure of the piece of cloth which leads it to selectively length, shape, motion of an object are all perceived differently
absorb and reflect light of different wave-lengths. (This sort of by different perceivers and by the same perceiver on different
explanation was already given by Newton.) If we say that having occasions. To ask whether a table is the same length as my image
such a microstructure is 'being red' in the case of a piece of cloth, of it or the same length as your image of it is to ask an absurd
then clearly whatever the nature of subjective red may be, the question. If the table is three feet long, and I have a good clear
event in my mind (or even my brain) that takes place when I view of it, do I have a three foot long mental image? To ask the
have a sensation of subjective red does not involve anything in question is to see its senselessness. Mental images do not have a
my mind (or brain) 'being red'. The properties of a physical thing physical length. They cannot be compared with the standard
which make it an instance of physical red and the properties of measuring rod in Paris. Physical length and subjective length
a mental event which make it an instance of subjective red are must be as different as physical redness and subjective redness.
quite different. A red piece of cloth and a red after-image are not To state Berkeley's conclusion another way, Nothing can be
literally similar. They do not share a Form. similar to a sensation or image except another sensation or
For those properties (shape, motion, position) which his Cor- image. Given this, and given the (still unquestioned) assumption
puscularian philosophy led him to regard as basic and irreduci- that the mechanism of reference is similitude between our 'ideas'
ble, Locke was willing to keep the similitude theory of reference, (i.e. our images or 'phantasms') and what they represent, it at
however. (Actually, some Locke scholars today dispute this; but once follows that no 'idea' (mental image) can represent or refer
Locke does say that there is a 'similitude' between the idea and to anything but another image or sensation. Only phenomenal
the object in the case of the primary qualities and that there is objects can be thought about, conceived, referred to. And if you
'no similitude' between the idea of red or warmth and the red- can't think of something, you can't think it exists. Unless we
ness or warmth in the object. 5 And the reading of Locke I am treat talk of material objects as highly derived talk about regu-
describing was the universal one among his contemporaries and larities in our sensations, it is totally unintelligible.
among eighteenth century readers as well.) 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
Berkeley's tour de force
really exist (except as a construction from sensations), and not
Berkeley discovered a very unwelcome consequence of the to anything peculiar about his premisses. But the fact that one
similitude theory of reference: it implies that nothing exists could derive such an unacceptable conclusion from the simili-
except mental entities ('spirits and their ideas', i.e. minds and tude theory produced a crisis in philosophy. Philosophers who
did not wish to follow Berkeley in Subjective Idealism had to
5
See An Essay Concerning Human Understanding, Book II, Ch. VIII.
come up with a different account of reference.
60 Two philosophical perspectives Two philosophical perspectives 61

true of all qualities - the simple ones, the primary ones, the sec-
Kant's account of knowledge and truth ondary ones alike (indeed, there is little point of distinguishing
want to say that, although Kant never quite says that this is them). 6
I

what he is doing, Kant is best read as proposing for the first time If all properties are secondary, what follows? It follows that
what I have called the 'internalist' or 'internal realist' view of 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 say about
all we
truth.
To begin with, it is clear that Kant regarded Berkeley's Subjec- any object describes the object as it is 'in itself, independently of
unacceptable (this much he explicitly says), its effect on us, on beings with our rational natures and our
tive Idealism as quite
and also regarded causal realism - the view that we directly per- biological constitutions. It also follows that we cannot assume

ceive only sensations, and infer material objects via some kind
of any similarity ('similitude', in Locke's English) between our idea
problematical inference, as equally unacceptable. A view on of an object and whatever mind-independent reality may be ulti-

which only a very dubious hypothesis that there is a table in

it is
mately responsible for our experience of that object. Our ideas
front of me as I write these pages is a 'scandal', Kant says.
of objects are not copies of mind-independent things.

Secondly, I take it that Kant saw clearly how Berkeley's argu- This is very much the way Kant describes the situation. He

ment works: he saw that it depends on the similitude theory of does not doubt that there is some mind-independent reality; for
reference, and that rejecting Berkeley's argument requires reject- him this is virtually a postulate of reason. He refers to the ele-
ing that theory. Here I am attributing a view to Kant that Kant ments of this mind-independent reality in various terms: thing-
in-itself (Ding an sich); the noumenal objects or noumena; col-
does not express in these words (indeed, talk of 'reference' as the
relation between mental signs and what they stand for is very
lectively, the noumenal world. But we can form no real concep-
recent, although the problem of the relation between
mental tion of these noumenal things; even the notion of a noumenal
signs and what they stand for is very ancient). But we shall see world is a kind of limit of thought {Grenz-Begriff) rather than a

that what Kant did say has precisely the effect of giving up the clear concept. Today the notion of a noumenal world is per-
similitude theory of reference. ceived to be an unnecessary metaphysical element in Kant's

Let me suggest a of reading Kant that may be helpful,

way thought. (But perhaps Kant is right: perhaps we can't help think-

although it is only a first approximation to a right interpretation.

Think of Kant as accepting Berkeley's point that the argument Kant gives a summary of his own view in precisely this way in the
much to the so-called Prolegomena:
from the relativity of perception applies as
'primary' qualities as to the secondary ones, but making a differ- Long before Locke's time, but assuredly since him, it has been
generally assumed and granted without detriment to the actual
ent response than Berkeley made. Berkeley's response, recall, existence of external things that many of their predicates may be
was to scrap the distinction between primary qualities and sec- said to belong, not to the things in themselves, but to their

ondary qualities and fall back on just what Locke would have appearances, and to have no proper existence outside our repre-

we can sentation. Heat, color, and taste, for instance, are of this kind.
called 'simple' qualities of sensation as the basic entities
Now, if I go farther and, for weighty reasons, rank as mere
refer to. Locke's own treatment of secondary qualities, recall,
appearances the remaining qualities of bodies also, which are
was to say that (as properties of the physical object) only we can called primary - such as extension, place, and, in general, space,

- - with all that which belongs to it (impenetrability or materiality,

conceive of them as Powers, as properties nature unspecified
shape, etc.) - no one in the least can adduce the reason of its
which enable the object to affect us in a certain way. Saying that being inadmissible. As little as the man who admits colors not to
something is red, or warm, or furry, is saying that it is so-and-so 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
in relation to us, not how it is from a God's Eye point of view. little can my thesis be named idealistic merely because I find that

I suggest that (as a first approximation) the way to

read Kant more, nay, all the properties which constitute the intuition of a
is as saying that what Locke said about secondary qualities is body belong merely to its appearance.
 Two philosophical perspectives Two philosophical perspectives 63
62

'external' [Link] sensations I call 'red' can no more be

ing that there is somehow a mind- independent 'ground' for our
lead at once to non-
directlycompared with noumenal objects to see if they have the
experience even if attempts to talk about it
same noumenal property than the objects I call 'pieces of gold'
sense.)
not can be directly compared with noumenal objects to see if they
At the same time, talk of ordinary 'empirical' objects is
have the same noumenal property.
talk of things-in-themselves but only talk of things-for-us.
The reason that 'All properties are secondary' is only a first
The really subtle point is that Kant regards all of these points
approximation to Kant's view is this: 'All properties are secon-
as applying to sensations ('objects of internal sense') as well as
dary'(i.e. all properties are Powers) suggests that saying of a
to external objects. This may seem strange: what is the problem
about whether or not an idea corresponds to a sensation? But
chair that it is madeof pine, or whatever, is attributing a Power

Kant is on to something profound.

(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
Suppose I have a sensation £. Suppose I describe E; say, by
asserting 'E is a sensation of red,
9
If 'red' just means like this,
different Power to that same noumenal object; and so on. On
*£ is like this (attending to 9 such a view there would be one noumenal object corresponding
then the whole assertion just means
to each object in what Kant calls 'the representation', i.e. one
£), i.e. E is like E - and no judgment has really been made. As
one reduced to virtually a grunt. On the
noumenal object corresponding to each thing-for-us. But Kant
Wittgenstein puts it, is
explicitly denies this. This is the point at which he all but says
other hand, if 'red' is a true classifier, if I am claiming that this
that he is giving up the correspondence theory of truth.
sensation E belongs in the same class as sensations I call 'red' at
Kant does not, indeed, say he is giving up the correspondence
other times, then my judgment goes beyond what is immediately
given, beyond the 'bare thatness', and involves an implicit refer-
theory of [Link] the contrary, he says that truth is the 'cor-
respondence of a judgment to its object'. But this is what Kant
ence to other sensations, which I am, not having at the present
called a 'nominal definition of truth'. On my view, identifying
instant, and to time (which, according to Kant, is not something
thiswith what the metaphysical realist means by 'the correspon-
noumenal but rather a form in which we arrange the 'things-for-
7 dence theory of truth' would be a grave error. To say whether
us'). Whether the sensations I have at different times that I clas-
a
Kant held what a metaphysical realist means by 'the correspon-
sify as sensations of red are all 'really' (noumenally) similar is
dence theory of truth' we have to see whether he had a realist
question that makes no sense; if they appear to be similar (e.g. if
conception of what he called 'the object' of an empirical judg-
I remember the previous sensations as similar to this one, and
ment.
anticipate that future sensations which I will so classify will in
seem to be similar to this one, as this one is then remem-
On Kant's view, any judgment about external or internal
their turn
objects (physical things or mental entities) says that the nou-
bered) then they are similar-for-me.
words, that the
menal world as a whole is such that this is the description that a
Kant says again and again, and in different
rational being (one with our rational nature) given the informa-
objects of inner sense are'wof transcendentally real (noumenal)
tion available to a being with our sense organs (a being with our
that they are 'transcendentally ideal' (things-for-us), and that
they are no more and no less directly knowable than so-called 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
Here I am being deliberately anachronistic and describing Kant's view by
correspondingly noumenal chairs and noumenal horses and nou-
means of an example taken from Wittgenstein's Phibsophical
Investigations. But Wittgenstein's example has deeply Kantian roots: menal sensations. There is not even a one-to-one correspondence
Hegel, writing shortly after Kant, and aware of Kant's doctrine, made between things-for-us and things in themselves. Kant not only
precisely the point that any judgment, even of sense impression, has to
gives up any notion of similitude between our ideas and the
go beyond what is 'given' to be a judgment at all.
64 Two philosophical perspectives Two philosophical perspectives 65

things in themselves; he even gives up any notion of an abstract occurring right now and the quarter's being in my pocket right
isomorphism. And this means that there is no correspondence now) are effects of my past actions; if I had not sat down to type,
theory of truth in his philosophy. I would not be having the sensation; and the quarter would/ not

What then is we have

a true judgment? Kant does believe that be in my pocket if I had not put it there. Both the sensation and
objective knowledge: we know laws of mathematics, laws of the quarter exist in the twentieth century. Both the sensation and
geometry, laws of physics, and many statements about individ- the quarter have been described in English. And so on and so on.
ual objects - empirical objects, things for us. The use of the term The number of similarities one can find between any two objects
'knowledge' and the use of the term 'objective' amount to the is limited only by ingenuity and time.

assertion that there still is a notion of truth. But what is truth if In a particular context, 'similarity' may have a more restricted
it is not correspondence to the way things are in themselves? meaning, of course. But to just ask 'are A and B similar?' when
As I have said, the only answer that one can extract from we have not specified, explicitly or implicity, what kind of simi-
Kant's writing is this: a piece of knowledge (i.e. a 'true state- larity is at issue, is to ask an empty question.

ment') is a statement that a rational being would accept on suf- From this simple fact it already follows that the idea that
ficient experience of the kind that it is actually possible for similitude is the privatemechanism of reference must lead to an
beings with our nature to have. 'Truth' in any other sense is infinite regress. Suppose, to use an example due to Wittgenstein,
inaccessible to us and inconceivable by us. Truth is ultimate someone is trying to invent a 'private language', a language
goodness of fit. 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
The empiricist alternative
are qualitatively identical with X. In effect, he intends that E
So far as our argument has gone, it is still possible for a philos- should apply to all and only those entities which are similar to
opher to avoid giving up the correspondence theory of truth X.
and the similitude theory of reference by restricting them to sen- If this is all he intends - if he does not specify the respect in
sations and images. And many philosophers continued to believe which something has to be similar to X to fall under the classi-
even after Kant that similitude is the mechanism by which we fication E-then his intention is empty, as we just saw. For
are able to have ideas that refer to our own (and, although this everything is similar to X in some respect.
was more controversial, other people's) sensations, and that this If, on the other hand, he specifies the respect; if he thinks the

is the primary case of reference from an epistemological point of thought that a sensation is E if and only if it is similar to X in
view. respect R; then, since he is able to think this thought, he is
To see why this doesn't work, recall that the heart of Berke- already able to refer to the sensations for which he is trying to
ley's argument was the contention that nothing can resemble an introduce a term E, and to the relevant property of those sensa-
'idea' (sensation or image) except another 'idea', i.e. there can be tions!But how did he get to be able to do this? (If we answer,
no resemblance between the mental and the physical. Our ideas 'By focussing his attention on two other sensations, Z, W, and
can resemble other mental entities, but they cannot resemble thinking the thought that two sensations are similar in respect R
'matter', according to Berkeley. if and only if they are similar to Z, W\ then we are involved in
At this point, we must stop and realize that this is in an impor- a regress to infinity.)
tant way false. In fact, everything is similar to everything else in The difficulty with the similitude theory of reference is the
infinitely many respects. For example, my sensation of a type- same as the difficulty with the 'causal chain of the appropriate
writer at this instant and the quarter in my pocket are both sim- kind' theory that we mentioned earlier. If I just say, The word
ilar in the respect that some of their properties (the sensation's "horse" refers to objects which have the property whose occur-
66 Two philosophical perspectives Two philosophical perspectives 67

rence causes me on certain occasions to produce the utterance fails, of course, because any two things Z, W are themselves sim-
"there is a horse in front of me" ', then one difficulty is that there ilar in more than one way (in fact, in infinitely many ways).
are too many such properties. For example, let H-A (for 'Horse Trying to specifiy a similarity relation by giving finitely many
Appearance') be that property of total perceptual situations examples is like trying to specify a function on the natural num-
which elicits the response 'there is a horse in front of me' from a bers by giving its first 1,000 (or 1,000,000) values: there are

competent normal speaker of English. Then the property H-A is always infinitely many functions which agree with any given
present when I say 'there is a horse in front of me' (even when I table on any finite set of values, but which diverge on values not

am experiencing an illusion), but 'horse' does not refer to situa- listed in the table.

tions with that property, but rather to certain animals. The pres- This is connected with another point that Wittgenstein makes
ence of an animal with the property of belonging to a particular in Philosophical Investigations and that was mentioned at the
natural kind and the presence of a perceptual situation with the end of Chapter 1. Whatever introspectible signs or 'presenta-
property H-A are both connected to my utterance There is a tions' I may be able to call up in connection with a concept can-
horse in front of me' by causal chains. In fact, the occurrence of not specify or constitute the content of the concept. Wittgenstein
horses in the Stone Age is connected with my utterance 'There is makes this point in a famous section which concerns 'following
a horse in front of me' by a causal chain. Just as there are too a rule' - Even if two species in two pos-
say, the rule 'add one'.

many similarities for reference to be merely a matter of similari- sible worlds argument in most un-Wittgensteinian
(I state the

ties, so there are too many causal chains for reference to be terminology!) have the same mental signs in connection with the
merely a matter of causal chains. verbal formula 'add one', it is still possible that their practice

On the other hand, if I say 'the word "horse" refers to objects might diverge; and it is the practice that fixes the interpretation:
which have a property which is connected with my production signs do not interpret themselves, as we saw. Even if someone

of the utterance "There is a horse in front of me" on certain pictures the relation 'A is the successor of B f
(i.e. A =B + 1) just

occasions by a causal chain of the appropriate type', then I have as we do and has agreed with us on some large finite set of cases
the problem that, if I am able to specify what is the appropriate (e.g. that 2 is the successor of 1, 3 is the successor of 2, . . .,

type of causal chain, I must already be able to refer to the kinds 999,978 is he may have a diver-
the successor of 999,977), still

of things and properties that make up that kind of causal chain. gent interpretation of 'successor' which will only reveal itself in
But how did I get to be able to do this? some future cases. (Even if he agrees with us in his 'theory' - i.e.
The conclusion is no terms which have the
not that there are what he says about 'successor of; he may have a divergent inter-
logic ascribed by the similitude theory, any more than the con- pretation of the whole theory, as the Skolem-Lowenheim
clusion is that there are no terms which refer to things which are Theorem shows.)
connected to us by particular kinds of causal chains. The conclu- This has immediate relevance to philosophy of mathematics,
sion is simply that neither similitude nor causal connection can as well as to philosophy of language. First of all, there is the

be the only, or the fundamental, mechanism of reference. 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

Wittgenstein on 'following a rule'
of our practice, then there are possible divergent interpretations
Consider the example I mentioned in passing, of the man who of even the natural number sequence - our practice, or our men-
attempts to specify the respect R (the respect in which sensations tal representations, etc., do not single out a unique 'standard

must be similar to X
if they are to be correctly classified as E) by model' of the natural number sequence. We are tempted to think
saying or thinking that two things are similar in the respect R they do because we easily shift from 'we could go on counting'
just in case they are similar in just the way Z, W are similar. This to 'an ideal machine could go on counting' (or, 'an ideal mind
68 Two philosophical perspectives Two philosophical perspectives 69

could go on counting'); but talk of ideal machines (or minds) is prevent such a 'misunderstanding'? Asking him to make
very different from talk of actual machines and persons. Talk of no use of prior experience and to avoid all concept-
what an ideal machine could do is talk within mathematics, it ualization will/obviously leave him speechless; for to
cannot fix the interpretation of mathematics. talk at all he must use words.
same way, Wittgenstein holds that talk of 'similarity'
In the
and same sensation' or 'the same experience' is talk within
'the
Grasp of 'Forms' and empirical association
psychological theory; it cannot fix the interpretation of psycho-
logical theory. That, the interpretation of psychological theory A Platonist or Neo-Platonist of an antique vintage would have
and terminology, is fixed by our actual practice, our actual stan- dealt with this issue in a much simpler way. Such a philosopher
dards of correctness and incorrectness. would have said that when we attend to a particular sensation
In Ways of Worldtnaking 8 Nelson Goodman makes a closely we also perceive a Universal or aForm, i.e. the mind has the
related point: it is futile to try to have a notion of what the per- ability tograsp properties in themselves, and not just to attend
ceptual facts 'really are' independently of how we conceptualize to instances of those properties. Such a philosopher would say it
them, of the descriptions that we give of them and that seem is the Nominalism of Wittgenstein and Goodman, their refusal

right to us. Thus, after discussing a finding by the psychologist to have any truck with Forms and with the direct grasp of
Kolers that a disproportionate number of engineers and physi- Forms, that makes it seem to them that there is any problem
cians are unable to see apparent motion at all, that is 'motion' with the similitude theory.
produced by lights which successively flash at different positions, While just positing a mysterious power of 'grasping Forms' is
Goodman comments (p. 92): hardly a solution, it might seem that an analogue of this power
is available to us. Properties of things do enter into causal expla-
Yet if an observer reports that he sees two distinct flashes,
nations; when I have a sensation and it elicits the response 'this
even at distances and intervals so short that most
is a sensation of red', my response is partly caused by the fact
observers see one moving spot, perhaps he means that he
that the sensation had a property. True, some philosophers are
sees the two as we might
we see a swarm of molecules
say
so nominalistic that they would deny the existence of such enti-
when we look at a chair, or as we do when we say we
ties as 'properties' altogether; but science itself does not hesitate
see a round table top even when we look at it from
to talk freely of properties. Can we not say that, when Wittgen-
an oblique angle. Since an observer can become adept at
stein's privateer (the man who wanted to invent a private lan-
distinguishing apparent from real motion, he may take
the appearance of motion as a sign that there are two
guage) attended to X and said (
E' then what caused the response
(
E* was a causal interaction involving a certain property, and
flashes, as we take the oval appearance of the table top as
that property (whatever it was) is the relevant 'similarity' that
a sign that it is round; and in both cases the signs may
be or become so transparent that we look through them
other sensations must have to X to be correctly classified as E?
The observation that talk of 'properties' is perfectly scientifi-
to physical eventsand objects. When the observer visually
cally legitimate is correct; but thisdoes not help rehabilitate Pla-
determines that what is before him is what we agree is
before him, we can hardly charge him with an error
tonism. We interact with properties only by interacting with
their instances;and these instances always are instances oimany
in visual perception. Shall we say, rather, that he mis-
properties at thesame time. There is no such thing as just inter-
understands the instruction, which is presumably just
acting with a property 'in itself. Talk of the properties causally
to tell what he sees? Then how, without prejudicing
associated with a sensation cannot do the work that the notion
the outcome, can we so reframe the instruction as to
of the (unique) Form of the sensation did in Platonistic philoso-
8
Published by Hackett, 1978. phy.
70 Two philosophical perspectives Two philosophical perspectives 71

when have a sensation of blue, I have a

To spell this out: I verbal processing unit of my brain), the point is that, just as the
sensation of blue, and I also have a sensation with the complex property A-H described a few pages back will elicit the report
property of being such as to be classed by me at that instance 'There is a horse in front of me' even on occasions when no horse
under that particular verbal label. Merely attending to this sen- is present in the environment, so there is a complex property of
sation does not constitute 'grasping'one of these properties. To my total mind-set which will elicit 'I am having a sensation of
pick out the property associated in just one of these ways with blue', when I am not having the sensation of blue (or, anyway,

my sensation or with the verbal label is our old friend, the prob- would deny that I was if I were queried). No mechanism of
lem of the Causal Chain of the Appropriate Type again. empirical association is perfect. If we decide to stipulate that I

To see this, observe, first of all, that when my total perceptual am having a sensation of blue whenever I am having a sensation
experience elicits 'I am
having the sensation of
the response which elicits that report (or which elicits that report and is such
blue', I am not always right. I myself have had the experience of that the report does not seem 'wrong' to me on second thought),
referring to 'the man in the blue sweater' two or three times then on folk psychological theory, and perhaps on scientific psy-
before someone pointed out that the sweater was green, I don't chological theory as well, there could be occasions when it will
mean the sweater looked blue; I realized that I had been misde- be true that am
having a sensation of blue by this criterion
I

scribing the sweater the instant the other person spoke. (I don't although, for one of a variety of reasons, the quality of the sen-
often have occasion to say 'I am having the sensation of blue', sation is not blue. Moreover, as Wittgenstein puts it, on such a
but if I did, then in such a case I two
would probably have said it criterion, whatever seems right to me is going to be right - i.e.

or three times until someone - wondering, perhaps, why I would the distinction between making a report of my sensation that
have the sensation of blue when I was looking at something that really is correct and making a report that seems to me to be
was obviously green - queried me, whereupon I would have correct will have been abandoned. Perhaps we should abandon
taken back my previous phenomenal report.) This already shows or at least qualify perhaps, as Goodman seems to be suggest-
it;

that the property of eliciting the report having the sensa- 'I am ing, the question of whether one is 'really' having the kind of
tion of blue', or whatever, is not the same property as the prop- sensation one thinks one is makes no sense, apart from special
erty of being a sensation of blue, or a sensation of whatever the cases, such as the case in which one would take the report back
relevant quality might be. if queried; but to abandon this distinction is not a possible move

Philosophers often refer to such a case as a 'slip of the tongue'. for a metaphysical realist, for the sharp distinction between what
This seems to me to be an unfortunate terminology. The word really is the case and what one judges to be the case is precisely
'green' might have been on my lips, and I might have found what constitutes metaphysical realism.
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
Could one always be wrong about the quality of one's past
was misdescribing until someone questioned my report (and
sensations?
might never have noticed otherwise).
Another explanation which is suggested is that when I said Another way to bring out what is involved is to consider the
'blue' I meant green. By now it should be clear that when we say question: 'Could one always be wrong about one's past sensa-
things we don't go around 'meaning' things in the sense of hold- tions?' On the similitude theory, the answer is clearly 'yes'. For
ing meanings in mind. To say I 'meant' green is just to say that according to that theory, my previous sensations either are or
I instantly accepted the correction (and felt funny when I realized aren't similar to the sensations I now describe by the various
the way I had been speaking). This is just to repeat what hap- verbal labels 'sensation of red', 'pain', etc., and whether they are
pened, not to explain it. or aren't is a totally different question from whether I then class-
Whatever the explanation may be (perhaps some slip-up in the ified them under those same verbal labels. Perhaps the world is
72 Two philosophical perspectives Two philosophical perspectives 73

such that what we call a 'sensation of red' at an even numbered that too many correspondences exist. To pick out just one cor-
minute from the beginning of the Christian Era is actually simi- respondence between words or mental signs and mind-
lar in quality to what we call a 'sensation of green' at an odd- independent things we would have already to have referential
numbered minute, but our memory always deceives us in such a access to the mind-independent things. You can't single out a

way that we never notice. Then the sensation I classified under correspondence between two things by just squeezing one of
the verbal label 'sensation of red' one minute ago would not be them hard (or doing anything else to just one of them); you can-
similar to the sensation I now classify under that same label. not single out a correspondence between our concepts and the
There is something very odd about this alleged possibility, supposed noumenal objects without access to the noumenal
however. For one thing, the sense in which 'I would never notice' objects.

is very strong: if I treat my 'sensations of red' at different times One way to see this is the following. Sometimes incompatible
as reliable signs of the various correlated physical occurrences theories can actually be intertranslatable. For example, if New-
(such as fire, the signal to stop, etc.) then I will be successful in tonian physics were true, then every single physical event could
all my actions. The 'wrong' similarity class (the class that lumps be described in two ways: in terms of particles acting at a dis-
together the sensations I call sensations of red, in spite of the fact tance, across empty space (which is how Newton described grav-

that they are not 'really' all of the same 'quality') would be the itation as acting), or in terms of particles acting on fields which
one that I had better use in connection with my problem-solving act on other fields (or other parts of the same field), which finally
activities. But then is it really the wrong similarity class? act 'locally' on other particles. For example, the Maxwell equa-

we don't suppose that the notion of similarity is self-

If tions, which describe the behavior of the electro-magnetic field,

interpreting, then this case could be redescribed as a case in are mathematically equivalent to a theory in which there are

which the relation called 'similarity' by the external observer only action-at-a-distance forces between particles, attracting and
who is telling us about the case simply differs from the relation repelling according to the inverse square law, travelling not

called 'similarity' by us. If we take this view, then the hypothesis instantaneously but rather at the speed of light ('retarded poten-
that we are 'really' wrong about our past sensations collapses: tials'). The Maxwell field theory and the retarded potential the-

from an internalist point of view there is no intelligible notion of ory are incompatible from a metaphysical point of view, since
sensations at different times being 'similar' apart from our stan- either there are or there aren't causal agencies (the 'fields') which
dards of rational acceptability. 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
The correspondence theory of truth again
which makes one of them true, then one can define another cor-
By now the reader may be convinced that the similitude theory respondence which makes the other theory true. If all it takes to
of reference is thoroughly dead. But why should we conclude make a theory true is abstract correspondence (never mind
that the correspondence theory of truth must be given up? Even which), then incompatible theories can be true.
if the notion of a 'similarity' between our concepts and what To an internalist this is not objectionable: why should there
they refer to doesn't work, couldn't there be some kind of an not sometimes be equally coherent but incompatible conceptual
abstract isomorphism, or, if not literally an isomorphism, some schemes which fit our experiential beliefs equally well? If truth
kind of abstract mapping of concepts onto things in the (mind- is not (unique) correspondence then the possibility of a certain

independent) world? Couldn't truth be defined in terms of such pluralism i^ opened up. But the motive of the metaphysical real-
an isomorphism or mapping? ist is to save the notion of the God's Eye Point of View, i.e. the

The trouble with this suggestion is not that correspondences One True Theory.
between words or concepts and other entities don't exist, but Not only may there be correspondence between objects and
74 Two philosophical perspectives

(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
out a correspondence between two domains
fact that to single
one needs some independent access to both domains. Mind and body
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 unfulfillable, thought that it Parallelism, interactionism, identity

was nonetheless built into our rational nature itself (he suggested In the seventeenth century the great philosophers Descartes, Spi-
sublimating this 'totalizing' impulse in the project of trying to
noza, and Leibniz all realized that there was a serious problem
realize 'the highest good by reconciling the moral
in the world'
about the relation of mind to material body. To some extent, the
and empirical orders in a perfected system of social institutions relation was already a problem for Plato, of course, and for all
and individual relationships). The continued presence of this of the philosophers that came after; but it became much more of
natural but unfulfillable impulse is, perhaps, a deep cause of the
a problem with the rise of modern physics. In the seventeenth
false monisms and false dualisms which proliferate in our cul-
century, people became aware that the physical world is strik-
ture; be this as it may, we are left without the God's Eye View.
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
16 Mind and body Mind and body 11

way of thinking, was like a gas with just a little bit of push. As soon as 'spirit' is
influenced the sufferers of the illness.) In such a
dropped out, and the mind is really thought of as totally imma-
it is not so surprising that mind can 'influence' body.
terial, then the push of the mind on even very ethereal matter in
In the time of the philosophers I mentioned, the mathematical

was beginning to appear and to push aside this the pineal gland appears very strange. One can't quite visualize
way of thinking
that.
older way of thinking. The new way of thinking did not fully
The most naive version of the interactionist view conceives of
develop until Newton, but in special cases Descartes already had
the parallelogram of forces, and in still more primitive cases
the mind as a sort of ghost, capable of inhabiting different bodies
(but without change in the way it thinks, feels, remembers, and
Leonardo da Vinci already had it. These thinkers saw that phys-
exhibits personality, judging from the spate of popular books
ics could be done in something like the way it is done now. They
about reincarnation and 'remembering previous lives') or even
saw that what physics deals with is force and motion, and they
capable of existing without a body (and continuing to think, feel,
rejected the qualitative style of explanation. Rather, they con-
had a logic of its own, had a remember, and exhibit personality). This version, which
ceived that the mechanical world
we would and that it followed that program amounts to little more than superstition, is vulnerable to the
'program', as say,
objection that there is enormous evidence (some of which was
unless something disturbed it.

It seemed to these thinkers that mental events could do one of already known in the seventeenth century) that the functions of

two They could parallel physical events, e.g. events in

things. (1)
thought, feeling, and memory involve the brain in an essential
way. Indeed, on such a version it is not clear why we should
the brain. The model is a pair of synchronized clocks: the body
have complicated brains at all. If all that is needed is a 'steering
is a clock which has been wound up and which runs its happy or
wheel', that could be a lot smaller than the human brain.
unhappy way until death, and likewise the entire physical world
runs its happy or unhappy way from creation to the Last Judg-
To avoid such scientific objections, sophisticated interaction-
such as Descartes maintained that the mind and the brain are
ment (or to gravitational collapse, in a modern version). And the ists

mental events run their happy or unhappy way, and somehow,

an essential unity. In some way it is the mind-brain unity that
thinks, feels, remembers, and exhibits personality. This means
perhaps by divine providence, it has been arranged so that brain
that what we ordinarily call the mind is not the mind at all, but
event B will always occur just when sensation S is occurring. (2)
the mind-brain unity. What this doctrine means, what it means
They could interact with physical events. The mental events
to say that something can consist of two substances as different
might actually be causing brain events, and vice versa.
form of the interactionist view, the as mind and matter are supposed to be and still be an essential
Descartes' rather notorious
mind can influence matter when the matter unity, is, however, very obscure.
suggestion that the
is very, very ethereal (and that, in fact, it in some way pushes the The parallelist alternative is also very strange. What makes
matter in the pineal gland), was less the crazy speculation that it the mental event accompany the brain event? One daring
might seem to be, and more a hangover from a set of medieval seventeenth-century philosopher suggested that mental events

doctrines. 1 In the earlier way of thinking, mind was thought

the might actually be identical with brain events and other physical
which in turn acted on 'matter', and events, and that was Spinoza. The suggestion in a contemporary
of as acting on the 'spirit'

spirit was not thought of as totally immaterial. 'Spirit' was just form is that the event of my being in pain on a particular occa-

the in-between sort of stuff that the medieval philosophers' ten-

sion might be the same event as the event of my brain being in

dency to introduce in-betweens between any two adjacent terms 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-
in the series of kinds of being naturally led them to postulate. It
ular sort of pain and being in brain state B are identical. I prefer
to talk in this way because I we have more of a logical
think
1
See The Discarded Image, by C S. Lewis, especially Chapter VII, sec. F,
theory of properties at the present time than we do of events, but
for a description of the medieval view (Cambridge, 1964).
78 Mind and body Mind and body 79

Ithink the idea can be couched in either way. The idea, in this might be. A disembodied spirit might exhibit a certain program,
terminology, is that the property of the person, that the person a brain might exhibit a certain program, a machine might exhibit
isexperiencing sensation Q, could be the same property as the a certain program and the functional organization of these three,
property of being in brain state B.) In this form the suggestion the disembodied spirit, the brain, the machine, could be exactly
was put forward by Diderot, for example, in the eighteenth cen- the same even though their matter, their stuff, is totally different.
tury, and became 'mainstream' in the 1940s and 1950s. Materi- Psychological properties exhibit the same characteristic; the
alism and the identity theory began to be taken seriously for the same psychological property (e.g. being angry) can be a property
first time, and the suggestion began to be advanced that some-
of members of thousands of different species which may have
thing like Spinoza's view (or Spinoza's view minus its elaborate quite different physics and chemistry (some of these species
theological and metaphysical embellishments) is right: we are might be extraterrestrial; and perhaps robots will someday
really dealing with one world, and the fact that we do not know
exhibit anger). The suggestion of the functionalist is that the
until we do a great deal of science that the states of having pains,
most plausible 'monistic' theory in the twentieth century, the
hearing sounds, experiencing visual sensations, and so on, are in most plausible theory that avoids treating Mind and Matter as
reality brain states doesn't mean that they can't be.
two separate sorts of substance or two separate realms of prop-
The first contemporary form of this identity theory was erties, is that psychological properties are identical with func-
advanced by several writers, one of the best known being the tional properties.
Australian philosopher J. J. C. Smart. At first the suggestion was Today I am still inclined to think that that theory is right; or
that a sensation, say, a particular sensation of blue, is identical
at least that it is the right naturalistic description of the
with a certain neuro-physiological state. A variant on this, sug-
mind/body There are other, 'mentalistic', descriptions
relation.
2
gested by myself, I believe, is a view called functionalism.
first
of this relation which are also correct, but not reducible to the
On the functionalist view there is indeed an identity here, but
world-picture we call 'Nature' (indeed the notions of 'rational-
Smart was looking at the wrong sort of brain property to figure ity', 'truth', and belong to such a 'mentalistic' ver-
'reference'
as the other term in the identity. According to the functionalist,
sion). I shall say something about this later (Chapter 6). This fact
the brain has properties which are in a sense not physical.
does not dismay me: for, as Nelson Goodman has emphasized,
Now, what do I mean by saying that the brain has non- one of the attractive features of non-realism is that it allows the
physical properties? I mean properties which are definable in
possibility of alternative right versions of the world. I am, how-
terms that do not mention the brains physics or chemistry. If it
ever, attracted to the idea that one right version is a naturalistic
seems strange that a system which is physical should have prop- version, in which thought-forms, images, sensations, etc. are
erties which are not physical, consider a computing machine. A
functionally characterized physical occurrences; and what I wish
computing machine has many physical properties. It has a cer- to discuss here is a difficulty with the functionalist theory that
tain weight, for example; it has a certain number of circuit chips,
occurred to me some years ago: that is that the theory has diffi-
or whatever. It has economic properties, such as having a certain
culty with the qualitative character of sensations. When one
price; and it also has functional properties, such as having a cer-
thinks of relatively abstract pure psychological states, e.g. what
tain program. Now this last kind of property is non-physical in
we called a 'bracketed' belief, i.e. a thought considered only in
the sense that it can be realized by a system quite apart from
its 'notional' content, or of such diffuse emotional states as being
what its, as it were, metaphysical or ontological composition
jealous or being angry, then the identification of these with func-
tional states of thewhole system seems very plausible; but when
2
N. Block's Readings in Philosophy of Psychology (Harvard 1980)
contains an excellent collection of articles on Functionalism. My own one thinks of having a presented quality, e.g. experiencing a par-
papers are reprinted as Chapters 14 through 22 of my Mind, Language ticular shade of blue, the identification is implausible.
and Reality, Philosophical Papers, Vol 2 (Cambridge, 1975).
An example I have used for many years in lectures is a variant
 Mind and body Mind and body 81
80

case the sensation (or the corresponding physical event in the

of the famous example of the 'inverted spectrum'. The inverted
3 brain) has the role of signalling the presence of objective blue in
spectrum example (which appears in the writings of Locke )
blue looks the environment'. This theory captures one sense of the phrase
involves a chap who walks about seeing things so that
'sensation of blue', but not the desired 'qualitative' sense. If this
red to him and red looks blue to him (or so that his subjective
functional role were identical with the qualitative character, then
colors resemble the colors on a color negative rather than the
one couldn't say that the quality of the sensation has changed.
colors on a color positive). One's first reaction on hearing of
(If this is not clear, then imagine that after the spectrum inver-
such a case might be to say, 'Poor chap, people must pity him.'
sion, and after learning to compensate for it linguistically, you
But how would anyone ever know? When he sees anything blue,
experience an attack of amnesia which wipes out all memory of
itlooks red to him, but he's been taught to call that color blue
what colors used to look like. In this case it would seem as if the
ever since he was an infant, so that if one asked him what color
sensation you are now calling a 'sensation of blue' could have
the object is he would say 'blue'. So no one would ever know.
almost exactly the functional role that the sensation you used to
My variation was imagine your spectrum
the following:
call the 'sensation of blue' used to have, while having a totally
becomes inverted at a particular time in your life and you
different character.) But the quality has changed. The quality
remember what it was like before that. There is no epistemolog-
doesn't seem to be a functional state in this case.
ical problem about 'verification'. You wake up one morning and
It seems to me that the most plausible move for a functionalist
the sky looks red, and your red sweater appears to have turned
to make if such cases are really possible is to say, 'Yes, but the
blue, and all the faces are an awful color, as on a color negative.
"qualitative character" is just the physical realization.' And to
Oh my God! Now, perhaps you could learn to change your way
say that for this special kind of psychological property, for qual-
of talking, and to call things that look red to you 'blue', and
ities, the older form of the identity theory was the right one. If
perhaps you could get good enough so that if someone asked
the reader is fairly materialistically inclined, he or she probably
you what color someone's sweater was you would give the 'nor-
thinks that the property of having the sensation is a brain prop-
mal' answer. But at night, let us imagine that you would moan,
was erty. Readers who are not materialistically inclined probably
'Oh, wish the colors looked the way they did when
I
I a
think that the property of having the sensation is correlated with
child. colors just don't look the way they used to.'
The
a brain state. Probably most people hold one of these two views:
In this case, it seems that one even knows what must have
the view that sensation-states are correlated with brain-states, or
happened. Some 'wires' must have gotten 'crossed' in the brain.
the view that sensation-states are identical with brain-states. As
The inputs from blue light, that used to go to one mechanism in
so often happens, the question becomes debated over and over
the brain, now go to another, and the inputs from red light go to
in the same way. The way it is always discussed is, 'given that B
the first. In other words, something has switched around the
realizations, the physical states. The physical state that formerly
is correlated with Q, isB actually identical with Qf We know that
this sensation- state parallels this brain-state, is it or is it not the
played the functional role of signalling the presence of 'objective'
case that 'the sensation-state is identical with the brain-state?'
blue in the environment now signals the presence of 'objective'
The more the discussion goes on in that way, the more the
red in the environment.
concept of correlation comes to seem unproblematical. Correla-
Now suppose we adopted the following 'functionalist' theory
tion isn't (much) discussed because everyone knows that there is
of subjective color: 'a sensation is a sensation of blue (i.e. has
now describe in that way) just in
at least a correlation. Identity is discussed because that is what
the qualitative character that I
is problematical. But I am
going to try to show you that even
1

correlation is problematical, not in the sense that there is evi-

3 An Essay Concerning Human Understanding, Book II, Chapter 32 (sec.
dence of non-correlation, but in the epistemo logical sense that if
14).
82 Mind and body Mind and body 83

there is a correlation, one can never know which it is. The prob- truth. Quine pointed out that many things we thought we knew

lem will not depend on assuming materialism, but it will depend a priori have had to be revised. Thus, consider the following:
upon the fact that we think that there is at least a correlation. 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 .

Identity theory and the a priori

said that it was a necessary truth that this couldn't happen.
What made the revival of interest in the identity theory and other According to the physical theory we accept today, it does hap-
'monist' theories possible, if not initially (not with Smart and pen. Light passing near the sun behaves as does not because it

some of the early identity theorists), at least starting around the lighr travels in curved lines, but because the light continues
1960, was the change in the epistemological climate. Identity to travel in straight lines and the straight lines behave in that
theory was not taken seriously prior to the 1960s for the reason way in our non-Euclidean world.
that philosophers 'knew' it was false. And they thought they Once we accept that, then some philosopher was bound to ask
9
knew it was false not on the basis of empirical evidence (for what the question, 'What is left of the a priori? and Quine did. ,

sort of empirical evidence could show that a sensation-state is (Quine also showed convincingly that the standard empiricist
not a brain-state?), but a priori. One thinks about it and one just accounts oia priority — e.g. the notion of 'truth by convention' —
sees a priori that a sensation-state couldn't be a brain-state, or were incoherent, but I shall not review his arguments.)
perhaps one sees that it is meaningless to say that a sensation- In some ways, I think, Quine went too far. Quine's assertion
state is a brain-state in the way in which it is meaningless to say that 'no statement is immune from revision' suggests that for
that the number three is blue. Prior to 1950 or 1960 people every statement there are circumstances under which it would be
thought they just knew, or many people thought they knew, that rational to reject it. But this is pretty clearly false: under what
sensation-states can't be physical. Other people thought they circumstances, after all, would it be rational to reject 'Not every
knew those people were wrong. But argument was impossible. statement is true', i.e. to accept 'All statements are true'? 5
The majority would say, 'Look, we can't prove to you that it is But if Quine does overstate the case against the a priori, what
impossible for a sensation-state to be a neurophysiological state, he is nonetheless right about our notions of rationality
is this:

we can't prove to you that every number has a successor, we and of rational revisability are not fixed by some immutable
can't prove to you that the number three is not blue, but these book of rules, nor are they written into our transcendental
are things we just know; these are truths of reason. We know natures, as Kant thought, for the very good reason that the
that it is nonsense or an impossibility for a sensation-state to whole idea of a transcendental nature, a nature that we have
be a neuro-physiological state as clearly as we know anything.' noumenally, apart from any way in which we can conceive of
One had the majority that knew that sensation-states couldn't ourselves historically or biologically, is nonsensical. Since our
be brain-states and a minority that knew the majority was notions of rationality and of rational revisability are the product
wrong. Each knew the other was wrong a priori. And there was of our all too limited experience and all too fallible biology, it is
no really significant possibility of argument or movement from to be expected that even principles we regard as 'a priori', or
this frozen state in the debate. 'conceptual', or whatever, will from time to time turn out to
In 1951 W. V. Quine published a paper titled 'Two Dogmas need revision in the light of unexpected experiences or unantici-
of Empiricism'. 4 From that time on, there has been a steady ero- pated theoretical innovations. Such revision cannot be unlim-
sion in philosophical confidence in the notion of an 'a priori' ited: otherwise we would no longer have a concept of anything

4
Two Dogmas of Empiricism' first appeared in The Philosophical 5 I on the notion of the a priori in 'Analyticity and
discuss Quine's attack
Review, 1951. It is reprinted in Quine's From a Logical Point of View. Apriority:Beyond Wittgenstein and Quine', Midwest Studies in
(New York, 1961). Philosophy, Vol. IV, 1979 (Minnesota).
84 Mind and body Mind and body 85

we could call rationality; but the limits are not in general possi- erally) that the physical magnitude temperature is one and the
ble for us to state. Apart from trivial cases (e.g. 'Not every state- same physical magnitude as mean molecular kinetic energy. If

ment is true') we cannot be sure that it would never be rational this is right, then since 'x has such-and-such a temperature' is

in any context to give up a statement that is regarded (and legit- not synonymous with 'x has blah-blah mean molecular kinetic
imately so, in a given context) as a 'necessary' truth. In general, energy', even when 'blah-blah' is the value of molecular energy

we have to admit that considerations of simplicity, overall util- that corresponds to the value 'such-and-such' of the tempera-

ity, and plausibility may lead us to give up something that was ture, it must be that what the physicist means by 'physical mag-
nitude' is something quite other than what philosophers have
formerly regarded as a priori, and that this is reasonable. Philo-
sophy has become anti-aprioristic. But once we have recognized called a 'predicate' or a 'concept'.

that most of what we regard as a priori truth is of a contextual To be specific, the difference is that, whereas synonymy of the
and relative character, we have given up the only good 'argu- expression 'X is
3
P and S
X is Q' is required for the predicates P
ment' there was against mind—body identity. Identity theorists and Q to be the 'same', it is not required for the property P to
were bound to point this out, and they did. So there was a be the same as the property Q. Properties, as opposed to predi-
changed situation. cates,can be 'synthetically identical'.

I seems to me
have been using the notion of a property; but it
If there is such a thing as synthetic identity of properties, then
that there are at least two notions of a 'property' that have why shouldn't it be the case that the property of being in a cer-

become confused in our minds. 6 There is a very old notion for tain brain-state is the same property as the property of having a

which the term 'predicate' used to be employed (e.g. in the sensation of a certain qualitative character (very much in line

famous question, 'Is existence a predicate?'), and there is the with Spinoza's thinking) - even though it is not a conceptual
truth that fact, it seems to many to be a
even though, in
notion thatwe use today when we speak of 'physical properties', it is,

'fundamental magnitudes', etc. When a philosopher has the priori false? This argument that was made. In short, we
is the

older notion in mind, he frequently regards talk of properties as had a wave of anti-apriorism, we had the new machinery of the
interchangeable with talk of concepts. For such a philosopher, synthetic identity of properties, and with these two the identity
theorist and in particular the functionalist seem automatically to
properties cannot be thesame unless it is a conceptual truth that
they are the same; in particular, the property of having a sensa- be in business.

tion with a certain qualitative character cannot be the same as

the property of being in a certain brain-state, since the corre- Now, I want to consider what happens when to these two
sponding predicates are not synonymous (in the wide sense of things we add a third. What happens if a philosopher is (1) an
'analytically equivalent'),and the principle of individuation for anti-aprioristic naturalist, who (2) allows that there is such a

predicates is P is one and the same predicate as

just that being thing as the synthetic identity of properties, and (3) also has a

being Q just in case 'is P' is synonymous with 'is Q'. hard-line realist view of truth? I wish to claim that such a phi-
Consider, however, the situation which arises when a scientist losopher will find himself confronted with serious epistemologi-
asserts that temperature is mean molecular kinetic energy. On cal difficulties.

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- Split brains
perature is really (in some sense of 'really') the same property as
the property of having a certain molecular energy; or (more gen- 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
6 See *On Properties', Chapter 19 of my Mathematics, Matter and
Method, Philosophical Papers, Vol. 1 (Cambridge, 1975). relevance of this kind of experiment to the identity theory and to
86 Mind and body Mind and body 87

what has so far been taken for granted in the whole discussion, sion, or as having a conventional component, which brain-state
the notion that there is a correlation. Q is The position is that as a matter of fact we
identical with.
live in a world in which what we experience as the qualitative
On the model of the brain as a cognitive system resembling a
computer, the brain has a language, an internal language (which characters of sensations really are one and the same properties

may be innate, or which may be a mixture of an innate 'lan- as some of the properties that we encounter in other ways as

guage', or system of representation, and a public language). physical properties of brain events. (Or better put, in which the

Some philosophers have even invented a name for this hypothet- property of having a sensation of a certain qualitative character
ical brain language, 'mentalese'. Let us consider what happens is really just the property of being in a certain brain-state.)

when one has a visual sensation on such a model (and I shall Let us stop for a moment and see what the view actually says.
make up my neurology, since I don't know enough, but I don't Suppose that red is the subjective quality we're attending to (pro-
think anyone really knows enough). Here is one possible story: duced, say, by staring at a green disk and then removing the disk
When one has a sensation a 'judgment' is made; the brain has to get an afterimage). Suppose that when I experience this red,

12 o'clock'. So the the sensation-state I am in is identical to a disjunction of brain

to 'print' something like 'red presented at
quality (call it 'Q') corresponds, among other things, to a record states. It can't be identical to one maximally specified brain state,
in mentalese. Also, there is an input to the verbal processing cen- because we know away any one neuron, or
that one can take

ter, the center which is connected with the voice box, which whatever, and one can have the experience. But the property
still

accounts for the brain's ability to report in the public language, might be a disjunctive one, say (implausibly), the even numbered
'red now'. It may be that the judgment in mentalese has to be neurons in area blah-blah are firing or the prime numbered ones
transmitted from one location to another before there is an input are firing. Actually, it would be a much bigger disjunction than
to the speech center. There are also events in the visual cortex that. There would be a huge collection of neurological states

(which have been studied by the neurologists Hubel and Wiesel), such that their disjunction would be the property of experiencing
which I am imagining as on the road to the 'record in mentalese', red.

and the verbal process. These 'records', 'inputs', and other But now we go a little further. If the even numbered neurons
events may take place in different lobes of the brain: if the cor- in area blah-blah are firing, I experience red. If the cerebroscope

pus collosum is split, the person's right lobe (the lobe that says, 'no, the prime numbered neurons in area blah-blah are fir-

doesn't have speech) can see red (or at least it will affirmatively ing', I still experience red. That tell which of these
is, I can't

signal in response to a written query visible only to that lobe), brain-states I'm in. If I have to be in one of
experience red I

but if one asks the subject what color the card is, he will reply them. But I can't distinguish between them. The even numbered
'I can't see the card.' And, finally, there is at some point the neurons in area blah-blah are firing is not an observable prop-
formation of a memory trace or of memory traces (one could erty. Even with the knowledge that the identity theory is true, I

break this up into short-term memory and long-term memory). can't tell from my sensations that I have this property. Call this
There almost certainly is not a linear causal chain; there are property T/ and call the property that the odd numbered neu-
probably branchings and rejoinings, a causal network. rons are firing T 2\ The sensation-state is identical with the dis-

The problem is that psychology divides up mental events in a junction (P 1 or P 2 )> where this is, of course, a third property. P t
way. Here is a sensation of blue. Now it started;
fairly discrete is not a sensation-state, and P2 is not a sensation-state; it is only
now stopped. Causal networks are not discrete. There isn't a
it their disjunction that is a sensation-state. In other words, in this
unique physical event which is the correlate of the sensation. ontology, the disjunction of two properties which are themselves
If the identity theory then the sensation-state
is right, is Q unobservable can be observable. It is a complicated logical func-
identical with some brain-state or other. A metaphysical realist tion of unobservable properties that I experience as a simple

cannot regard it as in any way a matter of convention or deci- given. That is the position.
88 Mind and body Mind and body 89

It may have made the view sound silly. Thus, a friend

be that I physical properties of us, and either those properties are or
of mine has remarked, 'Suppose the only device we have for aren't identical to this sensation-state.

detecting muons doesn't distinguish between muons and anti- Similarly for a philosopher like Daniel Dennett, who thinks
muons. Then muon isn't an observable property, and antimuon that sensation talk is highly vague, who doesn't think there is a
isn't an observable property, but the disjunction of them is. This well defined subjective property of being in this sensation-state,
only seems to be paradoxical to those who take observationality of having a sensation with this qualitative character. I think he
to be less of a pragmatic notion than it is.' My purpose, however, too could adopt an identity theory as a meaning stipulation, fix-

is not to ridicule the view, which, indeed, constitutes a very ing not the meaning of the physical object terms this time, but
important and legitimate research program in neurophysiology, the meaning of the psychological terms. But again, that wouldn't
but to make clear what it commits one to. What leads to diffi- be the position of a full-blown metaphysical realist.

culties, I shall argue, is not the identity theory by itself but the I am considering a full-blown realist who thinks 'yes, I know
identity theory taken in conjunction with metaphysical realism — what this psychological property (the sensation-state) is. I've had
i.e. taken in conjunction with what I called the 'externalist' per- it. I can recognize it. I think it's a definite psychological property
spective on the nature of truth. to which I refer. I know what P and P 2 x are, therefore what {P x
One can avoid committing oneself to such a perspective. Thus, or P2 ) is, and either the sensation-state is identical to this or it

Carnap would have said (at least in a certain period) that talk isn't.' Just in the way that a naive physicist might say, 'there's no
about physical objects is highly derived talk about sensations, element of convention' (I think he'd be wrong by the way);
and that the decision to say that a particular brain state is iden- 'there's no element of convention in the decision that tempera-

tical with a sensation-state Q is really a decision to modify the ture is mean molecular kinetic energy, either temperature is

language of talk about physical properties in a certainway, to mean molecular kinetic energy, or it's some other property'.
change our concept of the physical property in question. That is the standpoint I want to examine.

Since physical object and physical property talk is only highly The problem is that if one takes this metaphysical realist

derived talk about sensations, we can modify the rules. But that standpoint, then there are many more possibilities than people
standpoint isn't the standpoint of metaphysical realism, at least are wont to consider. The possibility that first comes to mind is

with respect to material objects and physical properties. Some- that the sensation-state is identical with the property of having

body who thinks like that might be a metaphysical realist about the appropriate events take place in the visual cortex and having
sensations, but he is not a metaphysical realist about material the 'record' in 'mentalese' appropriately registered and having
objects, and since he regards material object talk as somewhat the input to the speech center and having the memory traces
soft, he can adopt the identity theory by simply saying 'I adopt formed - i.e. thought of as identical to the
the sensation-state is

it as a kind of convention, as a further meaning stipulation.' conjunction of these several properties. But as soon as we con-
Since the meanings were not totally fixed beforehand, since there sider the possibility of disassociation, then we become unsure
was some openness of texture, there is no problem about 'how that we really want the whole conjunction. Perhaps the sensation
can you know that the sensation-state is identical with this prop- is just the event in the visual cortex? (I.e. the property of having
erty and not some other?' If what this property is is somewhat the sensation is 'really' the property of having the event take
vague, then we're allowed to simply postulate the identity as a place in the visual cortex.)
meaning But I'm talking to someone who really
specification. Let us make the assumption that it is, for the moment. Now
thinks there is a material world out there, and it is not just highly let us suppose we
can cut off the process that produces the
derived talk about sensations; who really thinks that there are record in mentalese, or at least cut off the input to the speech
physical properties; and who holds that such expressions as 'the center. Let us imagine that we have shown the subject a red card

neurons in such and such a channel are firing' predicate definite on the left side of his visual field (so that the card is only 'visible
90 Mind and body Mind and body 91

to the right lobe', as neurologists say). The appropriate event in

you am told
telling I didn't.' (I that actually it is more customary
the visual cortex will then take place in the right lobe, but if we for patients to 'reconcile' or rationalize situations of this kind
say to the subject, 'Do you see anything red?', the subject will than to describe them as I have just imagined.) Would such a
say, 'No'.
report show that there was a sensation of red without there
Now, by one criterion we employ to decide whether or not being an input to the speech center?
someone has a sensation, the criterion of sincere verbal reports, It would not. If Daniel Dennett (who at one time held the view
we should have to say that he didn't have the sensation of red, that the sensation is the input to the speech center, or a view
and therefore that we have refuted the theory that (the rele- Q close to this 8 ) wished to reconcile this subject's report with his
vant qualitative character) is identical with the visual cortex theory, all he would have to say is, 'I don't deny that at the later
property in question. But someone could object, 'No, you time the psychological event of remembering having had the sen-
haven't refuted that theory at all. Because, what kind of an sation earlier took place. I deny that the sensation took place at
observer is this?The chap's brain is cut in two.' As far as any the earlier time.' On either theory the subject later has the expe-
observer in a normal condition can tell, Q is identical with this rience of remembering rightly or wrongly that he had the sensa-
property of the visual cortex. And observers who are not in a tion of red earlier.
normal condition don't count. They can't count. The disagreement here is an actual one. Most neurologists do
The difficulty is that there are identity theories which are believe that in the 'split brain' patients the right lobe is 'con-
observationally indistinguishable, 7 by which I mean that they scious'. In effect, this amounts to saying that there is sometimes
lead to thesame predictions with respect to the experience of all a sensation of red, or whatever, even though there is no input to
observers in a normal condition. the speech center. (There are two loci of consciousness', is the
Consider the view that one doesn't have the sensation of red way it is frequently put.) At least one famous neurologist, Eccles,
unless one has the input to the speech center. How could one holds, however, that the disassociated right lobe (or left lobe, in
prove that or refute One might think, if we split the corpus
it?
the case of patients who have the speech center on the right) is
colossum but there is some memory that doesn't go through the not conscious. There is a unitary consciousness on Eccles' view;
verbal processing unit, then there is a way; namely, we first ask that the disassociated right lobe can 'simulate' conscious behav-
the chap whether he has a sensation of red. He says, 'No'. Then ior does not show that it is a second 'locus of consciousness', he
we sew the corpus colossum back together (a neat trick if you would say.
can do it!), you have a sensation of red?' He might
and ask, 'Did Nor will it help to appeal to methodological maxims, e.g.
say, 'Yes, but ityou know I had this sensation of red
was crazy, 'choose the simpler theory'; for there does not appear to be any
and you asked me whether I had it, and I heard myself sincerely relevant kind of 'simplicity' which is possessed by the 'unitary'
view and lacked by the 'two loci' view, or possessed by the 'two
7
The notion of 'Observational Indistinguishability' was introduced in
loci' view and lacked by the 'unitary' view. Perhaps the 'two loci'
papers on space-time theory by Clark Glymour and David Malament in
Foundations of Space-Time Theories, Earman, Glymour, and Stachel view is simpler in one respect; it says that certain behavioral
(eds.), Minnesota Studies in the Philosophy of Science, Vol. VIII capacities (which the right lobe possesses, even if it does not pos-
{Minnesota University 1977). The analogous problem in space-time sess speech) are sufficient for consciousness, and this agrees with
theory is the existence of 'possible* space-times (i.e. space—times allowed
by which differ in their global topological properties,
relativity theory)
the fact that we call animals (who do not possess speech either)
but in which observers would have exactly the same experiences. Such conscious. But there are many dissimilarities between an animal
examples are often dismissed on the grounds that 'simplicity with an intact brain, whose brain processes are still 'integrated',
considerations' would tell one which space-time one is living in; the
trouble with this (as Malament points out) is that the physical theory 8
Dennett's model of consciousness is presented in 'Towards a Cognitive
(general relativity) doesn't say we live in the simplest space-time
theory of Consciousness', reprinted in his Brainstorms (Bradford Books,
compatible with its laws.
1978).
92 Mind and body Mind and body 93

even if they do not involve speech, and one piece of a 'split sensations of the other sex) I probably can't imagine, but that
brain'. If the case were not one which touches us so nearly, if we doesn't mean I regard the psychological space of those other
did not have such a strong tendency to metaphysical realism humans asunimaginably different from my own. Why shouldn't
about sensations, would it not be in keeping with our best meth- I think of the bat's visual field, for example, as very much like
odological intuitions to regard this as a case to be legislated my visual field? (N.B. Bats see very well, contrary to folklore.)

rather than fought over? Allowing for some adjustments for the optics of the bat eye, or
In short, there are a number of observationally indistinguish- the bat's hearing within the range that overlaps with mine, it's

able identity theories. If the identity theorist is right, it would like my hearing, and its pains are like my pains.' Now could we
seem that there is no way on earth in which one could know settle this?

which way he is right; know with which brain-state a given Because the number of neurons is different, and because the
sensation-state is identical (or correlated with). The point is suf- arrangement is different (the acoustic center of the bat's brain is

ficiently important to deserve further illustration. enlarged to become 7/8ths of the brain), the properties at the

Thomas Nagel9 has made the plausible claim that one cannot most completely specified neurological level - number of neu-
imagine what it would be like to be a bat. But why should this rons firing where - which are identical with a quale in the case
be a plausible claim? Some years ago I read a delightful book of a bat on the assumption that the identity theory is correct,
about bats by Donald Griffin. I came to realize that bats are not cannot literally be the same as the properties which are identical
from any other mammals. We mostly do think
basically different with any quale in the case of a human. Or can they? Suppose
we can imagine what sensations our dogs and cats have. What is that when a bat has a certain visual sensation (produced by
the difficulty with bats? seeing red objects), that the bat's brain has the disjunctive prop-
Well, bats can hear sounds several octaves higher than we can. erty (?! or P 2 ), where P and P 2 r are maximally specified states of
I cannot imagine what it would be like to be a bat in the sense of the bat's brain. (It would really be a much more complicated
imagining what the echolocation sensation would be. But need disjunctive property with thousands of cases, but let us sim-

this be so used to be able to hear sounds an octave

difficult? I plify.) And let us suppose that when I have a certain visual sen-
higher than the highest sounds I can hear in middle age. But the sation (produced by seeing red objects), my brain has the dis-

subjective pitches have not changed: the highest sounds I can junctive property (?/ or P 2 ')- Consider the following two
hear may be an octave lower than the highest sounds I could theories: (1) that the qualitative character of the bat's sensation

hear when I was ten years old in objective pitch, but the highest (call it, Weds) is identical with (or at least correlated with) the

sounds I hear now have the same thin, squeaky quality that the disjunctive property (P x or P 2 and ) the qualitative character of
(

sounds on the threshold of being too high to hear always did for the human sensation (call it, red H') is identical with (or at least

me. Perhaps that's how a sound five octaves higher than those correlated with) the different property (P/ or P 2 '). (2) That the
we hear sounds to a bat: like a short high squeak. qualitative character of the bat's sensation is identical with the

Now, imagine a debate between two philosophers or psychol- qualitative character of my sensation (i.e. red B —red H and both
)

ogists, one of whom says no bat quale is at all like any human are identical with (or correlated with) the more complex disjunc-
quale. Batqualia are unimaginably different from human qualia. tive property (P t or P2 or P t
'
or P 2 ').
You will never be able to imagine what it feels like to be a bat On the first theory the bat and I have different experiences,
(or even a dog or cat). The other philosopher, we may imagine, while on the second we have
same experience; but these two
the
replies, 'Nonsense! Perhaps there arc some bat sensations I can't theories lead to the same predictions with respect to what human
imagine. There are some sensations of other humans (e.g. some observers, normal and abnormal will experience. Once again,
they are observationally indistinguishable.
9 'What is it like to be a bat?*, reprinted in N. Block, op. cit. Will methodological maxims ('choose the simpler theory')
94 Mind and body Mind and body 95

help? Once again, it is not clear that they can. Ned Block has functional state mentioned. It does not show that it is not iden-

pointed out that the first theory is simpler in one respect (the ticalwith a more complicated functional state, such as the state
quale is identified with a simpler physical property in each case), of being in whatever material state earliest in your life realized
but the second is simpler in another respect (the second theory is the above functional state. One might object that this is a funny
'non-chauvinist'; it allows that one doesn't have to have exactly property, a complicated logical function of functional proper-
our physical constitution to have our qualia). And once again, ties. But why is a complicated logical function of functional
we lack principles for determining a unique preferred trade-off. properties less likely to be identical with red H than a disjunction
Indeed, what reason is there to think there should or must be of complicated physical properties? Does the world prefer dis-
such principles? Why should we not, as Wittgenstein urged we junctions of physical properties to conjunctions of functional
do, abandon our metaphysical realism about sensations and properties?
about 'same' (as applied to sensations), and treat this too as a Let us consider (2). Let P 3 be the property of being a rock, and
case to be legislated rather than fought over? consider the hypothesis that red H is identical with the disjunctive
Finally, I want to present three theories which I am sure are property (?! or P 2 or P/ or P 2 or P 3 ). Of course, rocks have this
'

false, but which it is difficult or impossible to rule out if meta- property all the time. So on this hypothesis, events of the quali-
physical realism is right. These are: (1) that red H is identical with tative character red H are taking place in rocks all the time. (They
a functional (or quasi-functional) state after all, namely the state are not experiencing red in the functional sense of experiencing
of being in whatever material (e.g. physical) state earliest in your red, but an event of the qualitative character of the event that
life played the functional role of normally signalling the presence plays the functional role of being the sensation of red in us is
of objective red. (2) that rocks have qualia (i.e. events qualita- taking place in them all the time.) Or consider more complicated
tively similar to, as it might be, visual sensations, take place in hypotheses on which rocks are having different qualia at differ-
rocks). (3) that nations are conscious. ent times. Or just the hypothesis that some one of these
Let us first consider (1). Recall the argument I used to show hypotheses (which not specified) is correct. We might say 'Well,
that red H could not be a functional state. That argument was these hypotheses are crazy.' Yes, they are. But each of them leads
that if we identified red H with the functional state of being in to the same predictions with respect to all human observers as
whatever material state (e.g. brain-state) normally signals the the 'sane' theory. None of them can be ruled out on observa-
presence of objective red, then I would not have undergone a tional or experimental grounds, because each of them is obser-
spectrum inversion (at least in the 'amnesia' case), since I am in vationally indistinguishable from the more standard view.
that functional state when I see something objectively red both We might think that these theories can be ruled out by an
before the spectrum inversion and after the spectrum inversion appeal to the methodological principle that one shouldn't attrib-
(allowing time for linguistic adjustment to take place, and, if ute a property to an object with no reason. Of course this prin-
necessary, postulating an attack of amnesia). But on a metaphys- ciple doesn't say these theories are false (sometimes things we
ical realist position it is certainly possible that I have undergone have no reason to believe are true), but at least it says we are
a spectrum inversion (even though I don't remember it because justified in taking them to be false. But is there really no reason
of the attack of amnesia). The case even stronger if I don't
is to hold the least specific of these theories (the theory that some
have an attack of amnesia and recall that my spectrum has been such theory is and rocks have qualia) ? What of the argu-
correct,
inverted; even in this case, if the linguistic adjustments have ment that if we have qualia and physicalism is true (and many
become automatic, there is a sense in which what used to be 'the philosophers think there are many good reasons for accepting
sensation of green' now plays the functional role of 'signalling physicalism), then there is at least one physical object in which
the presence of objective red in the environment'. events with a qualitative character take place: so why shouldn't
This argument only shows that red H is not identical with the such events take place in all physical objects? If we could show
96 Mind and body Mind and body 97

something about the quale which requires that which they pass notes to one another. (Their time would have to
that there is itself
pass very fast relative to 'our' time, of course.) They could be
it have the particular functional 'role' that it does in the case of
alienated workers. 'Now,' the reply continues, 'you wouldn't
humans then this move would be blocked; but this is just what
believers in qualia as metaphysically real objects tell us we can't
call that thing "conscious" because you know that it is really
do.
only these little people moving the body. And that shows that an
appropriate functional organization (one like ours) is not suffi-
Last but not least, let us consider (3). Consider the hypothesis
cient to justify the application of such predicates as
that pain is an appropriate functional state which
identical with
"conscious".'
can be exhibited by either organisms or nations. In other words,
One reply to this reply (the one I actually made) was to deny
suppose that when the United States announces that 'the United
.' it really is. We would, of course, never
that the 'hydra-headed robot' (as this last thing has been called)
States is pained by . .

know. Perhaps the reader is at this moment finding it interesting does have the same functional organization we do. But there is

and mildly amusing that a group can behave in ways which a more radical reply I might have made. I might have said, 'Why

resemble the ways in which something that really does feel pain 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-
behaves when it manifests its pain; but the reader does not think
that the United States really feels pain. On this hypothesis, the
tronic brain would be conscious, why would the fact that the

reader would be wrong: the national Geist would really be feel-

neurons of the hydra-headed robot are more conscious mean
ing pain.
that the whole thing is less conscious? After all, we are in a sense

This hypothesis connects with an interesting discussion in the a society of small animals. Our cells are in a sense individual
animals. And perhaps they have some little bit of feeling, who
philosophy of mind. An argument that functionalists (including
me) like to employ is the following 'anti-chauvinism' argument:
knows? Over and above our feeling.' Now, if we move that way,
in principle, the differencesbetween a robot and a human (in
if we decide that the hydra-headed robot is conscious (even
functional organization, anyway) could be reduced to small
though its neurons are boy scouts and girl scouts), then why not
the United States?
details of the physics and chemistry. One might even have a
I don't, of course, claim that the United States has the same
robot that corresponded to us down to the neuron level. (It could
functional organization as homo sapiens. Clearly it doesn't. But
even have a 'flesh and blood' body, apart from the brain.) The
difference would be that whereas we have neurons made of car-
there are many similarities. The United States has defensive
organs. It has ingesting organs, it eats oil and copper and so on.
bon and hydrogen and proteins and so on, it would have neu-
It excretes (pollution) in vast quantities. Is it not perhaps as sim-
rons made of electronics, but from the neuron level up all the
circuitry would exactly correspond. Now, unless you are a
ilar in functional organization to a mammal as is a wriggling fly,

'hydrogen— carbon chauvinist' who and

thinks that carbon
to which we do attribute pain?

hydrogen are intrinsically more conscious, why shouldn't you

say that this robot is a person whose brain happens to have more How well-defined is 'qualitative character'?
metal in it and less hydrogen and carbon?
This argument has provoked the following reply: 'Well, So far we have not questioned the idea that it is perfectly clear

instead of these electronic gadgets, electronic neurons wired

what it means to say that two of one's own sensations have or

together in the same circuits that human neurons are wired in,
do not have the same 'qualitative character'. Even at an intro-
spective level, this is not the case, however. For one thing, just
let us suppose you have miniature people, little girl scouts and
boy scouts.' We don't even have to imagine that these little peo- what one's experiences seem to one to be is notoriously depen-
ple even know what the whole scheme is for, or that they see
dent on antecedent conceptualization, as when we report seeing

anything except a dimly lit room, or a lot of dimly lit rooms, in a round table top even when we view it from an angle.
98 Mind and body Mind and body 99

In the case of the round table top, psychologists and philoso- sense-data 'flipped back'? Or has he gotten used to altered sense
phers have argued since the nineteenth century about whether data, and reinterpreted 'up' and 'down'? Very likely the subject
one has 'eliptical sense data' and thinks that they are round himself cannot say at what point things became normal or which
(unless one is a 'trained introspectionist') or has round 'Gestalts' of these things happened. (Readers who like me wear bifocals
and only thinks that they are elliptical because of optical theory. can ask themselves: does the lower half of the visual field look
One can have experiences which fit each description; and many different even when one isn't noticing the difference?) While
experiences will fit either description. Nor is neurology likely to there are transformations to which subjects never accommodate
settle this dispute: the elliptical image on the retina doubtless (in fact, it is only relatively simple changes to which one accom-
produces events in the brain itself, and if we identify these with modates), and I have assumed that one would not accommodate
the 'visual sensation', then we may well get something like the to a color-inversion, the phenomenon of accommodation cer-
unconscious infer-
classical story of 'elliptical sense-data plus tainly casts doubt on the extent to which 'same qualitative char-
ences'; the judged character of the experience ('I see a round acter' is a well-defined notion.
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
Realism about qualia
one judges that something looks elliptical. Why shouldwe not
say that these two versions are both legitimate? As Goodman We have considered a set of sceptical difficulties. What they pur-
says about the case of the subject who is asked to describe appar- port to show is not that the identity theory is wrong (or that the
ent motion, correlation theory is wrong - note that they can all be stated as
difficulties for a 'correlation' view just as much as for an identity
The best we
can do is to specify the sort of terms, the
then there are a vast number of alter-
view), but that, if it is true,
vocabulary, he is to use, telling him to describe what he
native ways of specifying the details such that one can never
sees in perceptual or phenomenal rather than physical
know which one of them is true. And not knowing which one of
terms. Whether or not this yields different responses, it
them is true means not knowing what the answer is to a great
casts an entirely different light on what is happening.
many traditional sceptical questions, such aswhether rocks and
That the instrument to be used in fashioning the facts
other inanimate objects have qualia, whether bats and other spe-
must be specified makes pointless any identification of the
cies have the kind of qualia we have or don't have the kind of
physical with the real and of the perceptual with the
qualia we have, whether groups can feel pain, and so on.
merely apparent. The perceptual is no more a rather
But why should any philosopher think it is even a logical pos-
distorted version of the physical facts than the physical is
sibility that a rock can have a pain (i.e. that an event of the same
a highly artificial version of the perceptual facts. 10
'qualitative character' as a human pain can take place 'in' a

If I see a red tablecloth at two different times during the day, rock)? Perhaps Russell gives us some clue to the nature of this

do I have the same sensation of red? Or do I have different sen- kind of metaphysical realism. Russell was a realist about qualia
sations and not notice the difference (unless I happen to be a and a realist about universals. Moreover, he took qualia to be
painter)? paradigmatic universals. A universal is, above all, a way in
An especially baffling case is the case of accommodation. If a which things can be similar; and to Russell it seemed that the
subject is given glasses which turn the image upside down, after qualitative similarities of one's own sensations are the episte-

a time things will again look normal to the subject. Have the mologically most primitive and most fundamental examples of
'ways in which things can be similar'. Qualia, for Russell, are
10 Ways of Worldmaking, pp. 92-3. universals par excellence.
100 Mind and body Mind and body 101

by
Universals, however, are thought of as totally well-defined case as to whether two sensations (let alone two arbitrary events)
words may be vague, but universals them-
a traditional realist: are qualitatively similar or dissimilar.
selves can't be vague. (A vague word is vague because it stands Let £ be the event of my having a particular sensation at a
for a vague set of concepts, Godel once said in a conversation; particular time and £' be some physical event in a rock. The
but the concepts are perfectly well-defined.) £ (say, red H) might
suggestion that the qualitative character of
So, if qualia are universals and universals are by nature well- be identical with or correlated to some such property as {P x or
defined, it must be perfectly well-defined whether any given P2 or P 3 (where P 3 is the property of being a rock) offends any
)

thing or event - including a half of a split brain or some event in sane human sensibility. The suggestion that E and £' might be
it; including a rock or some event in it; including a nation or 'qualitatively similar' events is absurd. We have already dis-
group or some event in it- does or doesn't exhibit any given cussed one explanation of this absurdity: the explanation that
quale. And if the quale is thought of as independent of the func- the hypothesis is absurd because it violates the methodological
tional role it plays, if it is thought to be wholly contingent that maxim 'do not ascribe properties to an object without a reason'.
the qualitative character of a sensation of red is the qualitative Even if this explanation worked, what it would yield is far less
character of something which has that particular functional role, than the impossibility of rocks having qualia (or the incoherence
then it does seem to be a logical possibility that the split brain or of the notion that they do). If this is all that is wrong with the
the rock has that quale. 'hypothesis' that rocks have qualia then we are in the position of
A philosopher like myself who wishes to deny that every one having to say: it is possible for all we know that rocks have
of these possibilities makes sense (although some of them may - qualia, but it is a priori highly improbable that they do.
there is a temptation to treat the right lobe of the split brain as In fact, the hypothesis that rocks have qualia is incoherent in
a 'locus of consciousness', and I have suggested that it would be much the way that the Brain in a Vat hypothesis is incoherent:
legitimate to decide to do this) has to be careful to make clear like the Brain in a Vat hypothesis, this 'hypothesis' presupposes
that he is not espousing some form of behaviorism. Saying that a magical theory of reference. Any sane human being regards E
*qualia' are not well defined entities is not the same thing as say- and E as so dissimilar that the question of 'qualitative similar-
f

ing they don't exist, that just behavior, or whatever.

it is all
ity' (in the sense in which two sensations can be qualitatively
Many notions are vague and have some clear referents. The
still
similar, i.e. feel the same way) does not even arise. But the meta-
notion of a house, for example, is ill-defined in the case of igloos physical realist, while not denying this at all, thinks that £ and
(is an igloo a house?), in the case of hogans, perhaps in other £' might (logically possibly) be similar in this way, even though
cases as well. ButTne fact that there is no fact of the matter as to it is 'crazy' to think so. And he thinks this because he is under
whether or not an igloo is a house doesn't mean that houses the illusion that by having the sensation in question, with its
don't exist. And, similarly, the fact that there is no fact of the qualitative character, its 'the way it feels', with its functional
matter as to whether or not the right lobe is 'conscious' doesn't role, with the accompanying thoughts and judgments, he has
mean that conscious beings don't exist. somehow brought it about that the expression 'the way this sen-
Qualitative similarity of sensations is defined to some extent: sation feels' (or some technical substitute, e.g. 'the qualitative
if Ihave a sensation of red followed by a sensation of green, I character of this sensation', or 'red H \ or 'this quale') refers to
know that I have had dissimilar sensations (and I know this one definite 'universal', one absolutely well-defined property of
without comparing their functional roles), and if I have a sensa- metaphysical individual events. But this is not the case.
tion of red followed by the 'same' sensation of red, I know (up If were robots functionally isomorphic to us and
there actually
to the vagueness we discussed above) that I have had similar we worked with them, argued with them, had some of them as
sensations. But, for someone with an 'internalist' perspective on friends, we would quickly feel sure that they were conscious.
truth, it does not follow that there is a fact of the matter in every (We might still be puzzled as to whether they had the same
102 Mind and body

qualiawe 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
Two conceptions of rationality
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 In the preceding chapters I have spoken of rationality and of
there is nothing hidden here, no noumenal fact of the entities' 'rational acceptability'. But rationality is not an easy thing to

really being conscious or really not being conscious, or of the give an account of.

qualities' really being the same or really being different. There The problem is not without analogues in other areas. Some
are only the obvious empirical facts: that rocks and nations are years ago I studied the behavior of natural kind words, for
grossly dissimilar from people and animals; that robots of var- example, gold, and I came to the conclusion that the extension
ious kinds are in between sorts of objects; and so on. Rocks and of the term is not simply determined by a 'battery of semantical
nations aren't conscious; that is a fact about the notion of con- rules',or other institutionalized norms. The norms may deter-
sciousness we actually have. mine that certain objects arc paradigmatic examples of gold; but
What makes this line seem so disturbing is that it makes our they do not determine the full extension of the term, nor is it
standards of rational acceptability, justification and ultimately impossible that even a paradigmatic example should turn out
of truth, dependent on standards of similarity which are clearly not to really be gold, as it would be if the norms simply defined
the product of our biological and cultural heritage (e.g. whether what it is to be gold.
we have or haven't interacted with 'intelligent robots'). But We are prepared to count something as belonging to a kind
something like this is true of most of the language we use in even if our present tests do not suffice to show it is a member of
everyday life, of such words as 'person', 'house', 'snow', and the kind if it ever turns out that it has the same essential nature
'brown', for example. A realist who accepted this resolution of as (or, more vaguely, is 'sufficiently similar' to) the paradigmatic
the puzzles about qualia would be likely to express it by saying examples (or the great majority of them). What the essential
that 'qualia don't really exist', or that qualia belong to our 'sec- nature is, or what counts as sufficient similarity, depends both
ond class conceptual system'; but what is the point of a notion on the natural kind and on the context (iced tea may be 'water'
of 'existence' that puts houses on the side of the non-existent? in one context but not in another); but for gold what counts is
Our world is a human world, and what is conscious and not ultimate composition, since this has been thought since the
conscious, what has sensations and what doesn't, what is quali- ancient Greeks to determine the lawful behavior of the sub-
tatively similar to what and what is dissimilar, are all dependent stance. Unless we say that what the ancient Greeks meant by
ultimately on our human judgments of likeness and difference. chrysos was whatever has the same essential nature as the para-
digmatic examples, then neither their search for new methods of
104 Two conceptions of rationality Two conceptions of rationality 105

detecting counterfeit gold (which led Archimedes to the density (which is almost coextensive with theory of rationality). In a

test) nor their physical speculations will make sense. publication a few years 1 ago I described the desiderata for a
It is tempting to take the same line with rationality itself, and moral system, following Grice and Baker, and I included (1) the
to say that what determines whether a belief is rational is not the desire that one's basic assumptions, at least, should have wide
norms of rationality of this or that culture, but an ideal theory appeal; (2) the desire that one's system should be able to with-
of rationality, a theory which would give necessary and suffi- stand rational criticism; (3) the desire that the morality recom-
cient conditions for a belief to be rational in the relevant circum- mended should be livable.
stances in any possible world. Such a theory would have to The way to develop a better understanding of the nature of
account for the paradigmatic examples, as an ideal theory of rationality - the only way we know - is, likewise, to develop

gold accounts for the paradigmatic examples of gold; but it better philosophical conceptions of rationality. (An unending
could go beyond them, and provide criteria which would enable process; but that is as it should be.) It is striking that the desid-

us to understand cases we cannot presently see to the bottom of, erata I listed for a moral system, unchanged, could be listed as

as our present theory of gold enables us to understand cases the the desiderata for a methodology or a system of rational proce-
most brilliant ancient Greek could not have understood. A gen- dure in any major area of human concern. In analytical philos-
eral difficulty with the proposal to treat 'rational', 'reasonable', ophy the main attempts to better understand the nature of
'justified', etc., as natural kind terms is that the prospects for way have come from philosophers of
rationality in this science,
actually finding powerful generalizations about all rationally and two important tendencies have resulted from these efforts.

acceptable beliefs seem so poor. There are powerful universal

laws obeyed by all instances of gold, which is what makes it
Logical positivism
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 In the past fifty years the clearest manifestation of the tendency

we can find powerful universal generalizations obeyed by all to think of the methods of 'rational justification' as given by
instances of rationally justified belief? something like a list or canon (although one that philosophers of
That the chances are poor does not mean that there are no science have admittedly not yet succeeded in fully formalizing)
analogies between scientific inquiry into the nature of gold and was the movement known as Logical Positivism. Not only was
moral inquiry or philosophical inquiry. In ethics, for example, the list or canon that the positivists hoped 'logicians of science'
we start with judgments that individual acts are right or wrong, (their term for philosophers) would one day succeed in writing

('observation reports', so to speak) and we gradually formulate down supposed to exhaustively describe the 'scientific method';
maxims (not exceptionless generalizations) based on those judg- but, since, according to the logical positivists, the 'scientific
ments, often accompanied by reasons or illustrative examples, as method' exhausts rationality itself, and testability by that

for instance 'Be kind to the stranger among you, because you method exhausts meaningfulness (The meaning of a sentence is
know what it was
like to be a stranger in Egypt' (a 'low level its method of verification'), the list or canon would determine

generalization'). These maxims in turn affect and alter our judg- what is and what is not a cognitively meaningful statement.
ments about individual cases, so that new maxims supplement- Statements testable by the methods in the list (the methods of
ing or modifying the earlier ones may appear. After thousands mathematics, logic, and the empirical sciences) would count as
of years of this dialectic between maxims and judgments about meaningful; all other statements, the positivists maintained, are
individual cases, a philosopher may come along and propose a 'pseudo-statements', or disguised nonsense.
moral conception (a 'theory'), which may alter both maxims and 1
'Literature, Science, and Reflection', New Literary History, vol. VII,
singular judgments and so on. 1975-6, reprinted in my Meaning and the Moral Sciences, Routledge
The very same procedure may be found in all of philosophy and Kegan Paul, 1978.
106 Two conceptions of rationality
Two conceptions of rationality 107

Anobvious rejoinder was to say that the Logical Positivist

sceptic who challenges us to 'prove' such statements as 'I am
standing on the floor' testifies to the existence of social norms
criterion of significance was self-refuting: for the criterion itself
requiring agreement to such statements in the appropriate cir-
is neither (a) 'analytic' (a term used by the positivists to account
cumstances.
for logic and mathematics), nor (b) empirically testable.
Wittgenstein argued that without such public norms, norms
Strangely enough this criticism had very little impact on the log-
shared by a group and constituting a 'form of life', language and
ical positivists and did little to impede the growth of their move-
even thought itself would be impossible. For Wittgenstein it is
ment. I believe that the neglect of this particular philosophical
absurd to ask if the institutionalized verification I have been
gambit was a great mistake; that the gambit is not only correct,
but contains a deep lesson, and not just a lesson about Logical
speaking of is 'really' justificatory. In On Certainty Wittgenstein
remarks that philosophers can provide one with a hundred epis-
Positivism.
The point I am going to develop will depend on the following
temological 'justifications' of the statement 'cats don't grow on
trees' - but none of them starts with anything which is more sure
observation: the forms of 'verification' allowed by the logical
(in just this institutionalized sense of 'sure') than the fact that
positivists are forms which have been institutionalized by mod-
cats don't grow on trees.
ern society. What can be 'verified' in the positivist sense can be
Sceptics have doubted not only perceptual judgments but
verified to be correct (in a non-philosophical or prephilosophical
sense of 'correct'), or to be probably correct, or to be highly
ordinary inductions. Hume, whose distinction between what is
rational and what is reasonable I am not observing, would have
successful science, as the case may be; and the public recognition
said there is no rational proof that it will snow (or even that it
of the correctness, or the probable correctness, or the 'highly
will probably snow) in the United States this winter (although he
successful scientific theory' status, exemplifies, celebrates, and
reinforces images of knowledge and norms of reasonableness would have added that it would be most unreasonable to doubt
maintained by our culture.
that it will). Yet our response to a sceptic who challenges us to

On the face of it, the original positivist paradigm of verifica-

'prove' that it will snow in the United States this winter testifies
that there are social norms requiring agreement to such 'induc-
tion was not this publicly institutionalized one. In Carnap's Der
tions' just as much as to ordinary perceptual judgments about
Logische Aufbau der Welt (The Logical Construction of the
people standing on floors and about equal arm balances.
World) verification was ultimately private, based on sensations
whose subjective quality or 'content' was said to be 'incommu-
When we come to high-level theories in the exact sciences,

nicable'. But, under the urgings of Neurath, Carnap soon shifted

people's reactions are somewhat different. Ordinary people can-
not 'verify' the special theory of relativity. Indeed, ordinary peo-
to a more public, more 'intersubjective', conception of verifica-
tion.
ple do not at the present time even learn the special theory, or
the (relatively elementary) mathematics needed to understand it,
Popper has stressed the idea that scientific predictions are con-
although it is beginning to be taught in freshman physics courses
fronted with 'basic sentences', sentences such as 'the right pan of
the balance is down' which are publicly accepted even if they
in some of our colleges. Ordinary people defer to scientists for
an informed (and socially accepted) appraisal of a theory of this
cannot be 'proved' to the satisfaction of a sceptic. He has been
type. And because of the instability of scientific theories, a sci-
criticized for using 'conventionalist' language here, for speaking
entist is not likely to refer to even so successful a theory as spe-
as if it were a convention or social decision to accept a basic
cial relativity as 'true' tout court. But the judgment of the scien-
sentence; but I think that what sounds like a conventionalist ele-
ment in Popper's thought is simply a recognition of the institu-
tificcommunity is that special relativity is a 'successful' - in fact,
tionalized nature of the implicit norms to which we appeal in
like quantum electrodynamics, an unprecedentedly successful -
scientific theory, which yields 'successful predictions' and which
ordinary perceptual judgments. The nature of our response to a
 108 Two conceptions of rationality Two conceptions of rationality 109

is 'supported by a vast number of experiments'. And these judg- analysis is more or less nonsense. But his rejection of evolution
ments are, in fact, deferred to by other members of the society. is quite striking. 2 Wittgenstein contrasts Darwin's theory unfa-
The difference between this case and the cases of institutional- vorably with theories in physics ('One of the most important
izednorms of verification previously referred to (apart from the things about an explanation is that it should work, that it should
hedging of the adjective 'true') is the special role of experts and enable us to predict something. Physics is connected with Engi-
the institutionalized deference to experts that such a case The bridge must not fall down'
neering. (Lectures on Aesthetics,
involves; but this no more than an instance of the division of
is
p. 25)). And he which
says people were persuaded 'on grounds
intellectual labor (not to mention intellectual authority relations)
were extremely end you forget entirely every ques-
thin'. 'In the
in the society. The judgment that special relativity and quantum
tion of verification, you are just sure it must have been like that.'
electrodynamics are 'the most successful physical theories we Again, the great discussions about 'analyticity' that went on
have' is one which is made by authorities which the society has in the 1950s seem to me to be connected with the desire of phi-
appointed and whose authority is recognized by a host of prac- losophers to find an objective, uncontroversial foundation for
tices and ceremonies, and in that sense institutionalized.
their arguments. 'Analyticity', i.e. the doctrine of truth by virtue
Recently occurred to me that Wittgenstein may well have
it
of meaning alone, came under attack because it had been over-
thought that only statements that can be verified in some such used by philosophers. But why had philosophers been tempted
'institutionalized' way can be true (or right, or correct, or justi-
to announce that so many things which are in no intelligible
fied) at all. I don't mean to suggest that any philosopher ever
sense 'rules of language', or consequences of rules of language,
held the view that all things which count in our society as 'justi-
were analytic or 'conceptually necessary', or whatever? The
fications' really are such. Philosophers generally distinguish answer, I think, is that the idea that there is a definite set of rules
between institutions which are constitutive of our concepts
of language and that these can settle what is and is not rational,
themselves and those which have some other status, although had two advantages, as philosophers thought: (1) the 'rules of
there is much controversy about how to make such a distinction.
2
mean to suggest that Wittgenstein thought that Concerning evolution what Wittgenstein said was 'People were certain
I it was some on grounds which were extremely thin. Couldn't there have been an
subset of our institutionalized verification norms that determines attitude which said: "I don't know. It is an interesting hypothesis which
what it is right to say in the various 'language games' we play may eventually be well confirmed" ', Lectures on Aesthetics, p. 26, in
and what wrong, and that there is no objective lightness or
is
Cyril Burret (ed.) L. W. Wittgenstein: Lectures and Conversations,
Berkeley, University of California Press, 1967. Whatit would be like for
wrongness beyond this. Although such an interpretation does fit evolution to be 'well confirmed* Wittgenstein does not say, but the
much that Wittgenstein says - for instance, the stress on the paragraph suggests that actually seeing speciation occur is what he has
need for 'agreement in our judgments' in order to have concepts 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
at all - I do not feel sure that it is right. It is just too vague who
the bucket.')
the 'we' is in Wittgenstein's talk
of 'our' judgments; and I don't It is instructive to contrast Wittgenstein's attitude with Monod's:
know whether his 'forms of
correspond to the institutional-
life' Darwin himself had stated it,
the selective theory of evolution, as
ized norms I have mentioned. But this interpretation occurred to required the discovery of Mendelian genetics, which of course
was made. This is an example, and a most important one,
me upon reading Wittgenstein's Lectures and Conversations. In
of what is meant by the content of a theory, the content of an
this Wittgenstein rejects both psychoanalysis and Darwin's the- idea ... [A] good theory or a good idea will be much wider and
ory of evolution (although unlike the positivists he does not much richer than even the inventor of the idea may know at

regard such language as meaningless, and he has admiration for his time. The theory may be judged precisely on this type
of development, when more and more falls into its lap, even
Freud's 'cleverness'). Wittgenstein's view about psychoanalysis though it was not predictable that so much would come of it J.
{

(which he calls a 'myth') does not signify much, since so many Monod, 'On the Molecular Theory of Evolution*, in Harre,
R. (ed.), Problems of Scientific Revolution: Progress and
people have the view - mistakenly in my opinion - that psycho-
Obstacles to Progress in the Sciences, Oxford, 1975).
110 Two conceptions of rationality Two conceptions of rationality 111

language' are constitutive institutionalized practices (or norms The gambit I referred to at the outset, the gambit that refutes
which underlie such practices), and as such have the 'public' sta- the logical positivists' verification principle, is deep precisely
tus I have described; (2) at the same time, it was claimed that because it refutes every attempt to argue for a criterial concep-
only philosophers (and not linguists) could discover these mys- tion of rationality, that is because it refutes the thesis that noth-
terious things. was a nice idea while
It it lasted, but it was bound ing is rationally verifiable unless it is criterially verifiable.
to be exploded, and it was. The point is that although the philosophers
I mentioned often
I any conception according to which there are insti-
shall call spoke as if their arguments had the same kind of finality as a
tutionalized norms which define what is and is not rationally mathematical proof or a demonstration experiment in physics;
acceptable a criterial conception of rationality. The logical posi- that although the logical positivists called their work logic of
tivists, Wittgenstein, at least on the admittedly uncertain inter- science; although the Wittgensteinians displayed unbelievable
pretation have essayed, and some though not all of the 'ordi-
I arrogance towards philosophers who could not 'see' that all
nary language' philosophers 3 at Oxford shared a criterial philosophical activity of a pre-Wittgensteinian or non-Wittgen-
conception of rationality even if they differed on other issues, steinian kind is nonsensical; and although ordinary language
such as whether to call unverifiable statements 'meaningless', philosophers referred to each other's arguments and those of
and over whether or not some ethical propositions could be non-ordinary language philosophers as 'howlers' (as if philo-
'conceptually necessary'. sophical errors were like mistakes on an arithmetic test); no phil-
osophical position of any importance can be verified in the con-
3
One might develop an 'ordinary language* philosophy which was not clusive and culturally recognized way I have described. In short,
committed to the public and 'criterial' verification of philosophical if it istrue that only statements that can be criterially verified
theses if one could develop and support a conception in which the norms
which govern linguistic practices are not themselves discoverable by
can be rationally acceptable, that statement itself cannot be cri-
ordinary empirical investigation. In Must We Mean What We terially verified, and hence cannot be rationally acceptable. If
Say, Stanley Cavell took a significant step in this direction, arguing that there is such a thing as rationality at all - and we commit our-
such norms can be known by a species of 'self knowledge* which he
compared to the insight achieved through therapy and also to the
selves to believing in some notion of rationality by engaging in
transcendental knowledge sought by phenomenology. While I agree with the activities of speaking and arguing - then it is self- refuting to
Cavell that my knowledge as a native speaker that certain uses are argue for the position that it is identical with or properly con-
deviant or non-deviant is not 'externar inductive knowledge - I
can know without evidence that in my dialect of English one says 'mice'
tained in what the institutionalized norms of the culture deter-
and not 'mouses* - I am inclined to think this fact of speaker's mine to be instances of it. For no such argument can be certified
privileged access does not extend to generalizations about correctness to be correct, or even probably correct, by those norms alone.
and incorrectness. If I say (as Cavell does) that it is part of the rule
I don't at all think that rational argumentation and rational
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 justification are impossible in philosophy, but rather I have been
to explain my intuitionsabout specific cases, not just reporting those driven to recognize something which is probably evident to lay-
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
men if not to philosophers, namely that we cannot appeal to
of immunity from criticism by others, including criticisms which public norms what is and is not rationally argued and
to decide
depend on offering rival explanations, in either case. And if one allows justified in philosophy. The claim which is still often heard that
the legitimacy of such criticism, then the activity of discovering such
norms begins to look like social science or history - areas in which, I philosophy is 'conceptual analysis', that the concepts themselves

have argued, traditional accounts of 'The Scientific Method* shed little determine what philosophical arguments are right, is, when
light. (See my Meaning and the Moral Sciences, Routledge and
combined with the doctrine that concepts are norms or rules
Kegan Paul, 1978.)
In any case, whatever their status, I see no reason to believe that the
underlying public linguistic practices, just a covert form of the
norms for the use of language are what decide the extension of claim that all rational justification in philosophy and
is criterial,
'rationally acceptable*, 'justified', 'well confirmed', and the like. that philosophical truth is (barring 'howlers') as publicly demon-
112 Two conceptions of rationality Two conceptions of rationality 113

Such a view seems to me to be simply

strable as scientific truth. In sum, what the logical positivists and Wittgenstein (and per-

unreasonable in the light of the whole history of the subject, haps the later Quine as well) did was to produce philosophies
including the recent history. which leave no room for a rational activity of philosophy. This
What goes for philosophical argument goes for arguments is why
these views are self-refuting; and also why the little gam-

about religion and about secular ideology as well. An argument bit have been discussing represents a significant argument of
I

between an intelligent liberal and an intelligent Marxist will the kind philosophers call a 'transcendental argument': arguing

have the same character as a philosophical dispute at the end, about the nature of rationality (the task of the philosophers par
even if more empirical facts are relevant. And we all do have excellence) is an activity that presupposes a notion of rational
views in religion, or politics, or philosophy, and we all argue justification wider than the positivist notion, indeed wider than
them and criticize the arguments of others. Indeed, even in 'sci- institutionalized criterial rationality.

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 Anarchism is self-refuting
description of the 'scientific method' to include these subjects;
but so broadened it cannot be shown to clearly exclude anything Let me now discuss a very different philosophical tendency.

whatsoever. (See Chapter 8.)

Thomas Kuhn's The Structure of Scientific Revolutions {SSR)

The positivists, I will be reminded, conceded that the verifica- enthralled vast numbers of readers, and appalled most philoso-
tion principle was 'cognitively meaningless'. They said it was a phers of science because of its emphasis on what seemed to be
proposal and as such not true or false. But they argued for their irrational determinants of scientific theory acceptance and by
4 its use of such terms as 'conversion' and 'Gestalt switch'. In fact,
proposal, and the arguments were (and had to be) non-starters.
So the point stands. Kuhn made a number of important points about scientific theo-
ries and about how scientific activity should be viewed. I have
4
The weakest argument offered in defense of the Verification Principle
expressed a belief in the importance of the notions of paradigm,
construed as a proposal was that it 'explicated* the *pre-analytic' notion
of meaningfulness. (For a discussion of this claim, see my 'How normal science, and scientific revolution elsewhere; at this point
Not to Talk about Meaning*, in my Mind, Language and Reality, I want to focus on what I do not find sympathetic in Kuhn's
Philosophical Papers, Vol 2, Cambridge University Press, 1975.)
book, what I described elsewhere as 'Kuhn's extreme relativism'.
Reichenbach defended a form of the Verification Principle (in Experience
and Prediction) as preserving all differences in meaning relevant to The reading that enthralled Kuhn's more sophomoric readers
behavior. Against an obvious objection (that the non-empirical belief in was one according to which he is saying that there is no such
a divinity - Reichenbach used the example of Egyptian cat wor-
thing as rational justification in science, just Gestalt switches
shippers - could alter behavior) Reichenbach replied by proposing to
it's

translate 'Cats are divine animals' as 'cats inspire feelings of awe in and conversions. Kuhn has rejected this interpretation of the
cat-worshippers'. Clearly the acceptance of this substitute would SSR, and has since introduced a notion of 'non-paradigmatic
not leave behavior unchanged in the case of a cat worshipper!
rationality' which may be closely related to if not the same as
The most interesting view was that of Carnap. According to Carnap,
all rational reconstructions are proposals. The only factual questions what I just called 'non-criterial rationality'.
concern the logical and empirical consequences of accepting this or that The tendency that most readers thought they detected in
rational reconstruction. (Carnap compared the 'choice' of a rational
reconstruction to the choice of an engine for an airplane.) The con-
Kuhn's SSR certainly manifested itself in Paul Feyerabend's
clusion he drew was that in philosophy one should be tolerant of Against Method, Feyerabend, like Kuhn, stressed the manner in
divergent rational reconstructions. However, this principle of Tolerance, which different cultures and historic epochs produce different
as Carnap called it, presupposes the Verification Principle. For the
paradigms of rationality. He suggests that the determinants of
doctrine that no rational reconstruction is uniquely correct or corre-
sponds to the way things 'really are', the doctrine that all 'external our conceptions of scientific rationality are largely what we
questions' are without cognitive sense is just the Verification Principle. would call irrational. In effect, although he does not put it this
To apply the Principle of Tolerance to the Verification Principle
itself would be circular.
way, he suggests that the modern scientific-technological con-
114 Two conceptions of rationality Two conceptions of rationality 115

ception of rationality is fraudulent by its own standards. (I think responses to stimuli (including noises that curiously resemble
I detect a similar strain in Michel Foucault.) And he goes far English or Italian). To tell us that Galileo had 'incommensura-
beyond Kuhn or Foucault in suggesting that even the vaunted ble' notions and
then to go on to describe them at length is
instrumental superiority of our science may be somewhat of a totally incoherent.
hoax. Faith healers can do more to relieve your pain than doc- This problem is posed in a sympathetic essay on Feyerabend's
tors, Feyerabend claims. view by Smart: 5
It is not those terrifyingly radical claims that I want to talk

about, although they are the reason Feyerabend calls his position Surely it is a neutral fact that in order to see Mercury
'anarchism'. wish to discuss a claim Kuhn does make in both
I
we have to point the telescope over the top of that
the SSR and subsequent papers, and that Feyerabend made both
tree, say, and not, as predicted by Newtonian theory,
in AgainstMethod and in technical papers. This is trie thesis of over the top of that chimney pot. And surely one can
incommensurability. I want to say that this thesis, like the logical talk of trees, chimney pots, and telescopes in a way
positivist thesis about meaning and verification, is a self-refuting
which is independent of the choice between Newton-
thesis. In short, I want to claim that both of the two most influ-
ian and Einsteinian theory. However Feyerabend could
ential philosophies of science of the twentieth century, certainly
well concede that we use Euclidean geometry and non-
the two and non-philosophers gen-
that have interested scientists
relativistic optics for the theory of our telescope. He
erally, the only two the educated general reader is likely to have
would say that this is not the real truth about our tele-
even heard of, are self-refuting. Of course, as a philosopher of
scope, the tree, and the chimney pot, but nevertheless
science I find it a bit troublesome that this should be the case.
it is legitimate to think in this way in order to discuss
We shall shortly come to the question of what to make of this
the observational tests of general relativity, since we know
situation.
on theoretical grounds that our predictions will be un-
The incommensurability thesis is the thesis that terms used
affected (up to the limits of observational error) if we
in another culture, say, the term 'temperature' as used by a
avail ourselves of this computational convenience.
seventeenth-century scientist, cannot be equated in meaning or
reference with any terms or expressions we possess. As Kuhn
But the trouble with Smart's rescue move is that I must under-
puts it, scientists with different paradigms inhabit 'different
stand some of the Euclidean non-relativists' language to even say
worlds'. 'Electron' as used around 1900 referred to objects
the 'predictions' are the same. If every word has a different sig-
in one 'world'; as used today it refers to objects in quite a
different 'world'. This thesis is supposed to apply to observa- nificance, in what sense can any prediction be 'unaffected'? How
can I even translate the logical particles (the words for 'if-then',
tional language as well as to so-called 'theoretical language';
simply a 'not',and so on) in seventeenth-century Italian, or whatever, if
indeed, according to Feyerabend, ordinary language is
I cannot find a translation manual connecting seventeenth-
false theory.
century Italian and modern English that makes some kind of sys-
The rejoinder this time is that if this thesis were really true
tematic sense of the seventeenth-century corpus, both in itself
then we could not translate other languages - or even past stages
and in its extra-linguistic setting? Even if I am the speaker who
of our own language - at all/ And if we cannot interpret orga-
employs both theories (as Smart envisages) how can I be justified
nisms' noises at all, then we have no grounds for regarding them
as thinkers, speakers, or even persons. In short, if Feyerabend
5
(and Kuhn at his most incommensurable) were right, then mem- J. J.C. Smart, 'Conflicting Views about Explanation*, in R. Cohen and
M. Wartofsky (eds.), Boston Studies in the Philosophy of Science,
bers of other cultures, including seventeenth-century scientists,
Volume II: in Honor of Philipp Frank (New York, Humanities Press,
would be conceptualizable by us only as animals producing Inc., 1965).
116 Two conceptions of rationality Two conceptions of rationality 117

in equating any word in my Newtonian theory with any word in pretation itself. When we translate a word as, say, temperature
my general relativistic theory? we equate the reference and, to the extent that we stick to our
The point I am making comes into even sharper focus when translation, the sense of the translated expression with that of
we apply to it some of Quine's and Davidson's observations our own term 'temperature', at least as we use it in that context.
about meaning and translation practice. Once it is conceded that (Of course, there are various devices we can use, such as special
we can find a translation scheme which 'works' in the case of a glosses, to delimit or delineate the way we are employing 'tem-
seventeenth century text, at least in the context fixed by our perature', or whatever the word may be, in the context.) In this
interests and the use to which the translation will be put, what sense we equate the 'concept' with our own 'concept'
in question
sense does it have in that context to say that the translation does of temperature. But so doing is compatible with the fact that the
not 'really' capture the sense or reference of the original? It is seventeenth-century scientists, or whoever, may have had a dif-
not, after all, as if we had or were likely to have criteria for ferent conception of temperature, that is a different set of beliefs
sameness of sense or reference apart from our translation about it and its nature than we do, different 'images of knowl-

schemes and our explicit or implicit requirements for their edge', and different ultimate beliefs about many other matters as
empirical adequacy. One can understand the assertion that a well. That conceptions differ does not prove the impossibility of
translation fails to capture exactly the sense or reference of the ever translating anyone 'really correctly' as is sometimes sup-
original as an admission that a better translation scheme might posed; on the contrary, we could not say that conceptions differ
be found; but it makes only an illusion of sense to say that all and how they differ if we couldn't translate.
possible translation schemes fail to capture the 'real' sense or But, it may be asked, how do we ever know that a translation
reference. Synonymy exists only as a relation, or better, as a fam- scheme 'works' if conceptions always turn out to be different?
ily of relations, each of them somewhat vague, which we employ The answer to this question, as given by various thinkers from
to equate different expressions for the purposes of interpreta- Vico down to the present day, is that interpretative success does
tion. The idea that there is some such thing as 'real' synonymy not require that the translatees' beliefs come out the same as our
apart from all workable practices of mutual interpretation, has own, but it does require that they come out intelligible to us.
been discarded as a myth. This is the basis of all the various maxims of interpretative char-
Suppose someone tells us that the German word 'Rad' can be ity or 'benefit of the doubt', such as 'interpret them so they come
translated as 'wheel'. If he goes on to say his translation is not out believers of truths and lovers of the good', or 'interpret them
perfect, we naturally expect him to indicate how it might be so that their beliefscome out reasonable in the light of what they
improved, supplemented by a gloss, or whatever. But if he goes have been taught and have experienced', or Vico's own directive
on to say that 'Rad' can be translated as 'wheel', but it doesn't to maximize the humanity of the person being interpreted. It is
actually refer to wheels, or indeed to any objects recognized in a constitutive fact about human experience in a world of differ-
your conceptual system, what do we get from this? To say that ent cultures interacting in history while individually undergoing
a word A can be translated as 'wheel', or whatever, is to say slower or more rapid change that we are, as a matter of universal
that, to the extent that the translation can be relied upon, A human experience, able to do this; able to interpret one
refers to wheels. another's beliefs, desires, and utterances so that it all makes
Perhaps the reason that the incommensurability thesis some kind of sense.
intrigues people so much, apart from the appeal which all inco- Kuhn and Feyerabend, not surprisingly, reject any idea of con-
herent ideas seem to have, is the tendency to confuse or conflate vergence in scientific knowledge. Since we are not talking about
concept and conception. To the extent that the analytic/synthetic the same things as previous scientists, we are not getting more
distinction is fuzzy, this distinction too is fuzzy; but all interpre- and more knowledge about the same microscopic or macro-
tation involves such a distinction, even if it is relative to the inter- scopic objects. Kuhn argues that science 'progresses' only instru-
118 Two conceptions of rationality Two conceptions of rationality 119

mentally; we get better and better able to transport people from years ago with 'grass' today, no statement about the reference of
one place to another, and so on. But this too is incoherent. this word 200 years ago could be made. Nor is it only natural
Unless such locutions as 'transport people from one place to kind words that are so dependent for interpretation on principles
another* retain some degree of fixity of reference, how can we of charity; the artifact word 'bread' would pose exactly the same
understand the notion of instrumental success in any stable way? problems. Indeed, without interpretative charity we could not
The argument I have just employed is essentially related to equate a simple color term such as 'red' across different speak-
Kant's celebrated arguments about preconditions for empirical ers. We interpret discourse always as a whole; and the interpre-

knowledge. Replying to the contention that the future might be tation of 'observation' terms is as dependent on the interpreta-
wholly lawless, might defeat every 'induction* we have made, tion of 'theoretical' terms as is the interpretation of the latter on
Kant pointed out that if there is any future at all - any future for the former.
us, at any rate, any future we can grasp as thinkers and concep- What I have given is, once again, a transcendental argument.
tualize to say if our predictions were true or false - then, in fact, We are committed by our fundamental conceptions to treating
many regularities must not have been violated. Else why call it a not just our present time-slices, but also our past selves, our

future? For example, when we imagine balls coming from an urn ancestors,and members of other cultures past and present, as
in some 'irregular' order, we forget that we couldn't even tell
persons; and that means, I have argued, attributing to them
they were balls, or what order they came out in, without
tell shared references and shared concepts, however different the
depending on many regularities. Comparison presupposes there conceptions that we also attribute. Not only do we share objects
are some commensurabilities. and concepts with others, to the extent that the interpretative
There is a move Kuhn and Feyerabend could make in reply to exercise succeeds, but also conceptions of the reasonable, of the
all these criticisms, but it is not one they would feel happy mak- natural, and so on. For the whole justification of an interpreta-
ing, and that would be to introduce some kind of observa- tive scheme, remember, is that it renders the behavior of others
tional/theoretical dichotomy. They could concede commensura- at least minimally reasonable by our lights. However different
bility, translatability, and even convergence with respect to our images of knowledge and conceptions of rationality, we
observational facts, and restrict the incommensurability thesis to share a huge fund of assumptions and beliefs about what is rea-

the theoretical vocabulary. Even then there would be problems sonable with even the most bizarre culture we can succeed in
(why shouldn't we describe the meanings of the theoretical terms interpreting at all.

via their relations to the observational vocabulary a la Ramsey?)

But Kuhn and Feyerabend reject this alternative with reason, for
Why relativism is inconsistent
in fact the need for principles of interpretative charity is just as
pervasive in 'observational language' as in 'theoretical language'. That (total) relativism is inconsistent is a truism among philos-

Consider, for example, the common word 'grass'. Different ophers. After not obviously contradictory to hold a point
all, is it

speakers, depending on where and when they live have different of view while at the same time holding that no point of view is
perceptual prototypes of grass (grass has different colors and dif- more justified or right than any other? Alan Garfinkel has put
ferent shapes in different places) and different conceptions of the point very wittily. In talking to his California students he
grass. Even speakers must know that grass is a plant, on
if all once said, aping their locutions: 'You may not be coming from
9

pain of being said to have a different concept altogether, the where I'm coming from, but I know relativism isn't true for me
conception of a plant today involves photosynthesis and the con- ... If any point of view is as good as any other, then why isn't

ception of a plant two hundred years ago did not. Without inter- the point of view that relativism is false as good as any other?
pretative charity which directs us to equate 'plant' 200 years ago The plethora of relativistic doctrines being marketed today
with 'plant' today (at least in ordinary contexts) and 'grass' 200 (and marketed by highly intelligent thinkers) indicates this sim-
120 Two conceptions of rationality Two conceptions of rationality 121

pie refutation will not suffice. Why should an intelligent relativ- This Plato took to be a reductio ad absurdum. However,
ist concede that every view is as true {for him) as any other? He Plato's argument is not a good one as it stands. Why should
cannot prevent you (or Alan Garfinkel) from saying that his view Protagoras not agree that his analysis applies to itself? It doesn't
isnot true for you (or justified for you, or whatever): but if he follow that it must an
be self-applied infinite number of times,
has his wits about him, he can retort that truth for you is far less but only that it can be self-applied any finite number of times.
salient (for him) than is truth for him. What concept of anything But Plato had noticed something very deep.
is more salient than one's own, after all? Is it then really incon- When one first encounters relativism, the idea seems simple
sistent to treat true, justified, etc., as relative notions? enough. The idea, in a natural first formulation is that every per-

The answer is that it is inconsistent but it does require a more son (or, in a modern 'sociological' formulation, every culture, or
elaborate argument than the (nonetheless very nice) one-liner sometimes every 'discourse') has his (its) own views, standards,
produced by Garfinkel. The important point to notice is that if presuppositions, and that truth (and also justification) are rela-
all is relative, then the relative is relative too. But this takes a bit tive to these. One takes it for granted, of course, that whether X
of explaining! is true (or justified) relative to these is itself something 'absolute'.
Platowas perhaps the first to employ the sort of argument I Modern Structuralists like Foucault write as if justification rel-

have in mind (against Protagoras). Protagoras (a deep-dyed rel- ative to a discourse is itself quite absolute - i.e. not at all relative.

But if statements of the form X is true

f
ativist, apparently) claimed that when I say X, I really should say (
justified) relative to per-

'I think that X\ Thus when I say 'Snow is white', Protagoras son P' are themselves true or false absolutely, then there is, after
would say that I really mean that Hilary Putnam thinks that all, an absolute notion of truth (or of justification) and not only

snow is white, and that what Robert Nozick means by the same of truth-for-me, truth-for-Nozick, truth-for-you, etc. A total rel-
utterance is that Robert Nozick thinks that snow is white. A ativist would have to say that whether or not X is true relative
more sophisticated statement of the same idea would be that to P is itself relative. At this point our grasp on what the position
when I say 'Snow is white', I am using this utterance to claim even means begins to wobble, as Plato observed.
that snow is white is true-for-me, whereas when Robert Nozick Plato's line of attack on relativism does not seem to have been
says the same words he would normally be claiming that snow followed up until recently. But it was brilliantly extended by
is white is true-for-A/w (or at least he would count his statement Wittgenstein in, of all places, the Private Language Argument
as having been correct just in case it turned out to be true-for- (alluded to in Chapter 3).
him). It follows (on Protagoras' view) that no utterance has the Most commentators read the Private Language Argument as
same meaning for me and anyone else; there is, as we saw
for simply an argument against the 'copy theory' of truth. And Witt-
before, an intimate connection between relativism and incom- genstein's brilliant demonstration that the similitude theory of
mensurability. Plato's counter-argument was that, if every state- reference does not work even for reference to sensations is cer-
ment X means 'I think that X', then I should (on Protagoras' tainly part of a sustained attack on metaphysical realism. But I

view) really say prefer to read the argument as a pair of quite traditional argu-
ments Kant would have approved of both of them!)
(at least
(1) I think that I think that snow is white. against two one a realist position and one a relativist
positions,
position: for the attempt to read the whole argument as an anti-
But the process of adding 'I think' can always be iterated! On
realist one makes it come out looking rather contrived.
Protagoras' view, the ultimate meaning of 'Snow is white' is then
not (1) but The form of relativism Wittgenstein was concerned to attack
is known as 'methodological solipsism'. A 'methodological
(2) I think that I think that I think that I . . . (with solipsist' is a non-realist or 'verificationist' who agrees that truth
infinitely many 'I thinks') that snow is white. is to be understood as in some way related to rational accept-
122 Two conceptions of rationality Two conceptions of rationality 123

ability, but who holds that all justification is ultimately in terms make it rationally acceptable to say that something is true. Can
of experiences that each of us has a private knowledge of. Thus, the relativist interpret statements about what he would believe
I have my knowledge of what experiences of mine would verify under ideal conditions in this non-realist or 'internal' realist
that snow is white and Bob Nozick has his knowledge of what way?
experiences of his would verify that snow is white: every state- Let us recall that the non-realist position, as I described it
ment has a different sense for every thinker. (in Chapter 3), assumes an objective notion of rational accepta-
Wittgenstein's argument seems to me to be an excellent argu- bility. The non-realist rejects the notion that truth is corres-
ment against relativism in general. The argument is that the rel- pondence to a 'ready-made world'. That is what makes him a
ativist cannot, in the end, make any sense of the distinction «ow-(metaphysical)-realist. But rejecting the metaphysical 'cor-
between being right and thinking he is right; and that means that respondence' theory of truth is not at all the same thing as regard-
there is, in the end, no difference between asserting or thinking, ing truth or rational acceptability as subjective. Nelson Good-
on the one hand, and making noises (or producing mental man, who regards truth and rational acceptability as species of
images) on the other. But this means that (on this conception) I a more general predicate 'rightness', applicable to works of art
am not a thinker at all but a mere animal. To hold such a view as well as to statements, has put the point succinctly:
is to commit a sort of mental suicide.
Briefly, then, truth of statements and rightness of descrip-
To see that Wittgenstein was right, let us consider, as Wittgen-
tions, representations, exemplifications, expressions - of
stein does not, how the relativist might attempt to draw the dis-
design, drawing, diction, rhythm - is primarily a matter of
tinction that Wittgenstein denies him, the distinction between
fit: fit to what is referred to in one way or another, or to
being right and thinking he is right.
other renderings, or to modes and manners of organiza-
The might borrow the idea that truth is an idealiza-
relativist
tion. The differences between fitting a version to a world,
tion of rational [Link] might hold that X is true-for-
a world to a version, and a version together or to other
me if 'X is justified-for-me' would be true provided I observed
versions fade when the role of versions in making the
carefully enough, reasoned long enough, or whatever. But sub-
worlds they fit is recognized. And knowing or under-
junctive conditionals of the form 'If I were to ... then I would ,
standing is seen as ranging beyond the acquiring of true
think such-and-such', are, like all statements, interpreted differ-
beliefs to the discovering and devising of fit of all sorts.
ently by different philosophers.
A metaphysical realist can regard statements about what The whole purpose of relativism, its very defining characteristic,
would be the case if as themselves true or false in an absolute is,however, to deny the existence of any intelligible notion of
sense, independently of whether we ever will be justified in objective 'fit'. Thus the relativist cannot understand talk about
accepting or rejecting them. If the relativist interprets statements truth in terms of objective justification-conditions.
about what he would believe under such-and-such conditions in The attempt to use conditionals to explicate the distinction
this realist way, then he has recognized one class of absolute between being right and thinking one is right fails, then, because
truths, and so has given up being a relativist. the relativist has no objective notion of rightness for these con-
A non-realist or 'internal' realist regards conditional state- ditionals any more than he does for any other sort of statement.
ments as statements which we understand (like all other state- Finally, if the relativist of today, like the ancient Protagoras,
ments) in large part by grasping their justification conditions. simply decides to bite the bullet and say that there is no differ-
This does not mean that the 'internal* realist abandons the dis- ence between 'I am right' and 'I think I am right' - that a dis-
tinction between truth and justification, but that truth (idealized tinction between being justified and thinking one is justified can-
justification) is something we grasp as we grasp any other con- not be drawn in one's own case - then what is speaking, on such
cept, via a (largely implicit) understanding of the factors that a conception - beyond producing noises in the hope that one
124 Two conceptions of rationality Two conceptions of rationality 125

will have the feeling of being right? What is thinking - beyond dure, that embraces what is today standard 'second order logic'.
producing images and sentence-analogues in the mind in the The procedure is complete for the elementary theory of deduc-

hope of having a subjective feeling of being right? The relativist tion ('first order logic'). The fact that one can write down an

must end by denying that any thought is about anything in either algorithm for proving all of the valid formulas of first order

a realist or non-realist sense; for he cannot distinguish between logic -an algorithm which requires no significant analysis and
thinking one's thought is about something and actually thinking simulation of full human psychology- is a remarkable fact. It

about that thing. In short, what the relativist fails to see is that inspired the hope that one might do the same for so called

it is a presupposition of thought itself that some kind of objective 'inductive logic' - that the 'scientific method' might turn out to

'rightness' exists. be an algorithm, and that these two algorithms - the algorithm

There an interesting relation between the argument I just

is for deductive logic (which, of course, turned out to be incom-
analyzed (Plato- Wittgenstein) and the argument against incom- plete when extended to higher logic) and the algorithm-to-be-
mensurability I attributed to Quine and Davidson: Quine and discovered for inductive logic - might exhaustively describe or
Davidson argue, in effect, that a consistent relativist should not 'rationally reconstruct' not just scientific rationality, but all

treat others as speakers (or thinkers) at all (if their 'noises' are rationality worthy of the name.
that 'incommensurable', then they are just noises), while Plato When I was just starting my teaching career at Princeton Uni-
and Wittgenstein argue, in effect, that a consistent relativist can- versity I know Rudolph Carnap, who was spending two
got to
not treat himself as a speaker or thinker. 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-
What to make of this?
ate student in physics, studying logic in Frege's seminar. The text
The arguments I just set before you convinced me that the two was Principia Mathematica (imagine studying Russell and
most widely known philosophies of science produced in this cen- Whitehead's Principia with Frege!) Carnap was fascinated with
tury are both incoherent. (Of course, neither of them is just a symbolic logic and equally fascinated with the special theory of
'philosophy of science'.) This naturally led me to reflect on the relativity. So he decided to make his thesis a formalization of

meaning of this situation. How did such views arise? special relativity in the notation of Principia. It was because the

Logical positivism, I recalled, was both continuous with and Physics Department at Jena would not accept this that Carnap
different from the Machian positivism which preceded it. became a philosopher, he told me.

Mach's positivism, or 'empirio-criticism', was, in fact, largely a Today, a host of negative results, including some powerful
restatement of Humean empiricism in a different jargon. Mach's considerations due to Nelson Goodman, have indicated that
brilliance, his dogmatic and enthusiastic style, and his scientific there cannot be a completely formal inductive logic. Some
eminence made his positivism a large cultural issue (Lenin, afraid important aspects of inductive logic can be formalized (although
that the Bolsheviks would be converted to 'empirio-criticism', the adequacy of the formalization is controversial), but there is
wrote a polemic against it). Einstein, whose interpretation of always a need for judgments of 'reasonableness', whether these
special relativity was operationalist in spirit (in marked contrast are built in via the choice of vocabulary (or, more precisely, the

to the interpretation he gave to general relativity), acknowledged division of the vocabulary into 'projectible' predicatesand 'non-
that his criticism of the notion of simultaneity Owed much to projectible' predicates) or however. Today, no onevirtually

Hume and to Mach, although, to his disappointment, Mach believes that there is a purely formal scientific method (on this,

totally rejected special relativity. see Chapter 8).

But the most striking event that led up to the appearance of The story Carnap told me supports the idea that it was the
logical positivism was the revolution in deductive logic. By 1879 sucess of formalization in the special case of deductive logic that
Frege had discovered an algorithm, a mechanical proof proce- played a crucial role. If that success inspired the rise of logical
126 Two conceptions of rationality

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 any case, additional factors
this suggestion; in
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 Fact and value
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 Understood in a sufficiently wide sense, the topic of fact and
of logical positivism is and unashamed; but I think
quite overt value is a topic which is of concern to everyone. In this respect,
there is also a scientism hidden behind relativism. The theory it from many philosophical questions. Most edu-
differs sharply
that all there is to 'rationality' is what your local culture says cated men and women do not feel it obligatory to have an opin-
there is is never quite embraced by any of the 'anarchistic' think- ion on the question whether there really is a real world or only
ers, but it is the natural limit of their tendency: and this is a
appears to be one, for example. Questions in philosophy of lan-
reductionist theory. That rationality is defined by an ideal com- guage, epistemology, and even in metaphysics may appear to be
puter program is a scientistic theory inspired by the exact sci- questions which, however interesting, are somewhat optional
ences; that it is simply defined by the local cultural norms is a from the point of view of most people's lives. But the question
scientistic theory inspired by anthropology. of fact and value is a forced choice question. Any reflective per-
I will not discuss here the expectation aroused in some by son has to have a real opinion upon it (which may or may not
Chomskian psychology will discover
linguistics that cognitive
be the same as their notional opinion). If the question of fact and
innate algorithms which define rationality. I myself think that value is a forced choice question for reflective people, one partic-
this is an intellectual fashion which will be disappointed as the
ular answer to that question, the answer that fact and value are
logical positivist hope for a symbolic inductive logic was disap-
totally disjoint realms, that the dichotomy 'statement of fact or
pointed.
value judgment' is an absolute one, has assumed the status of a
All this suggests that part of the problem with present day cultural institution.
philosophy a scientism inherited from the nineteenth century -
is
By dichotomy a cultural institution, I mean to sug-
calling the
a problem that affects more than one intellectual field. I do not gest that it is an unfortunate fact that the received answer will
deny that logic is important, or that formal studies in confirma- go on being the received answer for quite some time regardless
tion theory, in semantics of natural language, and so on are of what philosophers may say about it, and regardless of
important. do tend to think that they are rather peripheral to
I
whether or not the received answer is right. Even if I could con-
philosophy, and that as long as we are too much in the grip of vince you that the fact-value dichotomy is without rational
formalization we can expect this kind of swinging back and basis, that it is a rationally indefensible dichotomy, or even if
forth between the two sorts of scientism I described. Both sorts some better philosopher than I could show this by an absolutely
of scientism are attempts to evade the issue of giving a sane and conclusive argument (of course there are no such in philosophy),
human description of the scope of reason. still the next time you went out into the street, or to a cocktail
128 Fact and value Fact and value 129

party, orhad a discussion at some deliberative body of which fact his criterion for a successful definition of 'true' was that it

you happen to be a member, you would find someone saying to should yield all sentences of the form T* is true if and only if P,

you, 'Is that supposed to be a statement of fact or a value judg- e.g.

ment?' The view that there is no fact of the matter as to whether (T) 'Snow is white' is true if and only if snow is white
or not things are good or bad or better or worse, etc. has, in a
sense, become institutionalized. as theorems of the meta-language (where P is a sentence of the
The strategy of my argument is not going to be a new one. I'm formal notation in question).
going to rehabilitate a somewhat discredited move in the debate But the equivalence principle is philosophically neutral, and so

about fact and value, namely the move that consists in arguing is Tarski's work. On any theory of truth, 'Snow is white' is

that the distinction is at the very least hopelessly fuzzy because equivalent to '
"Snow is white" is true.'

factual statements themselves, and the practices of scientific Positivist philosophers would reply that if you know (T)

inquiry upon which we rely to decide what is and what is not a above, youknow what " 'Snow white" is true' means: it is

fact, presuppose values. means snow is white. And if you don't understand 'snow' and
The reason this is a somewhat discredited move is that there is 'white', they would add, you are in trouble indeed! But the prob-

an obvious rejoinder to it. The rejoinder to the view that science

lem is not that we don't understand 'Snow is white'; the problem
presupposes values is a protective concession. The defenders of is that we don't understand what it is to understand 'Snow is

the fact— value dichotomy concede that science does presuppose white.' This is the philosophical problem. About this (T) says

some values, for example, science presupposes that we want nothing.

truth, but argue that these values are
not ethical values. I shall And indeed does this not accord with our intuitions about
imagine a somewhat strawman opponent who takes the view these matters? If someone approaches us with a gleam in his eye

that science presupposes one value, namely the value of truth and says, 'Don't you want to know the "Truth"?', our reaction
itself. is generally to be pretty leery of this person. And the reason that

As we have not a simple notion. The idea that

seen, truth is we are leery (apart from the gleam in the eye) is precisely because

truth is a passive copy of what

is 'really' (mind-independently,
someone's telling us that they want us to know the truth tells us
discourse-independently) 'there' has collapsed under the cri- really nothing as long as we have no idea what standards of

tiques of Kant, Wittgenstein, and other philosophers even if it rational acceptability the person adheres to: what they consider

continues to have a deep hold on our thinking. a rational way to pursue an inquiry, what their standards of
Some philosophers have appealed to the equivalence principle, objectivity are, when they consider it rational to terminate an

that the principle that to say of a statement that it is true is

is inquiry, what grounds they will regard as providing good reason
equivalent to asserting the statement, to argue that there are no for accepting one verdict or another on whatever sort of ques-
real philosophical problems about truth. Others appeal to the tion they may be interested in. Applied to the case of science, I

work of Alfred Tarski, the logician who showed how, given a would say that to tell us that science 'seeks to discover the truth'
formalized language formal notation for expressing certain
(a 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
statements, employing symbolic logic), one can define 'true' for
that language in a stronger language (a so-called 'meta- white, that they don't want to assert that there are electrons

language'). 1 flowing through a wire if electrons are not flowing through the

Tarski's work was itself based on the equivalence principle: in wire, and so on. But these purely formal statements are quite
empty as long as we don't have some idea what the system of
1
criteria of rational acceptability is which distinguishes scientific
For a non-technical account of Tarski's work see my Meaning and the
Moral Sciences, Part I, Lecture I.
ways of attempting to determine whether snow is white from
130 Fact and value Fact and value 131

other ways of attempting to determine whether snow is white, economists. (Why should an economist care if all the money in
ways of attempting to determine whether electrons are
scientific the world isn't physically real? Most of it isn't physically real on
flowing through a wire from other ways of attempting to deter- any theory!)
mine whether there are electrons flowing through a wire, and so I want the reader to imagine that this crazy (and, I would
on. claim, incoherent) theory, the theory that we are all brains in a
If the notion of comparing our system of beliefs with uncon- vat, is held not by an isolated lunatic, but by virtually all the

ceptualized reality to see if they match makes no sense, then the people in some large country, say, Australia. Imagine that in
claim that science seeks to discover the truth can mean no more Australia only a small minority of the people believe what we do
than that science seeks to construct a world picture which, in the and the great majority believe that we are Brains in a Vat. Per-
ideal limit, satisfies certain criteria of rational acceptability. That haps the Australians believe this because they are all disciples of
science seeks to construct a world picture which is true is itself a a Guru, the Guru of Sydney, perhaps. Perhaps when we talk to
true statement, an almost empty and formal true statement; the them they say, 'Oh if you could talk to the Guru of Sydney and
aims of science are given material content only by the criteria of look into his eyes and see what a good, kind, wise man he is, you
rational acceptability implicit in science. In short I am saying too would be convinced.' And if we ask, 'But how does the Guru
that the answer to the 'strawman' position I considered, that the of Sydney know that we are brains in a vat, if the illusion is as
only aim of science is to discover truth (besides pointing out that perfect as you say?', they might reply, 'Oh, the Guru of Sydney
science has additional aims, which is of course true), is that truth just knows/
is not the bottom line: from our criteria of
truth itself gets its life As I said before, this is not a scientific disagreement in the

rational acceptability, and these are what we must look at if we ordinary sense. We can imagine that the Australians are just as
wish to discover the values which are really implicit in science. good as we are at anticipating experiences, at building bridges
For the purpose of an example let me now imagine an extreme that stay up (or seem to stay up), etc. They may even be willing
case of disagreement. The disagreement I'm going to imagine is to accept our not as true, but as cor-
latest scientific discoveries,

not an ordinary scientific disagreement, although I hope our rect descriptions of what seems to go on in the image. We may

response to it will enable us to discover something about the or may not imagine that they disagree with us about some pre-
nature of scientific values. dictions concerning the very distant future (for example, they
The hypothesis
that the disagreement is going to be about, in might expect that some day the automatic machinery will break
the case I am
about to describe, is just the hypothesis we dis- down and then people will begin to have collective hallucina-
cussed in Chapter 1, the hypothesis that we are all Brains in a tions of a kind which will give evidence that their view is right), 3

Vat. We have argued that this hypothesis cannot possibly be but whether they do make such predictions, or whether they
true; but we shall suppose that our arguments have failed to commit themselves to no predictions different from the ones
convince one side in this disagreement (which is not improbable, afforded by standard theory, will not affect my argument. The
since philosophical arguments never convince everyone). In point is that here I've imagined a case where a vast number of
short, the hypothesis is that everything is a collective hallucina- people have a self-contained belief system which violently dis-

tion in the way we described before. agrees with ours.

Of course, if it were all one collective hallucination in this There is no question of a disagreement in 'ethical' values here;
way, there are many people to whom this need not make any
3
difference. It would make little or no difference to lovers, for If do make such predictions, then it does make this much difference:
they
their viewis no longer incoherent in the way we criticized in Chapter
example. 2 And I imagine it would make no difference at all to
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
2
But I keep changing my mind about whether it would or not. independent of justification) to be understood.
132 Fact and value Fact and value 133

the Australians can have ethics just as similar to ours as you like. thing that we have an algorithm for, but something that we ulti-

(Although an ancient Greek would have said that being wise is mately judge by 'seat of the pants' feel). The Australians, remem-
an ethical value; Judaism and Christianity have, in fact, nar- ber, have themselves postulated an illusion so perfect that there
rowed the notion of the ethical because of a certain conception is rational way in which the Guru of Sydney can possibly
no
of Salvation.) know that the belief system which he has adopted and persuaded
The first thing I to observe about the hypothetical Aus-
want all the others to adopt is correct. Judged by our standards of

tralians is world view is crazy* Sometimes, to be sure,

that their coherence, their belief system is totally incoherent.
used almost as a term of approval; but I don't mean it Other methodological virtues could be listed which their belief
(
crazy' is

in that sense here. I think we would regard a community of system lacks. Their belief system, as I described it, agrees with
human beings who
held so insane a world view with great sad- ours concerning what the laws of nature are in the image; but
ness. The Australians would be regarded as crazy in the sense of does it tell us whether or not the laws of nature that appear to
having sick minds; and the characterization of their minds as hold in the image are the laws of nature that actually hold out-
sick is an ethical one, or verges on the ethical. But how, other side the vat? If it fails to, then it lacks a certain kind of compre-

than by calling them names, could one argue with the Austra- hensiveness which we aim after, for it does not, even in its own
lians? (Or try to argue with them, for I shall suppose that they terms, tell what the true and ultimate laws of nature are. Cer-
are not to be convinced.) tainly it violates Ockham's razor. Again, Ockham's razor seems

One argument that one can immediately think of has to do difficult or impossible to formalize as an algorithm, but the very

with the incoherence of their view. I don't just mean the inco- fact that the Brain in a Vatist theory postulates all kinds of
herence that we found in the view in Chapter 1. That is a deep objects outside the vat which play no role in the explanation of

incoherence, which requires a philosophical (and hence contro- our experiences, according to the theory itself, makes it clear
versial) argument to expose. But the Australian's view is inco- that this is a case in which we can definitely say that the maxim
herent at amuch more superficial level. One of the things that . .'don't multiply entities without necessity' is violated. Let us
.

we aim at is that we should be able to give an account of how call a theory which obeys Ockham's razor, in spirit as opposed

we know our statements to be true. In part we try to do this by to just in letter, functionally simple.
developing a causal theory of perception, so that we can account What I have been saying is that the procedures by which we
for what we take to be the reliability of our perceptual knowl- decide on the acceptability of a scientific theory have to do with
edge, viewed from within our theory itself, by giving an account whether or not the scientific theory as a whole exhibits certain
within the theory of how our perceptions result from the opera- 'virtues'. I am assuming that the procedure of building up scien-

tion of transducing organs upon the external world. In part we tific theory cannot be correctly analyzed as a procedure of veri-

try to do this by a theory of statistics and experimental design, fying scientific theories sen tence by sentence. I am assuming that
so that can show, within our theory itself, how the proce-
we verification in science is a holistic matter, that it is whole theo-

dures that we take to exclude experimental error really do have retical systems that meet the test of experience 'as a corporate

a tendency in the majority of cases to exclude experimental body', and that the judgment of how well a whole system of
error. In short, it is an important and extremely useful constraint sentences meets the test of experience is ultimately somewhat of
on our theory itself that our developing theory of the world an intuitive matter which could not be formalized short of for-
taken as a whole should include an account of the very activity malizing total human psychology. But let us come back to our
and processes by which we are able to know that that theory is original question. What are the values implicit in science?
correct. I've been arguing that if we take the values to which we appeal
The Australians' system, however, does not have this property in our criticism of the Brain-in-a-Vatists, and add, of course,
of coherence (at least as we judge it, and 'coherence' is not some- other values which are not at issue in this case, e.g. our desire for
134 Fact and value Fact and value 135

instrumental efficacy, which we presumably share with the short, I am saying that the 'real world' depends upon our values
Brain-in-a-Vatists, then we get a picture of science as presuppos- (and, again, vice versa).

ing a rich system of values. The fact is that, if we consider the

ideal of rational acceptability which is revealed by looking at
At least some values must be objective
what and ordinary people consider rational to
theories scientists
accept, then we see that what we are trying to do in science is to The fact that science is not 'value neutral', as has been thought,
construct a representation of the world which has the character- does not, to be sure, show that 'ethical' values are objective, or
istics of being instrumentally efficacious, coherent, comprehen- that ethics could be a science. In fact, there is no prospect of a
sive, and functionally simple. But why? 'science'of ethics, whether in the sense of a laboratory science or
I would answer that the reason we want this sort of represen- of a deductive science. As Aristotle long ago remarked, 4
tation, and not the 'sick' sort of notional world possessed by the
Australians, possessed by the Brain-in-a-Vatists, is that having
We must be content, then, in speaking of such subjects
this sort of representation systemis part of our idea of human
and with such premisses to indicate the truth roughly and
in outline, and in speaking about things which are only
cognitive flourishing,and hence part of our idea of total human
for the most part true, and with premisses of the same
flourishing, of Eudaemonia.
kind, to reach conclusions which are no better. In
Of course, if metaphysical realism were right, and one could
the same spirit, therefore, should each kind of statement
view the aim of science simply as trying to get our notional
world to 'match' the world in itself, then one could contend that 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
we are interested in coherence, comprehensiveness, functional
simplicity, and instrumental efficicacy only because these are nature of the subject admits; it is evidently foolish to

instruments to the end of bringing about this 'match'. But the accept probable reasoning from a mathematician and to

notion of a transcendental match between our representation demand from a rhetorician scientific proofs.

and the world in itself is nonsense. To deny that we want this But the fact that rational acceptability in the exact sciences
kind of metaphysical match with a noumenal world is not to (which are certainly central examples of rational thinking) does
deny that we want the usual sort of empirical fit (as judged by depend on such cognitive virtues as 'coherence' and 'functional
our criteria of rational acceptability) with an empirical world. simplicity' shows that at least some value terms stand for prop-
But the empirical world, as opposed to the noumenal world, erties of the things they are applied to, and not just for feelings
depends upon our criteria of rational acceptability (and, of of the person who uses the terms.
course, vice versa). We use our criteria of rational acceptability If the terms 'coherent' and 'simple' do not stand lot properties
to build up a theoretical picture of the 'empirical world' and then
of theories, not even fuzzy or imperfectly defined ones, but only
as that picture develops we revise our very criteria of rational for 'attitudes' that some people have towards theories, then such
acceptability in the light of that picture and so on and so on
terms for rational acceptability as 'justified', 'well confirmed',
forever. The dependence of our methods on our picture of the
'best of the available explanations' must also be entirely subjec-
world is something I have stressed in my other books; what I tive: for rational acceptability cannot be more objective than the
wish to stress here is the other side of the dependence, the depen- parameters upon which it depends. But, as we argued in the pre-
dence of the empirical world on our criteria of rational accepta- ceding chapter, the view that rational acceptability itself is sim-
bility. What I am saying is that we must have criteria of rational
ply subjective is a self-refuting one. So we are compelled to con-
acceptability to even have an empirical world, that these reveal
4
part of our notion of an optimal speculative intelligence. In Ethica Nicomacbea, Book I, Ch. 3.
 Fact and value Fact and value 137
136

search for better conceptions of rationality as an intentional

elude that at least these value-terms have some sort of objective
application, some sort of objective justification conditions.
human activity, which, like every activity that rises above habit
and the mere following of inclination or obsession, is guided by
Of course, one might attempt to avoid conceding that there
our idea of the good.
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 Rationality in other areas
that there is no objective Tightness about our doing so. But this
have If the values implicit in science, especially in the exact sciences,
line runs into difficulties at once. 'Coherent' and 'simple'
reveal a part of our idea of the good, I think that the rest of our
too many characteristics in common with the paradigmatic value
words. Like 'kind', 'beautiful', and 'good', 'coherent' and 'sim-
idea of the good can be read from our standards of rational
off
coher- acceptability in yet other areas of knowledge. At this point, how-
ple' are often used as terms of praise. Our conceptions of
justification are just as historically condi- ever, it is necessary to broaden the notion of standards of
ence, simplicity, and
rational acceptability.
tioned as our conceptions of kindness, beauty, and goodness;
these epistemic terms figure in the same sorts of perennial
So far, we have only considered standards of rational accept-

philosophical controversies as do the terms for ethical and aes-

ability in the literal sense: standards which tell us when we
thetic values. The conception of rationality of a John Cardinal
should and when we should not accept statements. But stan-

obviously quite different from that of a Rudolf Car- dards of rationality in the wide sense have to do not only with
Newman is

nap. It is highly unlikely that either could have convinced the

how we judge the truth or falsity of systems of statements, but

other, had they lived at the same time and been able to meet.
also with how we judge their adequacy and perspicuousness.
the rational conception of rationality There are ways — purely cognitive ways — in which a system of
The question: which is

way an
that the justification of statements can fall short of giving us a satisfactory description
itself is difficult in exactly the
There no neutral conception of other than by being false.
ethical system is difficult. is

rationality to which to appeal. Had I chosen I could have made this point even in connection
attempt various conventionalist moves here, e.g. with theoretical science. I could have pointed out that the con-
One might
cern of exact science is not just to discover statements which are
saying that 'justifiedcamap' is one 'property' and justified NeW man|
'subjective value judgment' true, or even statements which are true and universal in form
is a different 'property', and that a
mean 'justifiedcamap' or 'justi- ('laws'), but to find statements that are true and relevant And
is involved in the decision to
the notion of relevance brings with it a wide set of interests and
fied Newm a„' by the word 'justified'
but that no value judgment
that a given statement S is justi- values. But this would have only been to argue that our knowl-
is involved in stating the fact
But from whose standpoint is the edge of the world presupposes values, and not to make the more
fied or
Carnap justified Newma „.
radical claim that what counts as the real world depends upon
word facf being used? If there is no conception of rationality one

ought to have, then the notion of a 'fact' is empty. our values.

objectively
Without the cognitive values of coherence, simplicity, and When we come to perceptual rationality, that is to the implicit

we have no world and no 'facts', not even standards and skills on the basis of which we decide whether
instrumental efficacy
facts about what is so relative to what, for those are in the same someone is able to give a true, adequate, and perspicuous

boat with all other facts. And these cognitive values are arbitrary account of even the simplest perceptual facts, then we see a large

considered as anything but a part of a holistic conception of

number of factors at play. Recently psychologists have stressed
just how much theory construction is involved in even the sim-
human flourishing. Bereft of the old realist idea of truth as 'cor-
respondence' and of the positivist idea of justification as fixed by
plest cases of perception. Not only is this true at the neurophys-

public 'criteria', we are left with the necessity of seeing our iological level, but it is also true at the cultural level. Someone
138 Fact and value
Fact and value 139

from a culture which had no furniture might be able to come blame and may be used without any intention of blaming at all.
into a room and give some kind of description of the room, but, (Sometimes one has a perfect right to be jealous.)
if he did not know what a table or a chair or a desk was, his
The use of the word 'inconsiderate' seems to me a very fine
description would hardly convey the information that a member
example of the way in which the fact/value distinction is hope-
of this culture would wish to have about the room. His descrip-
lessly fuzzy in the real world and in the real language. The
tion might consist only of true statements but it would not be
importance of terms like 'inconsiderate', 'pert', 'stubborn',
adequate.
'pesky', etc., in actual moral evaluation, has been emphasized by
What this simple example shows is that the requirement that Iris Murdoch in The Sovereignty of 'Good 5 Even though each
9
,

a description be adequate is implicitly a requirement that the

of the statements 'John is a very inconsiderate man', 'J onn
describer have available a certain set of concepts; we expect thinks about nobody but himself, 'John would do practically
rational describers with respect to certain kinds of descripta to
anything for money' may be simply a true description in the
be capable of acquiring certain concepts and of seeing the need
most positivistic sense (and notice 'John would do practically
to use them; the fact that the describer did not employ a certain anything for money' does not contain any value term), if one has
concept may be a ground for criticizing both him and his asserted the conjunction of these three statements it is hardly
description.
necessary to add 'John is not a very good person'. When we
What is true at the simple level of talk about tables and chairs think of facts and values as independent we typically think of
in aroom without people in it is also true at the level of descrip-
some
'facts' as stated in and
physicalistic or bureaucratic jargon,
tion of interpersonal relations and situations. Consider the terms the 'values' as being stated in the most abstract value terms, e.g.
we use every day in describing what other people are like, e.g. 'good', 'bad'. The independence of value from fact is harder to
considerate or inconsiderate. Considerate and inconsiderate may
maintain when the facts themselves are of the order of 'incon-
of course be used to praise or blame; and one of the many dis-
siderate', 'thinks only about himself, 'would do anything for
tinctions which have gotten confused together under the general
money'.
heading 'fact— value distinction' is the distinction between using
Just as we a describer who does not employ the con-
criticize
a linguistic expression to describe and using that linguistic
cepts of table and chair when their use is called for, so also,
expression to praise or blame. But this distinction is not a dis-
someone who fails to remark that someone is considerate or
tinction which can be drawn on the basis of vocabulary. The
spontaneous may open himself to the criticism that he is imper-
judgment that someone is inconsiderate may indeed be used to ceptive or superficial; his description is not an adequate one.
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-
The super-Benthamites
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 Let me go back and modify my previous example of the 'Brain-
matter what he feels about you he'll likely behave in such a way in-a-Vatists'. This time let us
imagine that the continent of Aus-
that your feelings will be hurt. He's just a rather inconsiderate tralia is peopled by a culture which agrees with us on history,

man, but don't think that it has anything to do with you.'

geography and exact science, but which disagrees with us in eth-
In this little imaginary speech someone is using the word ics. I don't want to take the usual case of super- Nazis or some-
'inconsiderate' not for the purpose of blaming Jones, but with thing of that kind, but I want to take rather the more interesting

the intention of predicting and explaining Jones' behavior to case of super-Benthamites. Let us imagine that the continent of

someone else. And both the prediction and the explanation may Australia is peopled with people who have some elaborate sci-

be perfectly correct. And similarly, 'jealous' may be a term of s

Routledge and Kegan Paul, 1970.
140 Fact and value Fact and value 141

entific measure of what they take to be 'hedonic tone', and who tions will be quite different from the vocabulary available to us.

believe that one should always act so as to maximize hedonic Not only will they lack, or have altered beyond recognition,
tone (taking that to mean the greatest hedonic tone of the great- many of our descriptive resources, but they will very likely
est number). I will assume that the super-Benthamites are invent new jargon of their own (for example, exact terms for
extremely sophisticated, aware of all the difficulties of predicting describing hedonic tones) that are unavailable to us. The texture
the future and exactly estimating the consequences of actions of the human world will begin to change. In the course of time
and so forth. I will also assume that they are extremely ruthless, the super-Benthamites and we will end up living in different
and that while they would not cause someone suffering for the human worlds.
sake of the greatest happiness of the greatest number if there In short, it will we and the super-Bentham-
not be the case that
were reasonable doubt that in fact the consequence of their ites 'agree on the facts and disagree about values'. In the case of
action would be to bring about the greatest happiness of the almost all interpersonal situations, the description we give of the
greatest number, that in cases where one knows with certainty facts will be quite different from the description they give of the
what the consequences of the actions would be, they would be facts. Even if none of the statements they make about the situa-
willing to perform the most horrible actions - willing to torture tion are false, their description will not be one that we will count
small children or to condemn people for crimes which they did as adequate and perspicuous; and the description we give will
not commit - if the result of these actions would be to increase not be one that they could count as adequate and perspicuous.
the general satisfaction level in the long run (after due allowance In short, even if we put aside our 'disagreement about the val-
for the suffering of the innocent victim in each case) by any pos- ues', we could not regard their total representation of the human
itive e, however small. world as fully rationally acceptable. And just as the Brain-in-a-
I imagine that we would not feel very happy about this sort of Vatists' inability to get the way the world is right is a direct result
super-Benthamite morality. Most of us would condemn the of their sick standards of rationality - their sick standards of
super-Benthamites as having a sick system of values, as being theoretical rationality - so the inability of the super-Benthamites
bureaucratic, as being ruthless, etc. They'new man' in
are the to get the way the human world is right is a direct result of their
his most horrible manifestation. And they would return our sick conception of human flourishing.
invective by saying that we are soft-headed, superstitious, pris-
oners of irrational tradition, etc.
Subjectivism about goodness
The disagreement between us and the super-Benthamites is

just the sort of disagreement that is ordinarily imagined in order It has often been claimed that the step from 'John is considerate,
to make the point that two groups of people might agree on all truthful, kind, courageous, responsible, etc' to'John is morally
the facts and still disagree about the 'values'. But let us look at good' involves at least one unproved (and unprovable) 'premiss',
the case more closely. Every super-Benthamite is familiar with namely, 'Consideration is morally good.' And it has been held
the fact that sometimes the greatest satisfaction of the greatest that the need for moral 'premisses' before one can draw moral
number (measured in 'utils') requires one to tell a lie. And it is conclusions from 'factual' statements shows that ethical state-
not counted as being 'dishonest' in the pejorative sense to tell lies ments are not rationally justifiable.

out of the motive of maximizing the general pleasure level. So This picture of ethics as a sort of inverted pyramid, with the
after a while the use of the description 'honest' among the super- tip (which is itself unsupported) consisting of 'ethical axioms'
Benthamites would from the use of that
be extremely different which support our whole body of moral belief and thinking, is
same descriptive term among us. And the same will go for 'con- naive. No one has ever succeeded in imposing an axiomatic
siderate', 'good citizen', etc. The vocabulary available to the structure upon ethics (as Aristotle remarked in the passage I
super-Benthamites for the description of people-to-people situa- cited a few pages ago, such moral maxims as we are able to list
142 Fact and value Fact and value 143

are almost always true only 'for the most part'). And the same values are objective; but the argument that there can't be any

trick, of picturing abody of thinking one wishes to cast into objective values at all has been refuted.
doubt as resting upon unsupportable 'axioms' is one which scep-
tics have employed in every area. Sceptics who doubt the exis-
In order to show what is wrong with arguments
for moral

tence of material objects, for example, argue that the principle

subjectivism, I must now arguments that were used
recall the

that 'if our sensations occur as they would if there were a mate- against metaphysical realism in Chapter 2. This may seem queer:
isn't subjectivism the opposite of metaphysical realism? If one
rial world, then there probably is a material world' is a rationally
unsupportable premiss which we tacitly invoke whenever we thinks so, then it will seem that any argument against metaphys-

claim to 'observe' a material object, or try otherwise to justify ical realism must support subjectivism; the strategy I am going

belief in their existence. In fact, ethics and mathematics and talk to follow of using the same argument against both metaphysical

of material objects presuppose concepts not 'axioms'. Concepts realism and subjectivism will seem an impossible one.

are used in observation and generalization, and are themselves But in fact, metaphysical realism and subjectivism are not sim-
made legitimate by the success we have in using them to describe ple 'opposites'. Today we tend to be too realistic about physics

and generalize. and too subjectivistic about ethics, and these are connected ten-
A more sophisticated attack on the idea of ethical objectivity dencies. It is because we are too realistic about physics, because

concedes that our ethical beliefs rest on observations of specific we see physics (or some hypothetical future physics) as the One
cases, 'intuitions', general maxims, and not on some collec-
etc.,
True Theory, and not simply as a rationally acceptable descrip-
tion of arbitrary 'ethical axioms', but makes the charge that eth- tion suited for certain problems and purposes, that we tend to be

ical 'observation' itself is infected with an incurable disease: pro- subjectivistic about descriptions we cannot 'reduce' to physics.

jection.
Becoming less realistic about physics and becoming less subjec-
tivistic about ethics are likewise connected.
According to this account, humans are naturally, if intermit-
tently, compassionate. So when we see something terrible hap-
The argument at the end of Chapter 2 was directed against the
pening, as it might be, someone torturing a small child just for 'physicalistic' or naturalistic version of metaphysical realism. To
his own sadistic pleasure, we are (sometimes) horrified. But the recall it, let us suppose that the standard interpretation J (Under
psychological mechanism of 'projection' leads us to experience which 'cat' refers to cats, 'cherry' to cherries, etc.) is either coex-
the feeling quality as a quality of the deed itself: we say 'the act tensive with or identical with physicalistic relation R. So R
was horrible' when we should really say 'my was
reaction to be holds between tokens of 'cat' (or physical events of someone's

horrified'. Thus we build up a body of what we take to be 'ethi- using those tokens suitably) and cats, etc. The non-standard
cal observations', which are really just observations of our own interpretation/ we described will then also be co-extensive with
subjective ethical feelings. a certain relation R ', definable in terms of R and the possible
This story has more sophisticated forms (like any other). worlds and permutations used in constructing/ (see Appendix).
Hume postulated a human tendency he called 'sympathy', which So R' holds between tokens of 'cat' (or the physical events of
has gradually become wider under the influence of culture. Con- someone's using those tokens in the standard way) and cherries,
etc. R and R are both 'correspondences': The same sentences
r

temporary sociobiologists postulate an instinct they call 'altru-

ism', and speak of 'altruistic genes'. But the key idea remains the are 'true' under both correspondences. The actions called for by

same: there are ethical feelings, but no objective value proper- the R '-truth of a sentence (i.e. the actions which will 'succeed',

ties.
from the agent's point of view) are the same as the actions called

We have already seen that this is not right: there are at least for when the sentence is It -true. If R is 'identical with reference';
some objective values, for example, justification. It could be still if R, R', and all the other relations which assign extensions to
claimed that the ethical values are subjective while the cognitive our words in ways which satisfy our operational and theoretical
144 Fact and value Fact and value 145

constraints are not equally correct; if R, R' and the others are So moral knowledge becomes problematical; perhaps
tific'.

not equally correct because one of them - R - just is reference; downright impossible.
then that fact itself is an inexplicable fact from a physicalist per- But what does 'unscientific' mean here? A belief that there is
spective. such a thing as justice is not a belief in ghosts, nor is a 'sense of
This argument is not just an argument against (the physicalist justice' apara -normal sense which enables us to perceive such
version of) metaphysical realism, but an argument against ghosts. Justice is not something anyone proposes to add to the
reductionism. If there is nothing in the physicalist world-picture list of objects recognized by physics as eighteenth-century chem-
that corresponds to the obvious fact that 'cat' refers to cats and istsproposed to add 'phlogiston' to the list of objects recognized
not to cherries, then that is a decisive reason for rejecting the by chemical theory. Ethics does not conflict with physics, as the
demand that all the notions we
use must be reduced to physical term 'unscientific' suggests; it is simply that 'just' and 'good' and
terms. For reference and truth are notions we cannot consis- 'sense of justice' are concepts in a discourse which is not reduc-
tently give up. If I think 'a cat is on a mat', then I am committed ible to physical discourse. As we have just seen, other kinds of
to believing that 'cat' refers to something (though not to a meta- essential discourse are not reducible to physical discourse and
physical realist account of 'reference') and to believing that 'a are not for that reason illegitimate. Talk of 'justice', like talk of
cat is on a mat' is true (though not to a metaphysical realist 'reference', can be wow-scientific without being unscientific.
account of truth).
Having reviewed the argument of Chapter 2, let us now see As a way of seeing what is going on, let us consider any basic
how it bears on the arguments for moral subjectivism. The 'pro- principle of logic or mathematics, say, the principle that the
jection' theory gave one account of moral experience: moral series of whole numbers can always be continued ('every number
experience is, so to speak, mislocated subjective feeling. Contrast has a successor'), or the principle that a non-empty set of whole
the 'projection' theory with the following account: 'all humans numbers must contain a smallest member. Suppose someone put
have, to some and some idea of the
extent, a sense of justice forward the following view: 'These principles are true for the
good. So we respond (intermittently) to such appeals as "be kind numbers and sets of numbers we deal with in practice. So they
to the stranger among you, because you know what it was like come to seem necessary. By the mechanism called "projection",
to be a stranger in Egypt". Our sympathy becomes broader, we attach this feeling of necessity to the principles themselves;
partly because we ought to be broader; we
are persuaded that it we statements have a mysterious "necessity". But in real-
feel xhz

feel that an atrocity is wrong (sometimes) even when we don't ity this has no justification. For all we know, these principles
easily or spontaneously find the victim a person we can sympa- may not even be true.'

thize with. We come to see similarities between injuries to others Virtually no one would agree with this. Virtually every math-
and injuries to ourselves, and between benefits to others and ematician would say, instead, something like this: 'Most humans
benefits to ourselves. We invent moral words for morally rele- have mathematical intuition to some extent. So we intuitively
vant features of situations, and we gradually begin to make "see", or can be brought by examples (or by skillful questioning,
explicit moral generalizations, which lead to still further refine- like the slave-boy in Plato's dialogue) to "see" that the principles
ment of our moral notions, and so on.' are necessarily true.'
This account is, on the face of it, simpler and more sophisti- Kurt Godel believed that 'mathematical intuition' was analo-
cated than the 'projection' theory. (For one thing, it acknowl- gous to perception; mathematical objects (which he called 'con-
edges the role of argument in shaping moral attitudes.) Never- cepts') are out there, and our intuition enables us to intellectually
theless, many intelligent people feel that today we must reject perceive these Platonic entities; but few mathematicians would
talk of a 'sense of justice' and talk of 'having an idea of the good' commit themselves to such a Platonic metaphysics. Godel's com-
(where this is not taken in a purely subjective sense), as 'unscien- parison of mathematical intuition to perception reveals an over-
 .

146 Fact and value Fact and value 147

simple idea of perception. Vision does not really give us direct or of reference and understanding, is not reducible to the lan-
access to a ready made world, but gives us a description of guage or the world-picture of physics. That does not mean phys-
objects which are partly structured and constituted by vision ics is 'incomplete'. Physics can be 'complete' - that is, complete
itself. If we take the physicist's rainbow to be the rainbow 'in for physical purposes. The completeness physics lacks is a com-
itself, then the rainbow 'in itself has no bands (a spectroscopic pleteness all particular theories, pictures, and discourses lack.

analysis yields a smooth distribution of frequencies); the red, For no theory or picture is complete for all purposes. If the irre-
orange, yellow, green, blue and violet bands are a feature of the ducibility of ethics to physics shows that values are projections,

perceptual rainbow, not the physicist's rainbow. The perceptual then colors are also projections. So are the natural numbers. So,
rainbow depends on the nature of our perceptual aparatus itself, for that matter, is 'the physical world'. But being a projection in

on our visual 'world making' as Nelson Goodman has termed it. this sense is not the same thing as being subjective. 6

(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 Authoritarianism and pluralism
as defective because it sees bands in the rainbow; someone who I have been arguing that it is necessary to have standards of
couldn't see them would have defective vision. Vision is certified rational acceptability in order to have a world at all; either a
as good by its ability to deliver a description which fits the world of 'empirical facts' or a world of 'value facts' (a world in
objects for us, not metaphysical things-in-themselves. Vision is
which there is beauty and tragedy). It should go without saying
good when it enables us to see the world 'as it is' — that is, the that it is not possible both to have standards of rational accept-
human, functional world which is partly created by vision itself. ability and not to accept them, or to stand at arm's length from
A proposed new axiom of set theory, such as the 'Axiom of them. (The kind of scepticism which consists in refusing to have
choice', may be adopted partly because of its agreement with the any standards of rational acceptability commits one to not hav-
'intuition' of expert mathematicians, and partly for its yield. ing any concepts at all. As Sextus Empiricus recognized, that
If the axiom of choice did not deliver results which count as kind of empiricism ultimately is unexpressible in language.) We
successful mathematics the fact that some people find it 'intu- have just as much right to regard some 'evaluational' casts of
itive' would have little interest. Mathematical intuition itself is
mind as sick (and we all do) as we do to regard some 'cogni-
demonstrated or tested by grasping mathematical principles and tional' casts of mind as sick. But to say this is not to reject plu-
by following proofs. In short, mathematical intuition is good ralism or to commit oneself to authoritarianism.
when it enables us to see mathematical facts 'as they are' - that Even in science, holding that science is an objective enterprise
is, as they are in the mathematical world which is constructed by
(by a standard of 'objectivity' which is admittedly anthropocen-
human mathematical practice (including the application of tric, but, as David Wiggins once remarked, the only standard of
mathematics to other subject matters).
A physiological or psychological description of vision cannot 6
An unintentionally funny version of the projection theory
is cited by

C. Lewis in The Abolition of Man (Macmillan, 1947). Lewis quotes a

S.
tell us whether seeing bands in the rainbow counts as seeing 'cor-
secondary school English text {which he does not identify, out of
rectly' or not. Even less could a physiological or psychological charity). 'You remember that there were two tourists present [Lewis is
description of the brain-process which goes on when one 'grasps' talking about the well known story of Coleridge at the waterfall]: that
one called it "sublime" and the other "pretty"; and that Coleridge
the Principle of Mathematical Induction tell us whether that
mentally endorsed the first judgment and rejected the second with
principle is true or not. Once one sees this, it should be no sur- disgust. Gaius and Titius [Lewis' pseudonyms for the unidentified

prise that a description of the brain process which goes on when authors of the text] comment as follows: "When the man said That is
sublime, he appeared to be making a remark about the waterfall
one 'sees' that an action is unjust cannot tell us whether the
. .

Actually he was not making a remark about the waterfall, but about his
action really is unjust. own feelings. What he was saying was really / have feelings in my mind
'

Talk of moral 'perception', like talk of mathematical intuition, associated with the word 'Sublime', or shortly, I have sublime feelings."
148 Fact and value Fact and value 149

objectivity available to us) is not to hold that every scientific moral authoritarianism. It is perhaps this confusion that has lead
question has a determinate answer. Some scientific questions one outstanding philosopher 7 to espouse what he himself
may have objectively indeterminate answers, i.e. there may be no regards as a limited version of 'non-cognitivism', and to say
convergence with respect to an answer to them even in the ideal 'Concerning what "living most fully" is for each man, the final
limit of scientific inquiry; and some scientific questions may have authority must be the man himself.' (Notice the ambiguity in
determinate but context-relative answers (e.g. 'What was the 'the final authority': does he mean the final political authority?
cause of John's heart attack?' may have different correct answers The final epistemological authority? Or does he mean that there
depending on who asks the question and why). And, similarly, is no fact of the matter, as his use of the term 'non-cognitivism'

holding that ethical inquiry is objective in the sense that some suggests?) Respect for persons as autonomous moral agents
'value judgments' are definitely true and some are definitely requires that we accord them the right to choose a moral stand-
false, and more generally that some value positions (and some point for themselves, however repulsive we may find their
'ideologies') are definitely wrong, and some are definitely infe- choice. According to the philosophy of political liberalism, it

rior to some others, is not the same thing as holding the silly also requires that we also insist the government not preempt
position that there are no indeterminate cases at all. (One espe- individual moral choices by setting up a state religion or a state
cially important kind of indeterminate case has been emphasized morality. But diehard opposition to all forms of political and
by Bernard Williams: this is the case where all the alternatives moral authoritarianism should not commit one to moral relativ-
are so horrible that there is no one of the alternatives that would ism or moral scepticism. The reason that it is wrong for the gov-
clearly be chosen by an ideally rational and wise person.) And ernment to dictate a morality to the individual citizen is not that
that there are context relativities in ethics goes without saying. there is no fact of the matter about what forms of life are fulfill-
today we differ with Aristotle it is in being much more plu-
If ing and what forms of life are not fulfilling, or morally wrong in
ralisticthan Aristotle was. Aristotle recognized that different some other way. (If there were no such thing as moral wrong,
ideas of Eudaemonia, different conceptions of human flourish- then it would not be wrong for the government to impose moral
ing, might be appropriate for different individuals on account of choices.) The fact that many people fear that if they concede any
the difference in their constitution. But he seemed to think that sort of moral objectivity out loud then they will find some gov-
ideally there was some
sort of constitution that every one ought ernment shoving its notion of moral objectivity down their
to have; that in an ideal world (overlooking the mundane ques- throats is without question one of the reasons why so many peo-
tion of who would grow the crops and who would bake the ple subscribe to a moral subjectivism to which they give no real
bread) everyone would be a philosopher. We agree with Aris- assent.
totle that different ideas of human flourishing are appropriate
7
David Wiggins in 'Truth, Invention, and the Meaning of Life', Proceed-
for individuals with different constitutions, but we go further
ings of the British Academy, vol. LXII, 1976.
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
 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
7 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
Reason and history 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
With the rise of science has come the realization that many ques- of poetry'. Prejudice aside, pushpin is as good as poetry.
tions cannot be settled by the methods of the exact sciences, ide- What makes this so shocking to the modern reader is how
ological and ethical questions being the most obvious examples. deeply it with our current cultural values. The arts have
conflicts
And with the increase in our admiration and respect for the been exalted by us to a place much higher than any they occu-
physicist, the cosmologist, the molecular biologist, has come a pied in Plato's day or in the middle ages. As a number of authors
decrease in respect and trust for the political thinker, the moral- have remarked, for a certain sort of educated person, art today
ist, the economist, the musician, the psychiatrist, etc. is religion, i.e. the closest thing to salvation available.
In this situation some have gone with the cultural tide and Bentham is saying that a preference for 'the arts and sciences
argued that, indeed, there is no knowledge to be found outside of music and poetry' over the childish game of pushpin is merely
of the exact sciences (and the social sciences to the extent that cream over chocolate
subjective, like a preference for vanilla ice
they succeed in aping the exact sciences, and only to this extent). ice cream. He does not wish to deny that music and poetry do
This view may take the form of positivism or materialism, or have greater value than pushpin ('prejudice aside' is an impor-
some combination of these. Others have tried to argue that sci- tant part of the sentence); in the context of his Utilitarianism,
ence too is 'subjective' and arbitrary - this is the popular reading the very fact that a large majority do prefer music and poetry to
of Kuhn's immensely successful book The Structure of Scientific pushpin gives music and poetry greater 'utility' and hence
Revolutions, even if it is not the one Kuhn now says he intended. greater value. But the value is, as it were, the product of 'preju-
Others - e.g. the Marxist philosophers and the religious philos- dice' (i.e. purely subjective interest); there is no fact of the matter
ophers - adopt a sort of double-entry bookkeeping, leaving about the relative value of pushpin and poetry apart from the
technical questions to the exact sciences and engineering and ide- fact that people prefer poetry to pushpin. We don't prefer poetry
ological or ethical questions to a different tribunal: the Party, to pushpin because poetry has greater value than pushpin, Ben-
the Utopian future, the church. But few can feel comfortable tham is saying, rather, it's the other way around, and poetry has
with any of these stances - with extreme scientism in either its greater value than pushpin because people prefer poetry to push-
positivist or materialist forms, with subjectivism and radical rel- pin. (For no reason apparently.)
ativism, or with any of the species of double-entry bookkeeping. Stating the position so baldly already makes it look implausi-
It is just because we feel uncomfortable that there is a real prob- ble. Let us consider for the moment a really 'subjective' prefer-
lem for us in this area. ence.
To be sure, the problem is in one way unreal. The same person One model that people sometimes seem to have in mind for
152 Reason and history Reason and history 153

subjective preference is this. There is something C which is the objective (or warranted) value judgments at all, very likely we
taste of chocolate ice cream; there is something V which is the think some hotly disputed judgments are objectively right. The
taste of vanilla ice cream. There are two feelings L, D which are Nazis disputed the judgment that wanton killing of Jews just
'liking' and 'disliking'. And what goes on and all that goes on, because of their racial affiliation is wrong, but anti-Nazis did not
when Jones likes and Smith dislikes vanilla (and Smith likes and regard their disagreement with the Nazis over this judgment as
Jones dislikes chocolate), is that Jones experiences V + L when 'subjective'. Those who think homosexuals should have full

he eats vanilla and C +D when he eats chocolate whereas Smith rights in our society violently disagree with those who think

experiences V + D when he eats vanilla and C + L when he eats homosexual activity or civil rights of homosexuals should be
chocolate. legally proscribed; but neither side in this dispute regards its own
Such an account is naive psychologically, however, as Kohler position as 'subjective'. Indeed, disagreement frequently makes
long ago argued. What vanilla tastes like to Jones, who likes people more sure that their moral position is warranted. So it

vanilla ice cream, not what it tastes like to Smith,

is can't who isn't just the fact that 'some people prefer chocolate and some
stand vanilla ice cream. Rather it's like this: Call the quality people prefer vanilla' that makes the Smith/Jones disagreement
vanilla has for Smith V s V$ is an 'unpleasant' taste; it may be
. in preference subjective.

imaginable that one could experience V s and like it, but just Part of the story may be that most people don't have strong
barely, and, even if one did, there would be some kind of disas- preferences between vanilla and chocolate, but this cannot be
sociation or repression. In short, psychologically if not meta- decisive. If half of the population couldn't stand chocolate but
physically, V s is 'intrinsically' unpleasant. And Smith feels D liked vanilla, and the other half couldn't stand vanilla but liked
(dislike for the taste) when he eats vanilla because vanilla has the chocolate, we would still (if we were reasonable about our pref-
taste quality Vs (for him). Similarly, Vj5 the taste quality of erences) regard this as a 'matter of taste', i.e. as subjective. It

vanilla for Jones is intrinsically 'pleasant' (which is why Jones isn't the existence of 'neutrals' that is decisive.
feels L, liking). In Moore, the taste Vj and
the language of G. E. What is decisive, in my opinion, is that whatever biological or
the positive value are an organic unity for Jones,and the taste Vs psychological idiosyncrasies are responsible for Smith's and
and the negative value are an organic unity for Smith. Phenom- Jones' preferences are not correlated with important traits of
enologically, they cannot really ever be separated into two parts mind and character. If we try the thought experiment of imag-
(
in the way the notation V S + D\ 'Vj + U
suggests. Almost cer- ining the contrary, of imagining that there was a caste of char-
tainly (barring special factors of repression or disassociation), acter that we regarded as good, both for sake and
its own
Smith would like vanilla ice cream too if it evoked Vj in his because of its effects on feeling, judgment and action, and
mouth and not VSi and Jones would dislike vanilla ice cream too another caste of character that we regarded as bad, both in itself

if itevoked V8 in his mouth and not V^ and for its effects, and that everyone knew that the good caste
Why do we regard the preference for vanilla over chocolate as of mind and character always revealed itself in a preference for
'subjective' then? I mean, why do we regard it as subjective even vanilla and the bad in a preference for chocolate, then I think we
when we don't think all value judgments are subjective or agree will find that the more vividly we can succeed in making this

with Bentham that, prejudice aside, pushpin is as good as case real to ourselves, the more we will feel that in such a world
poetry? Obviously, if we think all preferences are subjective, we the first preference would be seen as 'normal' and 'right' and the
will think this one is too, but the interesting question is why this second as 'perverse' or 'monstrous' or something of that sort.
judgment doesn't even seem objective to us unless, perhaps, we I don't mean to claim that all preferences are judged morally
are Smith or Jones, why it doesn't have the kind of objectivity by the traits of character they are thought to express. Some 'pref-

that many value judgments do undeniably have. erences' are terribly important in themselves: someone who
It isn't just that there is disagreement. If we think there are thought it was just wonderful to torture small children for the
 Reason and history 155
154 Reason and history

point was recognized by Plato and the medievals - we are per-

fun of it he was serious) be condemned on the basis of
would (if
haps the first culture to conceive of experience as neutral).
that one attitude. But if the matter preferred is not regarded as
The non-neutrality of experience also bears on the
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
pushpin/poetry case. We find it virtually impossible to imagine
that someone who really appreciates poetry, someone who is
on what, if anything, we think the preference shows. Value judg-
capable of distinguishing real poetry from mere verse, capable of
ments often come in clumps; and clumps of value judgments fre-
responding to great poetry, should prefer a childish game to arts
quently express durable traits of mind, personality, and charac-
which enrich our lives as poetry and music do. We have a reason
ter. The independence of 'I prefer vanilla to chocolate ice
for preferring poetry to pushpin, and that reason lies in the felt
cream' from any interesting and significant 'clump' of this kind
experience of great poetry, and of the after effects of great
is just what makes it 'subjective' (along, of course, with the
poetry - the enlargement of the imagination and the sensibility
absence of any intrinsic importance to the choice itself).
through the enlargement of our repertoire of images and meta-
Even if Smith's preference for vanilla is 'subjective', that does
- phors, and the integration of poetic images and metaphors with
not make it irrational or arbitrary. Smith has a reason the best
- mundane perceptions and attitudes that takes place when a
possible reason for liking vanilla, namely the way it tastes to
him. Values can be 'subjective' in the sense of being relative and
poem has lived in us for a number of years. These experiences

still be objective; it is objective that vanilla tastes better than

too are prima facie good— and not just good, but enobling, to
use an old fashioned word.
chocolate to Smith. In The Sovereignty of Good, Iris Murdoch
That there can be reasons for value judgments — reasons
pointed out that philosophers as different as the French Existen-
which really are good reasons for particular people to make par-
tialists and the logical positivists actually shared a common
ticular value judgments — does not mean that all value judg-
model of value judgment, the model of reason as supplying the
ments are rational, of course. Value judgments, judgments that
mind with neutral 'facts' on the basis of which the will must
people have cared passionately about, and in whose name people
arbitrarily choose 'values' - the choice of values must be arbi-
have killed and tortured other people, have often enough been
trary, precisely because 'facts' are (by definition) neutral. But,
how based on an unwholesome mixture of aggressive impulses and
since the will is given no clues by reason as to to judge
(reason only supplies 'facts', on this picture) it has no reason for
narcissistic ideas. Not surprisingly, when a relativist

historian/philosopher like Michel Foucault writes about the past,

its which is why the French philosophers called
arbitrary choice;
he often focusses our attention on these irrational ideas and
it and why more naturalistically inclined philosophers
'absurd',
value judgments. But it is important to see why he does this.
see instinct and emotion (the historic successors to Bentham's
Foucault writes about the early modern era (the sixteenth and
all-purpose category of pleasure) as the ultimate basis of moral
seventeenth centuries) and about ideology and culture in general.
choice.
His knowledge of fact is legendary, even though many specialists
In the case we have just examined, the Existentialist-Positivist
dispute Foucault's 'facts'. While some of his books are highly
model does not fit however. The 'fact' - the taste itself- and the
- - abstract (e.g. The Archeology of Knowledge), some are rather
'value' the goodness of the taste are one, at least psycholog-
specific, e.g. The Birth of the Clinic and The Birth of the Prison.
ically. Presented experiential qualities aren't, in general, neutral
The Birth of the Clinic is perhaps Foucault's best book, and it
and they frequently seem to demand responses and attitudes.
makes an important part of the case for Foucault's more abstract
One may override these felt demands for good and sufficient rea-
theories.
son, as when a child learns to bear the pain of an injection for
the sake of the benefits conferred by the immunizing agent
What Foucault tries to show is that the 'clinic', i.e. the insti-
tution of the hospital and the related medical institutions, was
injected, but the prima facie goodness and badness of particular
the reflection of the growth of a certain ideology about disease
experiences can hardly be denied. (Interestingly enough, this
156 Reason and history Reason and history 157

and health as much as the result of any increase in scientific believe in God, if we don't believe that the Church has special
knowledge and technique. This ideology, in turn, was connected access to his wishes, we will think the Divine Right of Kings was
with wider ideological changes, especially with the growth of and is an irrational doctrine. And finally, even believing Catho-
individualism in the seventeenth century. And he suggests both concede that the Church's support for monarchy in the
lics will
that the 'clinic' is not a very good way of treating most patients Middle Ages was based as much on political considerations as
and that our belief that it is is a kind of ideological prejudice, in on revelation or sound theology. In short, the belief in the Divine
short, a kind of folly. Right of Kings lacks, and always lacked, an adequate rational
The wider suggestion that emerges is that ideological convic- basis.
tionsand the associated value judgments are a rather arbitrary How, then, did the belief arise? The usual answer would
affair. 1 There is no objective place to stand in ideological matters appeal partly to political and economic factors (one does not
(except of course, for the mysterious standpoint of Foucault's have to be a Marxist to concede that these factors are among the
own allegedly objective 'Archeology of Knowledge'). determinants of ideology) and partly to psychological factors.
To see what Foucault is driving at, let us consider a more The comfort provided by belief in a personal God and a here-
familiar and less controversial example. In the Middle Ages, it after is obvious, and so, perhaps, is the comfort provided to the
was monarchy was the natural and proper form of
believed that believer in an infallible Church and a divinely appointed social
government. This belief was based partly on factual beliefs now order. In short, narcissistic ego-gratification and social condi-
thought to be unwarranted (e.g. that democracy would inevita- tioning were the real determinants of this belief. And, if this
bly lead to anarchy and tyranny), and partly on the authority of belief is really typical, if it is really representative of all our 'ide-
the Church. The view of the Church was in fact based partly on ological' beliefs, then such factors are the real determinants of
political considerations (the Church was the state religion), but all ideology.
this was not perceived because the Church itself was thought to It is because they believe something like this that so many
be the divinely inspired and divinely appointed interpreter of the modern French Marx, Freud, and Nietzsche in
thinkers hold
word and will of God. What Foucault is suggesting is that beliefs such high esteem. Marx, Freud, and Nietzsche have this in com-
held in the recent past, and, by implication, the beliefs we hold mon: they see our cherished religious and ethical ideas as reflec-
right now, are no more rational than the medieval belief in the tions of the irrational, of class interest (in the case of Marx), of
Divine Right of Kings. the unconscious (Freud and Nietzsche), of the Will to Power
Let us consider the belief in the Divine Right of Kings for a (Nietzsche). 2
Below what we are pleased to regard as our most
moment. If we don't think there is a good reason to believe in profound spiritual and moral insights lies a seething cauldron of
the existence of a personalGod, one who commands that we live power drives, economic interests, and selfish fantasies. This is
by certain kinds of social forms and structures, then this belief the view that is the cutting edge of relativism today.
will be immediately stamped as irrational (which is not to deny At the same time, no
relativist can himself use the term 'irra-
that the belief answered to real psychological needs). Even if we tional' in the way
have just used it in describing the relativist
I

view. Such a use is ruled out by relativism itself. When I showed

1
And also that we are determined in our thinking by the very language
these pages to a relativist friend he was indignant at my state-
we use. Foucault speaks of 'implicit systems which determine our
most familiar behavior without our knowing it* {see J. Simon, *A ments about the Divine Right of Kings. Was I not aware that
Conversation with Michel Foucault*, Partisan Review, No. 2, 1971, intelligent men had been convinced by philosophical arguments
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 2 Iam not accusing Marx, Freud, or Nietzsche of drawing relativist
science*. conclusions from this.
158 Reason and history Reason and history 159

that this doctrine was correct? Was I offering a cheap Marxist or rational factors, but those ideologies that are the product of the
Freudian explanation? Of course, belief in the Divine Right of interests of the working class are (in the present era) 'just', and
Kings was rational! tend in the direction of human liberation, while those ideolo-
My reply to him was that there may be a sense of 'rational' in gies that springfrom the interests of the exploiting class are
which any view that can be intelligently and persuasively 'unjust'and produce misery. But Althusser distinguishes himself
defended from the shared assumptions of a culture is a 'rational' from previous expounders of this class-relativist view by refusing
view, but that sense cannot be the only or the normatively to say that even Marxist ideology ('working class' ideology) is

important one. The Jews accepted Moses as lawgiver and true or closer to the truth than bourgeois ideology. Ideologies
prophet because his doctrine filled real religious, cultural, and can be 'just' or 'unjust' according to Althusser, but not true or
national needs; that is not the same thing as being convinced by false. 3 ('True' and 'false' apply, he says, in 'laboratory science',
rational argument. Later, prophets anointed Jewish kings (after and, presumably, to those ordinary empirical statements that
trying to dissuade the Jews from having kings at all); that hardly have clear empirical test conditions.) Foucault also seems to be
proves that later Christian kings are divinely appointed. Chris- moving towards a class-interest view in his most recent work,
tianity, which shared the Jewish bible, became the religion of the although it is hard to be sure. The point of such a view, at least
Roman Empire - hardly because the population or the Emperor in its radical Althusserian form, is that it seeks to preserve the
had rational proof that Christianity was true. Roman emperors radical relativist claim that no 'ideology' can be rational while
were then anointed (as Jewish Kings had been); that hardly saving the idea that some ideologies (the preferred one - Marx-
proves that they were divinely appointed. Finally, after the ism-Leninism in the case of Althusser) can be good by distin-
assumptions of Christianity had been accepted, one could give guishing between good and bad or 'just' and 'unjust' ideologies
'rational arguments' for the Divine Rights of Kings from those on grounds other than rational acceptability. Hie idea is that
assumptions. But to express this by saying that in the late Roman although all ideologies are adopted for irrational or non-rational
Empire or the Middle Ages, 'belief in the Divine Right of Kings causes,some non-rational causes (working class interests) are
was perfectly rational' is to debase the notion of rationality. good, and produce good ideologies (by definition?) and some
Hegel, who introduced the idea that Reason itself changes in non-rational causes are bad and produce bad ideologies. Instead
history, operated with two notions of rationality: there is a sense of judging ideologies by their reasons (which are always ratio-
in which what is rational is measured by the level to which Spirit nalizations) we should judge them by their causes.
has developed in the historical process at a given time; it is in This way of limiting one's own relativism is clearly unwork-
something like this sense that it is claimed by some that 'belief in able however. For on what is the judgment based that the victory
the Divine Right of Kings was rational at the time'. And there is of 'working class interests' will lead to such manifestly desirable
a limit notion of rationality in Hegel's system; the notion of that consequences as a world free from war and racism, and not to
which is destined to be stable, the final
self- awareness of Spirit totalitarianism and imperialism disguised as 'socialism'? If the
which not itself be transcended. When present day relativists
will
'naturalize' Hegel by throwing away the limit-concept of true 3
According to Althusser, 'Philosophical propositions are Theses.*
rationality, they turn the doctrine into a self-defeating cultural 'Philosophical Theses can be held negatively as dogmatic propositions,
relativism. insofar as they are not susceptible of demonstration in strict scientific
sense of the term (in which one talks of demonstration in mathematics
No relativist wants to be a relativist about everything how- or in logic), nor of proof in the strict scientific sense {in which one talks
ever. How do these French thinkers put limits on their own rel- of proof in the experimental sciences) . Philosophical Theses, since
. .

ativism? The answer varies with the thinker. In the case of a they can neither be demonstrated nor scientifically proved, cannot be
said to be "true" (demonstrated or proved, as in mathematics and in
Marxist like Althusser, the answer adopted is a version of the physics). They can only be said to be "just" \ Philosophie et Philosophie
'class interest' theory: all 'ideologies' are the product of non- Spontanee des Savants, pp. 13-14, Maspero (1967).
160 Reason and history Reason and history 161

answer is that the latter wouldn't be 'true' socialism, or reflect our present our present ideology is the product of forces
lights. If

'true* working class interests, then on what is the judgment that are irrational byits own lights, then it is internally incoher-

based that any particular institution (e.g. the French Communist ent. The French thinkers are not just cultural relativists; they are

Party, of which Althusser is a leading member) or policy will attacking our present notion of rationality from within, and this
promote 'true' working class interests and 'true' socialism? If is what the reader feels and
is troubled by.

these beliefs can be rationally justified, then not every ideology Cultural relativismnot a new doctrine at all. Anthropolo-
is

is irrational; if they cannot be, then the claim that any institution gists have been preaching cultural relativism to us as long as

or policy is 'just' must be as utterly irrational as every other 'ide- there have been anthropologists. But it would be a mistake to
ological' claim is asserted to be. If all human thought about ide- assimilate the relativism of Foucault to the older relativism.
ological questions is self-serving folly, then thought about which When an anthropologist preaches relativism to us, normally
beliefs spring from 'working class interests' and which do not he and beliefs of an exotic tribe which initially
cites practices

must also be self-serving folly. strike us as irrational or repulsive or both, and proceeds to show
Coming back to Foucault, however, and ignoring the signs in that these actually promote welfare and social cohesion. In
Foucault's most recent work that he too is becoming radicalized, short, he shows (to the extent that the example is a reasonable
his motive for focussing on the cases he chooses is precisely to one) that what is considered wrong or irrational in our society
suggest the utterly non-rational (and, in fact, irrational) character may actually be reasonable and right in different natural and
of the real reasons that people have for adopting ideological social circumstances.

positions. And the notion of the ideological here is very wide; it Of course, the wrong conclusion is frequently drawn by
is not just Communism, Fascism, Democracy, the Divine Right anthropologists from their own examples (and some of the
of Kings, etc., that are under discussion. The someone
belief that examples are rather less clear than the anthropologist thinks).
is 'diseased' and needs a 'cure', the belief that someone is a 'crim- Very often an anthropologist will say 'it's all relative', meaning
and should, if possible,
inal' be 'rehabilitated', and many, many that there is no fact of the matter as to what is right and wrong
more of our everyday beliefs are 'ideological' in the sense of at all. And Richard Boyd has suggested to me that very often the
these thinkers. Indeed, to the eagle eye of the Marxist sociologist motive is a political one; to convince us to stop destroying pri-
or the French philosopher, almost every belief is 'ideological'. mitive cultures by attacking our belief in the superior rationality
Perhaps, 'if I drop this glass, it will break' is ideologically neu- and morality of our own. Unfortunately, the argument is very
tral, but little else is. confused. The anthropologist's examples (when they are good
It missed the real point of what Foucault
may seem that I have ones) show that right and wrong, for example, are relative to
is saying. His real point, he himself would say, is not that the circumstances, not that there is no right and wrong at all, even
ideological perspectives of the past were foolish or irrational at in specified circumstances. His argument against cultural impe-
all, but rather that all ideology in the very wide sense in which rialism amounts to this: other cultures are not objectively worse
he uses the term, including our present ideology, is culture- than ours (because there is no such thing as objective better and
relative. He is trying to show us how every culture lives, thinks, worse, according to him); therefore they arc just as good as ours;
sees, makes love, by a set of unconscious guiding assumptions therefore it is wrong to destroy them.

with non-rational determinants. If previous ideologies now seem This argument equivocates. The conclusion requires that 'just
'irrational' it is because we judge them by our culture-bound as good as'means objectively just as good as (at least by our
notion of rationality. lights); but what follows from the non-existence of objective val-
What is troubling about Foucault's account is that the deter- ues cannot be that everything is (in the required sense) 'just as
minants he and other French thinkers point to are irrational by good' as anything else, but rather that there is no such thing as
162 Reason and history Reason and history 163

know' could not be one he knew), the modern relativist, were he

'just as good as'. If values really were arbitrary, then why should
consistent (and how
could one consistently hold a doctrine
we not destroy whatever cultures we please?
which makes nonsense of the notion of consistency?) should end
Fortunately, there are better grounds for criticizing cultural
by regarding his own utterances as mere expression of feeling.
imperialism than the denial of objective values. The anthropol-
To say this is not to deny that we can rationally and correctly
ogist's motive may be a good one, but he has chosen the wrong
think that some of our beliefs are irrational. It is to say that there
argument. Another term on which he equivocates is the notion
are limits to how far this insistence that we are all intellectually
of being 'relative'. examples actually confirm is
What his
Certain things are right -objec-
damned can go without becoming unintelligible. We do, for
Dewey's 'objective relativism'.
example, discuss just such doctrines as those advanced by Fou-
tively right - in certain circumstances and wrong - objectively
cault; we make an effort to be impartial; we try to adopt what
wrong - in others, and the culture and the environment consti-
Popper calls 'the critical attitude', and actively to seek evidence
tute relevant circumstances. About this the anthropologist is
and argumentation we might overlook, even when it bears
right. But this is not the same thing as values being 'relative' in
against our own views. None of this would make the slightest
the sense of being mere matters of opinion or taste.
sense if we did not think that these practices of discussion and
Still, freed of its conceptual confusions, the anthropologist's
communication, and these virtues of criticism and impartiality
argument should not trouble us. We should welcome his obser-
tend to weed out irrational beliefs, if not at once, then gradually,
vations, for they tend to widen our sensibilities and attack our
over time, and to improve the warranted assertibility of our final
smug assumption of cultural superiority. But the very compari-
conclusions. Rationality may not be defined by a 'canon' or set
son of Foucault's argument with the anthropologist's brings out
of principles, but we do have an evolving conception of the cog-
their difference: Foucault is not arguing that past practices were
nitive virtues to guide us.
more rational than they look to be, but that all practices are less
It will be objected that this conception does not 'get us very
rational, are, in fact, mainly determined by unreason and selfish
far'. Rudolf Carnap and John Cardinal Newman were both
power. The similarity of this doctrine to the older cultural rela-
careful and responsible thinkers, and both were committed to
tivism is a superficial one.
the cognitive virtues just mentioned, but no one thinks that one
The fact is that the position we have been discussing caters to
could have convinced the other, had they lived at the same time
an intellectual temptation which is the product of our increased
and been able to meet. But the fact that there is no way to resolve
knowledge about and sensitivity to psychological and sociologi-
all disputes to everyone's satisfaction does not show that there is
cal mechanisms. The knowledge and the sensitivity are in part
no better and worse in such a case. Most of us think that New-
pretense and in part real; the temptation is to fall into the trap
man's Catholicism was somewhat obsessive; and most philoso-
of concluding that all rational argument is mere rationalization
phers think that, brilliant as he was, Carnap employed many
and then proceeding to try to argue rationally for this position.
weak arguments. That we make these judgments shows that we
If all 'rational argument' were mere rationalization, then not
do have a regulative idea of a just, attentive, balanced intellect,
only would it make no sense to try to argue rationally for any
and we do think that there is a fact of the matter about why and
view, but it would make no sense to hold any view. If I view my
how particular thinkers fall short of that ideal. Some will say,
own and dissent as crazy behavior, then I should stop
assent
'So what; we are no better off when it comes to resolving an
assenting and dissenting - something to which there can be no
actual dispute than if there were no notion of rational accepta-
rational assent or dissent, only crazy parody of rational discus-
bility external to the views under debate to which we could
sion,cannot be called a statement Like Sextus Empiricus, who
appeal!' This is truewhen it comes to any one unresolvable dis-
eventually concluded that his own scepticism could not be
pute such as the Carnap-Newman dispute just imagined; but it
expressed by a statement (because even the statement, 'I do not
164 Reason and history Reason and history 165

is not true that we would be just as well off in the long run if we one's co-disputant in such a discussion is interestingly mixed. On
abandoned the idea that there are really such things as impar- the one hand, one recognizes and appreciates certain intellectual
tiality, consistency, and reasonableness, even if we only approx- virtues of the highest importance: open-mindedness, willingness
imate them in our lives and practice, and came to the view that to consider reasons and arguments, the capacity to accept good
there are only subjective beliefs about these things, and no fact criticisms, etc. But what of the fundamentals on which one can-
of the matter as to which of these 'subjective beliefs' is right. not agree? It would be quite dishonest to pretend that one thinks
Perhaps the analogy I have (occasionally) drawn between phil- there are and worse reasons and views here. I don't
no better
osophical discussion and political discussion may be of help. think it is whether one thinks that the obli-
just a matter of taste
One of my colleagues is a well-known advocate of the view that gation of the community to treat its members with compassion

all government spending on 'welfare' is morally impermissible. takes precedence over property rights; nor does my co-disputant.
On his view, even the public school system is morally wrong. If Each of us regards the other as lacking, at this level, a certain
the public school system were abolished, along with the compul- kind of sensitivity and perception. To be perfectly honest, there
sory education law (which, I believe, he also regards as an imper- is each of us something akin to contempt, not for the other's
in
missible government interference with individual liberty), then mind — for we each have the highest regard for each other's
the poorer families could not afford to send their children to minds - nor for the other as a person -, for I have more respect
school and would opt for letting the children grow up illiterate; for my colleague's honesty, integrity, kindness, etc., than I do for
but on his view, is a problem to be solved by private charity.
this, that of many people who agree with my 'liberal' political views -
If people would not be charitable enough to prevent mass illit- but for a certain complex of emotions and judgments in the
eracy (or mass starvation of old people, etc.) that is very bad, other.
but it does not legitimize government action. But am I not being less than honest here? I say I respect Bob
In my view, his fundamental premisses — the absoluteness of Nozick's mind, and I certainly do. I say I respect his character,
the right to property, for example — are counterintuitive and not and I if I feel contempt (or something in that
certainly do. But,
supported by sufficient argument. On his view I am in the grip ballpark) for a certain complex of emotions and judgments in
of a 'paternalistic' philosophy which he regards as insensitive to him, is that not contempt (or something like it) for him?
individual rights. This is an extreme disagreement, and it is a This is a painful thing to explore, and politeness normally
disagreement in 'political philosophy' rather than merely a keeps us from examining with any justice what exactly our atti-
'political disagreement'. But much political disagreement tudes are towards those whom we love and disagree with. The
involves disagreements in political philosophy, although they are fact is that none of us who is at all grown up likes and respects
rarely as stark as this. everything about anyone (least of all one's own self). There is no
What happens in such disagreements? When they are intelli- contradiction between having a fundamental liking and respect
gently conducted on both sides, sometimes all that can happen for someone and still regarding something in him as an intellec-
is that one and delineates the source of the
sensitively diagnoses tual and moral weakness, just as there is no contradiction
disagreement. Often, when is less fundamental
the disagreement between having a fundamental liking and respect for oneself and
than the one I described, both sides may modify their view to a regarding something in oneself as an intellectual and moral (or
larger or smaller extent. If actual agreement does not result, per- emotional, etc.) weakness.
haps possible compromises may be classed as more or less I want to urge that there is all the difference in the world
acceptable to one or another of the parties. between an opponent who has the fundamental intellectual vir-
Such intelligent political discussion between people of differ- tues of open-mindedness, respect for reason, and self-criticism,
ent outlooks is, unfortunately, rare nowadays; but it is all the and one who does not; between an opponent who has an impres-
more enjoyable when it does happen. And one's attitude toward sive and pertinent store of factual knowledge, and one who does
 Reason and history 167
166 Reason and history
Socialism, Rights, Autonomy, and so on, is not frequently non-
not; between an opponent who merely gives vent to his feelings
Even when what we mean to say about some general issue
sense.
and fantasies (which is all people commonly do in what passes
is right, frequently we have trouble expressing it well, especially
for political discussion), and one who reasons carefully. And the
if we are not trained in the expression of abstract ideas. The case
ambivalent attitude of respectful contempt an honest one:
is
of the anthropologist who says there are no objective values
respect for the intellectual virtues in the other; contempt for the
when what he means is that values are relative to circumstances
intellectual or emotional weaknesses (according to one's own
is a case in point. Even when we do succeed in expressing what
lights, of course, for one always starts with them). 'Respectful
we mean to say effectively, there are powerful forces of a non-
contempt' may sound almost nasty (especially if one confuses it
rational kind tending to sway our judgment. My purpose here is
with contemptuous respect, which is something quite different).
not to deny that power can corrupt our judgment and narcissism
And it would be nasty if the 'contempt' were for the other as a seduce it; it is to deny that we are helpless in the face of these
person, and not just for one complex of feelings and judgments
powerful forces, so helpless that it would be idle (and in fact,
in him. But it is a far more honest attitude than false relativism;
self-deception) to attempt to judge with intelligence and justice.
that is, the pretense that there is no giving reasons, or such a
To say we can be rational is different from saying we can be
thing as better or worse reasons on a subject, when one really
infallible. On the contrary, as Iris Murdoch points out, the striv-
does feel that one view is reasonable and the other is irrational.
ing for a reasoned and rational stance is essentially something
progressive, something 'infinitely perfectible'. 4
It may be helpful to descend from the abstract level at which

we have been discussing and consider once more a relatively sim- What have said so far might, perhaps, be conceded by an
I
ple example. Consider the judgment that most ordinary people
moral relativist. A relativist need not be concerned to
intelligent
are prepared to make at most times, that peace is preferable to
undermine the rationality of all 'value' judgments, or to defend
war. (Such judgments are never discussed by Foucault, just as
Foucault's picture of history as a discontinuous series of 'dis-
they are never described by Swift, and for the same reason: both
courses' or 'ideologies' which succeed one another for no
are satirists. Only social folly interests them, not -when it
rational reason. A more modest relativist might be happy to
exists - social sanity.) There are no doubts about the sources of
agree with Dewey 5 that some values are objectively relative — i.e.
such a judgment. We are too familiar with the horrors of war,
rational given the circumstances, the nature and history, of those
with what war does to adults and children, to combatant and
who make them. What is important, such a modest relativist
non-combatant, to the very land and foliage. Even if this judg-
maintains, is precisely the relativity of all values. The 'objectiv-
ment springs partly from self-interest, that does not make it irra-
ity' he is denying is not the objectivity Dewey was affirming,
tional, quite the contrary.
which is simply the objectivity of any judgment that is warranted
Yet whole populations can make the opposite judgment, that
in its actual existential setting, but it is rather the objectivity a
war is preferable to peace, and not for reasons of legitimate self-
Plato would affirm, the spurious objectivity (the relativist would
defense. Aggression and fantasy can whip people up to a
say) that purports to speak from an absolute point of view, apart
national blood thirst. But, again, what this shows is not that all
from all circumstances and valid for all circumstances.
value judgments are irrational, but only that some are; and that
If we are not content to accept such a modest relativism, if we
it is very hard to tell which are which when one is not able to
feel troubled by Dewey's own ethical writings, it is not, I think,
put aside partisanship or criticize one's own beliefs (which is
because we really do hanker for Absolutes. When 1 claim that
why we assign so much importance to impartiality and the crit-
4
ical attitude among the cognitive virtues). The Sovereignty of Good, p. 23.
5
See Dewey's Theory of Valuation, in The Encyclopedia of Unified
That some value judgments are rational and objective does not
Science, vol. II, no. 4, University of Chicago (1939).
mean that our abstract talk about Capitalism, Democracy,
 Reason and history Reason and history 169
168

because the assumption made that the choice of means

the murder and suffering of innocent people is wrong, I do not,
cisely is

this judgment subject to rational criticism, while the choice of ends is not.
I think, really care about the question whether
is

would be valid for a being of a totally alien constitution and This whole conception loses much of its persuasive appeal,

there are beings on, say, Alpha Centauri, who however, when we see what an oversimplified psychological the-
psychology. If

cannot feel pain and who do not mind individual death, then ory it rests upon. In the Benthamite scheme, goals, ends, prefer-
ences are treated either as fixed individual parameters (i.e. the
very likely our fuss about 'murder and suffering' will seem to
individual's learning pictured as a process of learning to better
them to be much ado about nothing. But the very alienness of is

estimate consequences and probable consequences of actions

such a life form means that they cannot understand the moral
issues involved. If our 'objectivity' is objectivity humanly speak- and to attain ends more but not as a process of
efficiently,

ing, it is still objectivity enough. acquiring new ends) or as individual parameters which, if not
actually fixed, change only as a result of factors which have no
What of concern is that Dewey's doctrine of 'objective rel-
is
and which the theorist cannot take account of.
rational status
6
cannot handle the case of the Nazi (although Dewey Bernard Williams has pointed out that there are a number of
ativism'
this). We want to say that the Nazi's goals
ways in which an individual's goals, and not just the means he
would have disputed
wrong; and the claim that 'this is true relative to chooses to attain them, can be rationally criticized; ways which
were deeply
your interests and false relative to the Nazi's interests' is just the become apparent the moment we pass beyond this narrow 'Ben-

we find repulsive. Objective relativism thamite' psychology.

kind of 'moral relativism'
seems the right doctrine for many moral cases; but not for cases The 'Benthamite' conception does allow one case in which an
and duties are manifest and sharp and the choice individual can be persuaded to abandon a goal (or, at any rate,
where rights
seems to us to be between right and wrong, good and evil. to abandon pursuit of the goal) by rational criticism: this is the

Indeed, there is a sense in which the modern instrumental case in which he had misestimated consequences in the direction

notion of rationality is itself 'objective relativist'. The core of this of badly underestimating the costs of attaining the goal (relative

notion is a deceptively simple dichotomy: the idea is that the to other goals he has). This opens the door to a question which
choice of 'ends' or 'goals' is neither rational nor irrational (pro- has to do as much with imagination with propositional intel-
as
ligence: the question of what it would actually be like, experien-
vided some minimal consistency requirements are met); while
tially, to attain the goal. Many human beings pursue goals they
the choice of means is rational to the extent that it is efficient.
Rationality a predicate of means, not ends, and it is totally
is
would not actually enjoy attaining or would not enjoy nearly as
conflated with efficiency. Thus, Jones' preference for vanilla over >long or nearly as much as they think. Even within a Benthamite
chocolate ice cream is neither rational or irrational, but the framework, it would be possible to improve the account of
action of choosing vanilla over chocolate on a particular occa- rational decision making by taking into account the possibility

sion would be rational for Jones, given his 'preference ordering'. of misestimating the actual existential feel of various goals. And
this begins to introduce a sense in which goals themselves can be
This conception, which goes back to Hume's dictum that 'reason
criticized as irrational, and not just means.
is and ought to be the slave of the passions', and which
deeply
influenced Bentham, is widely assumed to be the right one by Again, people often overlook goals they might pursue if they

modern social scientists. It has played a role in welfare econom- thought of them. Or, even if they think of them (or someone
ics and in many other areas. The modern economist's notion of
suggests them) they may lack the imagination (imagination

a Pareto optimum is an attempt to have a notion of economic again!) to visualize what the attainment of these goals would
optimality which considers only efficiency of means, and
involves no 'value judgments' concerning the goals of the various 6
Iam here summarizing a lecture titled 'Internal and External Reasons*,
of contemporary interest pre- given at Harvard, Nov. 1978.
economic agents; this notion is
170 Reason and history Reason and history 171

really be like - all the more if these goals are long-term traits of haps some Nazis would not have been Nazis if they had had the
character, such as developing an appreciation of poetry. The intelligenceand imagination to appreciate these consequences,
man who prefers pushpin to poetry may not actually be able to or to appreciate more vividly the alternative life, the life of a
imagine what it would be like to have a developed sensitivity to good man. But doubtless many Nazis would still have been
the nuances of real poetry, and if his intelligence could be raised Nazis, because they did not care about the suffering their actions
or his imagination improved he might be brought to see that he caused and because no matter how vivid they might make the
is making a mistake. alternativelife seem to their imaginations, it would no more

It is significant that the ability to rationally criticize one's own speak to anything in them than the military life did to the young
goals (and those of others) may depend just as much on one's man in Bernard Williams' story. There is no end in them to
imagination as on one's ability to accept true statements and which we can appeal, neither an actual end or even a potential
disbelieve false ones. And it is significant that one's goal may be one, one which they would come to realize if they were more
a long-term trait of mind or character, and not a thing or event. intelligent and more imaginative. Even without 'Benthamite psy-
There are still further ways besides misestimating the real chology', we are faced again with the problem of moral relativ-
experiential significance of one's goals or of possible alternative ism.
goals in which one may make errors in the choice of goals. Wil- Let us consider a case less inflammatory than the Nazi case.
liams points out (reviving an observation that goes back to Aris- Imagine a society of farmers who, for some reason, have a total
totle) that very often a goal is general (e.g. 'having a good time disinterest in the arts, in science (except in such products as assist
this evening') and the problem is not so much to find a means to them in farming), in religion, in short, in everything spiritual or
the 'end', but to find an overall pattern of activity that will con- cultural. (I don't mean to suggest that actual peasant societies
stitute an acceptable specification of the goal (e.g. 'going to a are or ever have been like this.) These people need not be imag-
movie' or 'staying home and reading a book'). Whether one can ined as being bad people; imagine them as cooperative, pacific,
think of creative and novel specifications of one's goal or only of reasonably kind to one another, if you like. What I wish the
commonplace and banal specifications will depend again on reader to imagine is that their interests are limited to such mini-
imagination and not just propositional intelligence. mal goals as getting enough to eat, warm shelter, and such sim-
The problem, as Williams pointed out, is that even if one ple pleasures as getting drunk together in the evenings. In short,
replaces the narrow Benthamite psychology with an account that imagine them as living a relatively 'animal' existence, and as not
does justice to all of these things, one still seems to be left with wishing to live any other kind of existence.
a certain relativism. Williams' example was a hypothetical case Such people are not immoral There is nothing impermissible
of a young man whose father wished him to undertake a military about their way of life. But our natural tendency (unless we are
career. The old man appeals to family traditions (the males have entranced with Ethical Relativism) is to say that their way of life
been army officers for generations) and patriotism, but in vain. is some way contemptible. It is totally lacking in what Aris-
in
Even when the young man makes as vivid to himself as he can They are living the lives of swine - amiable
totle called 'nobility'.
what it would be like to be an army officer, there is nothing in swine, perhaps, but still swine, and a pig's life is no life for a
this goal which appeals to him. It just is not his end; and not man.
because of some failure of intelligence or imagination. At the same time - and this is the rub - we are disinclined to
Even the case of the Nazi could be like this. Suppose the Nazis say the pig-men are in any way irrational This may be the result
had won the war, so that we could not appeal to Germany's of our long acculturation in the Benthamite use of 'rational' and
defeat as a practical reason for not being a Nazi. Perhaps some 'irrational', but, be that as it may, it is our present disposition.
Nazis were simply lacking in knowledge of the actual conse- The lives of the pig-men are not as good as they might be, we
quences of Nazism, the suffering brought about, and so on. Per- want to say, but they are not irrational
172 Reason and history Reason and history 173

We do not want to say that it is just a matter of taste whether [Link] neither ancient philosophers nor the medievals
one lives a better or worse life. We don't see how we can say that saw anything strange about saying that if A is a better life than
it is rational to choose the better life and irrational to choose the B then that fact is a reason, the best possible reason, for choosing
worse. Yet not saying some such things seems precisely to be A over B. We have lost the ability to see how the goodness of an
saying that 'it's all relative'; the ground crumbles beneath our end can make it rational to choose that end.
feet. Of course, this is very largely explained by the fact that we
Perhaps some of the corrections to Benthamite psychology don't regard 'goodness' as anything objective. But now we are
suggested by Bernard Williams will help with this case. Let us confronted with a circle, or rather two curves. There is the mod-
assume the pig-men are born with normal human potential (if
ern circle: the instrumentalist conception of rationality supports
they aren't, then their lives aren't 'worse than they might be', the claim that the goodness of an end doesn't particularly make
and we are not justified in feeling contempt, but only, at most, it irrational not to choose that end, or to choose
an end which is
pity). Then they might be led to appreciate artistic, scientific, and downright bad, which in turn supports the claim that goodness
spiritual aspects of life; to live more truly human lives, so to and badness are not objective, which in turn supports the claim
speak. And if any of them did this, they would doubtless prefer that the instrumentalist notion of rationality is the only intelli-
those lives (even though they might be less fun) to the lives they gible one. And there is the traditional arch: reason is a faculty
are now living. People who live swinish lives feel shame when which chooses ends on the basis of their goodness (as opposed
they come to live more human lives; people who live more to the 'passions', which try to dictate ends on the basis of the
human lives do not feel ashamed that they did when they sink appetites; or 'inclination'); a claim which supports the view that
into swinishness. These facts give one grounds for thinking that it is rational to choose the good, which in turn supports the

the pig-men are making the sort of error, the sort of cognitive claim that goodness and badness are objective. Clearly we can-
short-fall, that Williams discussed; grounds for thinking that not simply go back to the ancient or medieval world- view, what-
they have overlooked alternative goals, and certainly grounds ever conservatives might wish; but is the Benthamite circle really
for thinking that they have never made vivid to themselves what the only alternative left to us?
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
 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-
2 the results of 'pos-
ality always means scientific rationality; it is

8 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 obvious. Scientific rationality undeniably aids us in
is

achieving various practical goals. While few educated people

The impact of science on would subscribe to the view that we ought to pursue science
solely for the sake of technological success, there is no doubt

modern conceptions of rationality that the technological success of science overwhelming in the
is

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

If the discussion that we have reviewed — and it is a discussion Hermeneutics and Ethics: Types of Rationality Beyond Science and
Technology', Research in Phenomenology No. IX, 1979. Thus Apel
',

that has been going on for many decades — seems inconclusive, writes (p. 36):
perhaps because the discussion always assumes a kind of
it is
negative answer
Max Weber, however, also proposed a strictly
priority of rationality over goodness. The question is always with regard to my
question as to possible types of rationality
whether there is any sense in which it can be called irrational to beyond value-free science and technology. And this answer has
become paradigmatic, I suggest, for the present system of Western
choose a bad end, as if goodness were on trial and rationality
ideology. Weber restricted the scope of methodical understanding
were the judge. To assume this stance, especially when one's to 'value-free' understanding which he centered around the 'ideal

assumptions about rationality are a largely unexamined collec- type* of 'purposive-rational understanding' of 'purposive-

tion of cultural myths and prejudices, is to prejudge the question

rational actions'. Now
'purposive-rational actions' may also be
called 'instrumental actions'; and in those cases where these
of the status of value judgments in advance. In the remainder of actions are successful, they may be analyzed or reconstructed as
this essay I propose to reverse the terms of the comparison and being based on successful transpositions of the if-then-rules of
nomological science into the if-then-rules of technological pre-
to ask not how rational is goodness, but why is it good to be
Hence Max Weber thus restricted the business of
scriptions.
rational? Asking what value rationality itself has will both force methodical understanding to the attempt of grasping the (value-
us to become clearer about the nature of rationality itself and free) technological means-ends-rationality beyond the human
actions. And it is this idea of instrumental rationality which
about the assumptions we are prone to make concerning ration-
indeed constituted Weber's paradigm of rationality in a restrictive
ality and may enable us to see what is wrong with the way we sense.
think about the former question. It has to be pointed out, though, that for a purposive-national

understanding in sociology it is not necessary to fulfill the maxi-

Let us recall that when Max Weber introduced the modern
mal requirement of making sure that the agent succeeded in
fact-value distinction, his argument against the objectivity of transposing nomological rules into his technological maxims
value judgments was precisely that it is not possible to establish about means-ends-relations. In order to understand his actions in
the light of that type of instrumental rationality, it is enough to
the truth of a value judgment to the satisfaction of all possible
make sure that it was rational for the agent to act as he did under
rational persons. * From the very beginning it was the impossi- the presupposition of his aims and his beliefs about means or
ways or strategies as being suited to reaching his aims. Thus it
1
Cf. especially 'Die Objektivitat sozialwissenschaftlicher Erkenntnis', in becomes the empiric-hermeneutic business of understanding to
Archiv fur Soztalwiss. und Sozialpolitik, vol. 19 (1904), pp. 24-87, hypothetically find out and verify those goal-intentions and
and *Der Sinn der Wertfreiheit der soziologischen und okonomischen means-beliefs on the side of the agent, in the light of which his
Wissenschaften', in Logos, VII (1917), pp. 49-88 and 'Wissenschaft actions can be understood as being rational in the sense of tech-
als BeruP, Vortrag, 1919. All three texts were reprinted in The nological means-ends rationality.
Methodology of the Social Sciences, Glencoe, Illinois, 1949.
 176 The impact of science The impact of science 177

series of technological revolutions - 'the Industrial Revolution', times thought to consist, or to consist in part, in the fact that at
'the Electronic Revolution' - which constantly remind us how least the predictions of science can be publicly demonstrated;
momentous a force science is in shaping our lives. Even before that everyone can be brought to agree that these results do
the Industrial Revolution the apparently unique success of New- Of
obtain, that the phenomena predicted by the theory do occur.
tonian physics impressed a number of minds. For example, when we identify rational
course there is a threat of circularity here: if
the notion of 'progress' first began to be discussed in the seven-
procedures as those which lead to conclusions on which we can
teenth century, the progressivists clinched their case with the correct, that
get majority assent, then Weber's argument, even if
claim that 'Newton knew more than Aristotle.' No one could in ethical matters one cannot get the consent of all rational peo-
argue convincingly that Shakespeare is a better dramatist than ple,would mean that one cannot get consent of all those who
any of the ancient tragedians, or a better poet than Homer; but use methods which are guaranteed to produce the assent of the
itseemed undeniable that the scientist Newton had made a real majority or the overwhelming majority. That is, the way in
and undeniable advance in knowledge upon the scientist Aris- which we determine that value judgments cannot be verified to
totle.
the satisfaction of all rational people is simply by observing that
Although the encyclopedists and others were quick to gener- they cannot be verified to the satisfaction of the overwhelming
alize thenotion of progress from science to political institutions majority of all people. It is not, afterwe had a test for
all, as if
and morality, that generalization has appeared as dubious to the rationality; Weber's formulation suggests that somehow we first
twentieth century as it appeared evident to the nineteenth. take a headcount of those members of the population who are
Auguste Comte built a philosophy, positivism, celebrating the rational and then see whether or not they can all be brought to
success of science. History, as Comte tells it, is a success story: agree on whether or not some value judgment is true. But noth-
we start with primitive myths, these become refined and purified ing like this really goes on. All that Weber's examples (Chinese
until eventually the high religions appear, the high religions in Mandarins and so on) really show is that value judgments can-
turn give way to the metaphysical theories of a Plato or a Kant, not be verified to the satisfaction of all educated people or all
and finally in our own day metaphysics itself has to give way to intelligent people (which is by no means the same thing as all
'positive science'. Evidently, there no doubt who the hero of
is rational people). In a disguised form Weber's argument is a
this success story is: the hero is Science. And if what impressed
Majoritarian argument; he is appealing to the fact that we can
the Few about science from the start was its stunning intellectual
get the agreement of educated people on 'positive science'
success, there is no doubt that what has impressed the Many is
whereas we cannot get such an agreement on ethical values. It is
its overwhelming material and technological
success. We are interesting to contrast this stance with Aristotle's: Aristotle said
impressed by this even when it threatens our very lives.
that of course in ethics we should always try to get the agreement
One reason, then, for doubting that value judgments have any
of the Many, but very often we know that realistically we can-
cognitive status is that they cannot be 'verified by the methods of sounds to present day ears to say it,
not. Sometimes, elitist as it
science', as gets put over and over again. Then too there is the
it
we are only able to convince the wise; and of course we have to
fact, that we have already seen play a central role in Foucault's rely on our own judgment to tell who are and who are not the
discussion, that one cannot get universal, or even majority, wise.
agreement on ethical matters. Whether abortion is right or Of course it is not really true that one can get overwhelming
wrong or whether homosexuality is right or wrong do not seem agreement on the truth of an arbitrary accepted scientific theory.
to be questions on which some answer can be demonstrated
to The fact is that most people are woefully ignorant of science and
everyone's satisfaction; whereas it is widely believed that the much
many theories, especially in the exact sciences, require so
correctness of a scientific theory can be demonstrated to every-
mathematics for their comprehension that most people are not
one's satisfaction. The very rationality of science itself is some- even capable of understanding them. While this is of course con-
178 The impact of science The impact of science 179

ceded, to most peopleit does not seem to affect the point. For, Majoritarianism is also intenable. To be sure it is nice to
according to the watered-down operationism which seems to get agreement on what one takes to be always nice
true. And it is

have become the working philosophy of most scientists, the con- to avoid conflict with one's fellows. But people have lived for
tent of the scientific theory consists in testable consequences, and centuries with the uncomfortable knowledge that on some mat-
these can be expressed by statements of the form if we perform ters one has to rely on one's judgment even when it differs from
such and such actions, then we will get such and such observable the judgment of the majority. Many have gloried in relying on
results. Statements of this form, if true, can be demonstrated to their judgment when it differed from the judgment of the major-
be true by repeating the appropriate experiment often enough. It ity. The idea that on some matters, ethical matters among them,

is true that there are many difficulties with this account: experi- the considerations to be weighed are just so complex, and so
ments are much harder to design, perform, and evaluate than the imprecise, that we cannot hope to rely on anything like scientific
layman may think. But there is no doubt that as a matter of fact proof or scientific definitions but have to rely on perception and
it has been possible to achieve widespread agreement on the judgment is an old one. And it is plausible that one of the highest
experimental adequacy of certain theories in the exact sciences. manifestations of rationality should be the ability to judge cor-
The layman's acceptance of these theories may be a matter of his rectly in precisely those cases where one cannot hope to 'prove'
deference to experts, but at least the experts seem to be in agree- things to the satisfaction of the majority. It seems strange indeed
ment. that the fact that some things should be impossible to prove to
Intellectually, of course, Instrumentalism does not simply in everyone's satisfaction should become an argument for the irra-
and of itself constitute a tenable conception of rationality. No tionality of beliefs about those things. 4
doubt scientific results have enormous practical value; but, as we Even if these conceptions are intellectually weak however, it
have already said, no educated person thinks that science is val- seems to be the case that both Instrumentalism and Majoritari-
uable solely for the sake of its practical applications. And even if anism are powerfully appealing to the contemporary mind. The
science were valued solely for the sake of its applications, why contemporary mind likes demonstrable success; and the contem-
should rationality be valuable solely for the sake of applications? porary mind is uncomfortable with the very notions of judg-
To be sure it is of value to have an instrument that helps us select ment and wisdom. I am not a sociologist, and I will not attempt
efficient means for the attainment of our various ends; but it is to explore the question why industrial society, in both its capi-
also valuable to know what ends we should choose. It is not
surprising that the truth of value judgments cannot be 'rationally
ment' ['Entzauberung']. By 'rationalization* he understood the
demonstrated' if by definition limited to
'rational verification' is progress in putting into force means-end-rationality in all sectors
the establishment of means-ends connections. But why should of the socio-cultural system, especially in the sphere of economics
and bureaucratic administration, under the constant influence of
we have such a narrow conception of rationality in the first the progress in science and technology. By the process of 'disillu-
3
place? 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.
3
Attributing just such a narrow conception of rationality to Weber, Apel And he was prepared to draw practical consequences from this
writes {he. cit. t p. 37): development for his personal world-view in so far as he suggested
that a rigorous and sincere thinker had to accept the following
This issue of Weber's methodology of 'understanding' was in insight: Human progress in the sense of 'rationalization' has its
perfect accordance with his (more or less implicit) philosophy of complement in giving up the idea of a rational assessment of last
history. For in the context of his own reconstruction of the his- values or norms in favor of taking recourse to ultimate pre-
tory of Western civilization, he started out from the heuristic rational decisions of conscience in face of a pluralism, or, as
hypothesis that at least this part of history could be conceived of Weber said, 'polytheism' of last norms or values.
as a continuous progress of 'rationalization* and at the same time, 4
Indeed, we saw in Chapter 5 that the consensus theory of rational
as a process of disillusionment or, as he liked to say, 'disenchant- acceptability is self-refuting!
 The impact of science The impact of science 181
180

talist and its socialist versions, should be so taken with the philosophical analysis to show it, just knowing a great many
themes of instrumental success and majority consent. But doubt- facts of theform if you perform such and such actions, then you
less the sociological fact has something to do with the ever will have such and such experiences. Whatever our reason for

increasing prominence of the conception that rationality equals being interested in them, all facts are ultimately instrumental At
scientific rationality, and the conception of scientific rationality the same time, there seems to be no way of making out the claim

as itself based on the demonstration of instrumental connections that to call something good is to make any prediction of the

to the (potential) satisfaction of the overwhelming majority. form: if you perform such and such actions then you will have
If the conception of rationality we have just described, the
such and such experiences. Thus statements about the goodness
conception of rationality as consisting in methods (whose nature or badness of anything have no cognitive meaning on this con-
is usually left rather vague) which, whatever their nature, result ception; in the words of the twentieth-century Logical Empiri-
in the discovery of effective means/ends connections and the cistssuch statements are purely 'emotive'. Phenomenalism came
establishment of these connections 'publicly', is not as it stands to grief on two points, however. In the first place the claim that

intellectually tenable, philosophical attempts to make it respect- statements about material objects are translatable into state-
able have not been lacking. One of these attempts grows out of ments about actual and possible sensations seems as a matter of
the older empiricism of Locke, Berkeley and Hume. By the time fact to be false. Careful logical investigation of this claim, start-

of Mill this empiricism had solidified into what philosophers call ing with thework of Carnap and the Vienna Circle in the 1930s,
phenomenalism: that is the doctrine that all we can really talk convinced the phenomenalists themselves that the claim was
about are sensations. Even everyday objects, e.g. tables and unfounded. Scientific theories as a whole undoubtedly lead us to
chairs, are really just sets of objective regularities in actual and expect that we will have certain experiences if we perform cer-
possible human sensations, on this conception. As Mill put it, tain actions; but the idea that the statements of science are trans-

physical objects are 'permanent possibilities of sensation'. latable one by one into statements about what experiences we
Another way to put the same idea is to say that all talk that will have if we perform certain actions has now been given up as
appears to be about the physical world, or whatever, is really an unacceptable kind of reductionism. In the second place sen-
just highly derived talk about sensations.
we may be able
sations are necessarily private objects; although

The virtue of this point of view, in the eyes of its holders, was in practice to decide whether or not someone had a sensation by

that enabled them to say clearly what the content was, not
it simply asking them, we are immediately in some kind of epis-
only of science, but of all cognitively meaningful talk whatso- temological trouble if someone raises the question 'how do you
ever. Any scientific theory is really just an 'economical' way of know same sensations with his
that the person associates the

stating a number of facts of the form: if you perform such and descriptions that you do?' If the content of science consists in
such actions, then you will have such and such experiences. The predictions about what sensations any rational being will have if
holder of this view does not have to defend the untenable claim that rational being performs certain actions, then to know what

that scientists are interested only in applications, or only in the that content itself means we would have to be able to tell, for

attainment of practical goals, and disinterested in knowledge for example, whether extraterrestrials if we
encounter any, have the

its own sake. The phenomenalist does not have to deny that we same sensations that we do For
or not, this reason, philoso-
etc.

want to know the nature of black holes, that we want to know phers like Rudolph Carnap and Sir Karl Popper insisted that the
observational predictions of science should be stated in the form
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 if anyone performs such and such actions, then such and such
not just because knowing them might enable us to build better publicly observable events will take place - where both the
machines. But knowing these things is, although it takes refined actions to be performed and the observable events that should
 ^^
182 The impact of science
The impact of science 183

be expected must be described in terms of 'public' objects, e.g.

The contention of Carnap and Popper, that the observation
meter readings, and not in terms of such private objects as sen-
statements of science are couched in physical thing language and
sations.
not in sensation language, is obviously correct if taken as a gen-
To sum up, the older empiricism, or phenomenalism, seemed
eralization about the practice of scientists. When it is erected
to provide us with a tidy criterion of cognitive significance: a
into an epistemological absolute, however, its import becomes
statement is cognitively significant if it is translatable into a
rather momentous. For one thing, if no observation statement at
statement about sensations. But it turned out that either the
all is allowed to talk about sensations, then introspection is ruled
notion of 'translation' is hopelessly vague, or else that statements
out as a mode of scientific observation. Although many psychol-
of science itself failed to satisfy this criterion of cognitive signif-
ogists would agree should be ruled out, the fact is that
that it
[Link] trouble with drawing a sharp line between factual some psychologists, undeterred by both philosophical and psy-
statements and value judgments on the grounds that the former
chological dogma, have gone on performing experiments which
but not the latter are 'translatable' into statements about sensa-
on introspective reports. In fact,
involve at least in part reliance
tions is that the alleged translatability of the first class of state-
Carnap would not have been as dogmatic in prohibiting this as
ments has not been, and apparently cannot be, demonstrated.
some behaviorist psychologists were; he would have permitted
Empiricist reductionism drew a sharp line between the factual
the use of sensation reports, provided that they were construed
and the evaluational, but at the price of giving a wholly distorted not as observation reports but as behavioral data, the 'behavior'
picture of the factual.
being the making of the verbal reports themselves. But what it
Our was to consider answers to
original purpose, however,
means to construe the acceptance of an introspective report as
the question'why is it good to be rational?' The first answer we
an 'inference from verbal behavior' is not altogether clear. Even
considered, and rejected as too narrow, was that rationality
in the case of reports which are not about sensations but about
enables us to discover reliable means/ends connections. Phenom-
physical objects, e.g. 'there is a table in front of me', we do not
enalism came into the story because if phenomenalism were true
normally accept the report unless we have some theory accord-
then the apparent conflict between being interested in a scientific
ing to which the person was in a position to observe the fact that
theory for the sake of its instrumental consequences, and being
he reports. In this sense it is a part of our whole demand for
interested in the theory in order to learn what it tells us about coherence in our world picture that observations should them-
natural processes would dissolve. The conflict between instru- selves be theoretically explainable; if someone claims to have
mental interests and purely theoretical interests could in a sense
observed that there was a table in a certain place by clairvoyance
be finessed. Of course there would still be some kind of differ-
we do not accept this 'observation report' because it does not
ence between these interests; but even purely theoretical interests
cohere with our total body of theory. In this sense every obser-
would be interests in facts which, in the ultimate logical analysis, vation report has some component which could be described as
would have been revealed to be of an instrumental nature. All 'inferential'. On the other hand, when, say, a doctor accepts the
knowledge worthy of the name would have been shown to be
report of a patient that the patient feels pain, it is hard to know
knowledge of means/ends connections; it is just that when we
on the basis of what 'scientific theory' the doctor infers that the
are interested in the means/ends connection because we hope to
patient feels pain from the patient's verbal report; if the general
exploit it for the attainment of some goal, then we call our inter-
assumption that people are in a good position to tell whether or
ests 'practical', and when we are interested in knowing the
not they have a pain counts as a theory, then the general assump-
means/ends connection out of pure curiosity then we call it a
tion that people are in a good position to tell whether or not
theoretical interest. This attempt to reduce all the statements of
thereis a table in front of them also counts as a theory; but it is
science to statements of the form if you perform action A then
hard to see that there is any fundamental methodological differ-
you will get result B has, as we remarked, failed.
ence between accepting someone's say so that there was a table
 The impact of science 185
184 The impact of science

doctrine, after the disappearance of phenomenalism, is like the

in front of them and accepting someone's say so that they had a
appearance of 'primitive' material in a patient's associations in
pain.
therapy, after the 'defenses' have been stripped away. To say
Popper and Carnap would reply that the methodological dif-
that the aim of science is successful prediction (or successful pre-
ference is that the former statement but not the latter is publicly
diction plus something simply described as 'simplicity'), seems
checkable; but they both exaggerate the extent to which obser-
dangerously close to saying that science is pursued only for prac-
vation reports about physical objects are always publicly check-
tical goals; and this is something that no philosopher has wanted
able. Many such reports are made with the aid of instruments
to be put in the position of maintaining. Indeed, the philosophers
which it takes a good deal of training to use. (It is notorious that
who defended a purely instrumentalist conception of science did
learning tothrough a very high-powered microscope
'see'
so not because they were themselves worshippers of the practi-
requires a good deal of specialized training and skill and that not
cal, or narrow minded men who could not appreciate the beauty
everyone is capable of acquiring the skill.) What accepting this
of abstract scientific knowledge for its own sake, but they did so
epistemological dogma does is make it a part of the definition of
rather because they felt that by identifying what is 'cognitively
rationality that rational beliefs are capable of being publicly
significant' with what has value for the making of predictions
checked. Making it part of the definition of rationality is very
they could once and for all rule out all forms of obscurantism
convenient; it makes it unnecessary to provide any argument for
and metaphysics. 'Metaphysics' was for these philosophers sim-
this contention. Perhaps the argument is, at bottom, that what-
ply another name for various kinds of transcendental specula-
ever is not publicly checkable may become a matter of disagree-
tion; it was and 'metaphysical' speculations
religious (in their
ment, and that wherever there is unsettleable disagreement there
sense of 'metaphysical') that they were afraid of.
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
I am suggesting that the appearance in the culture of a philo-
sophical tendency which was hypnotized by the success of sci-
very notion of rationality that what is rationally verifiable is ver-
ence to such an extent that could not conceive of the possibility
it
ifiable to the satisfactionof the overwhelming majority.
of knowledge and reason outside of what we are pleased to call
The Empiricism was fundamentally a sophis-
fact that Logical
the sciences is something that was to be expected given the enor-
ticated expression of the broad cultural tendencies to instrumen-
mously high prestige that science has in the general culture, and
talism and majoritarianism becomes evident, I think, in the later
given the declining prestige of religion, absolute ethics, and tran-
history of this movement. Although the Logical Empiricists had
scendental metaphysics. And I am suggesting that the high pres-
abandoned phenomenalism as early as 1936, for the next twenty
tige of science in the general culture is very much due to the
years, that is until the movement began to break up and disap-
enormous instrumental success of science, together with the fact
pear as a recognizable philosophical tendency, Logical Empiri-
that science seems free from the interminable and unsettleable
cist philosophers of science were fond of talking about 'the aim
debates that we find in religion, ethics and metaphysics.
of science' and fond of identifying the aim of science with pre-
Since, however, the professional philosophers who, as it were,
diction (with some additions and qualifications which I shall dis-
rationalized the instrumentalist tendency in the culture were not
cuss in a moment). The idea that the aim of science is prediction
themselves vulgar-minded or purely practical persons, it is not
was the fundamental idea of positivism from its beginnings in
surprising that they themselves felt inclined to widen the descrip-
the writings of Auguste Comte. As we saw, this idea had at least
tion of 'the aim of science' somewhat, so as more explicitly to
some kind of serious philosophical rationale as long as phenom-
leave room for aims other than just successful prediction. And
enalism was in vogue; for then one could argue that all cogni-
so we find other aims being listed by Logical Empiricist writers
tively meaningful statements were predictions in disguise, or
in the 1940s and 1950s: the discovery of laws, retrodiction (i.e.
infinite sets of predictions in disguise. The reappearance of this
186 The impact of science The impact of science 187

and the dis-

the prediction of past as opposed to future events), reason that we call the sciences? The answer, of course, is that
covery of 'explanations', by which these writers meant simply these philosophers did not doubt that 'science'
seriously
the deduction of predictions and retrodictions from laws. exhausts reason. But why did they not doubt this? They did not
What happened here is interesting. In order to make it explicit doubt this because for them the opposition was not between sci-
that science is interested in discovering laws of nature for their ence, in the sense of knowledge proceeding by essentially the
own sake, and not merely for the sake of the predictions to methods of the empirical and mathematical sciences, and infor-
which those laws lead, these writers replaced the simple formula mal reason, proceeding by methods which might be adapted to
'the aim of science is successful prediction' with a list. The list but no less capable
interests different than those of the sciences,

is in fact open-ended: laws of nature turn out to include not only of having genuine standards. Rather the opposition was between
laws of nature in the strict sense, i.e. statements which it is phys- knowledge proceeding by the methods of the sciences, and
ically impossible to falsify, but also the so-called 'laws' of evo- pseudo knowledge pretending to proceed by revelation, or some
lutionary theory, which are really descriptions of general ten- kind of funny transcendental faculties. Reason had to be co-
dencies which may at some time, owing to the action of extensive with science because What Else Could It Be? Never-
intelligent life, cease to hold, and even statements about the theless the claim of these philosophers that reason is co-extensive
purely contingent dispositions of individual groups and even with science landed them some
in peculiar predicaments. Since
individual organisms. To
say that scientists are trying to discover they did not wish to deny that there is such a thing as historical
'laws of nature', including physically contingent generalizations knowledge, for example, they were committed to the position
which hold for long periods of time and which have wide that history is a science, and even to the position that what the
explanatory significance, such as those upon which evolutionary historian is really trying to do is to subsume individual state-

theory is based, or those upon which the science of economics is ments about the past under laws - a claim about history that
based, and seek to discover significant truths concerning the dis- seems false on the face of it.
positions of groups and individual organisms, and seek to organ- It is, perhaps, not surprising that the Logical Empiricist ten-
ize all of these into a deductive (and inductive) structure, is, of dency began to disintegrate by 1950. We have been looking at
course, quite correct. But why this particular list? this tendency solely from the point of view of one question; the

The reason for the list is that it is thought to be embracing Logical Empiricists had a great many different philosophical
enough to include of the kinds of truths that scientists seek to
all interests,and they made many valuable contributions. Neverthe-
discover, certainly in physical science, and narrow enough not less, from the point of view of the question we have been asking,

to include any of the objectionable ('cognitively insignificant') which is 'what good is rationality?', the Logical Empiricist
material. The old search for a 'criterion of cognitive signifi- movement represented a reasoned philosophical defense of the
cance', such as 'a sentence is meaningful if and only if it is pos- view that the answer, and the sole answer, to the question is that
sible to verify or falsify it', has been replaced by a list of types of rationality is good for the discovery of means/ends connections.
statements, such that a statement is to be admitted if it is of one The philosophical doctrine of phenomenalism provided the Log-
of these types and otherwise to be rejected. But why was this at ical Empiricists with an interesting philosophical defense of this

all a plausible move for a philosopher to make? Even if it is true claim. When phenomenalism was given up, and the philo-
the
that all of the statements in the disciplines that we call 'sciences' sophical defense of the claim was replaced by the bare claim, and
are of these types (and it is not at all clear that this is the case - even more when the bare claim was made more 'reasonable' by
is historical explanation really just subsumption of retrodictions allowing exceptions, modifications, etc., the whole cutting
under 'laws'?), does it follow that the verification of these types power of the movement disappeared. The trouble with the posi-
of statements and just these types of statements is the aim of tion that the aims of reason itself are the discovery of predic-
reason itself, and not just the aim of the special applications of tions, retrodictions, laws of nature, and the systematization of
188 The impact of science The impact of science 189

all of these, and that these are all the aims of reason, is that there ble answer which has had a considerable amount of appeal at
is simply no reason to believe it, I don't mean to say that there is different times. Many philosophers have believed that science
reason to believe that it is false; if the notion of a law of nature proceeds by following a distinctive method; if there is in fact a
is widened so that the discovery of laws of nature includes the method with the property that by using that method one can
discovery of dispositional statements about individual orga- reliably discover truths, and if no other method has any real

nisms, and the notion of a disposition is so wide (or so vague) chance of discovering truths, and if what explains the extraor-
that the statement that a certain scientist is envious of his col- dinary success of science, and the persistence of controversy in
league's reputation counts as a statement of a 'disposition', and fields other than science, is that science and science alone has

the statement that that scientist told a certain joke because he consistently employed this method, then perhaps rationality, to

was jealous of his colleague's reputation is a 'subsumption of a the extent that there is such a thing, should be identified with the
particular event under law', then it may be that everything one possession and employment of this method. The answer to the
says can be interpreted as either stating general laws or as sub- question 'Why is it good to be rational?' would then be that it is
suming descriptions under general laws. Perhaps even saying of good to be rational because if one is rational one can discover
truths (of whatever kind one interested in), whereas if one is
someone that he is morally good can be construed as ascribing a is

'disposition' to that someone. No, the trouble with trying to not rational one has no real chance of discovering truths, save
specify the aims of cognitive inquiry in general by means of a list by luck. Like the instrumentalists' view this view went through
of this kind is that the list itself has to be construed: if the terms a philosophical history of rise, stagnation, and decline. From the
in the list are construed in a more or less literal way, then the publication of Mill's Logic in the 1840s until the publication of
kinds of statements in the list would not even include all of the Carnap's Logical Foundations of Probability, influential philos-
sorts of statements that scientists are interested in discovering, ophers of science continued to believe that something like a for-
certainly not if 'scientist' includes historian, psychiatrist, and mal method ('inductive logic') underlies empirical science, and
sociologist; while if the terms in the list are construed so leni- that continued work might an explicit statement of this
result in

ently that there is no difficulty in construing the statements made method, a formalization of inductive logic comparable to the
by historians (and descriptive statements in the language of formalization of deductive logic that was achieved starting with
everyday psychology) as belonging to types included in the list, the work of Frege in 1879. If such a method had been discov-
then the list becomes worthless. In any case, in the absence of ered, then even if this did not by itself prove that the method
any epistemological explanation of why statements of these exhausts rationality, still the burden of proof would have been
kinds and only statements of these kinds should be capable of very much upon those who claimed that there were truths which
rational verification such a list would only be a mere hypothesis could be justified or shown to be rationally acceptable by any
about the limits of rational inquiry. A mere hypothesis,
whether other method.
in the form of a list or in some other form, could not have the According to the most influential school, the so-called 'Baye-

exclusionary force that the Logical Empiricists wanted 'criteria sian' school, the general character of this inductive method that

of cognitive significance' to have. 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-
'Method' fetishism
ble observations expressible by 'observation sentences' in this
Since the answer to the question 'Why is it good to be rational?' formalized language. We also assume that the various
cannot be simply that rationality enables one to attain practical hypotheses under consideration are expressed by formulas of
goals, and cannot be simply that rationality enables one to dis- this language. The problem of inductive logic is taken to be the
cover means/ends connections, we may consider another possi- problem of defining a 'confirmation function', that is a probabil-
190 The impact of science
The impact of science 191

ity function which will determine the mathematical probability

the method itself is a set of substantive factual beliefs (or degrees
of each one of the hypotheses relative to the observational evi-
of belief) about the world. This is the way in which many phi-
dence or, in another terminology, the 'degree of support' the
losophers of science today view the matter; increasingly it is
evidence lends to each of the alternative hypotheses. Usually one
coming to be believed it is not possible to draw a sharp line
assumes that one knows the probability that the given evidence
between the content of science and the method of science; that
would have been obtained if each of the alternative hypotheses
the method of science in fact changes constantly as the content
were true; this is the so-called 'forward probability', i.e. the
of science changes. Bayes' theorem, if it really does capture the
probability of the evidence given the hypothesis. What we wish logic of theory confirmation, provides a way of formalizing this
to calculate is the so-called 'inverse probability', that is the prob-
dependence of the method of science upon the content of science,
ability of the hypothesis given the evidence. Bayes' theorem gives
through the need for a prior probability function.
this 'inverse probability' as a simple function of the forward
To put the matter somewhat more abstractly, we might say
probabilities and certain other probabilities, the so-called 'prior
that the 'method' fetishist assumes that rationality is inseparable.
probabilities' of the alternative hypotheses,
i.e. the probabilities
But Bayes' theorem indicates that this is not the case; that we
or 'subjective degrees of belief assigned by scientists to those
can separate rationality, even in the area of science, even in the
alternative hypotheses prior to examining the observational evi-
special area where we are dealing with theories for which the
dence.
forward probabilities are computable, into two parts: a formal
The 'forward probabilities' are indeed easy to calculate in the
part, which can be schematized mathematically and pro-
two most common cases: they are easy to calculate when (a) the
grammed on a computer, and an informal part which cannot be
hypothesis actually implies the evidence (in this case the 'for-
so schematized and which depends on the actual changing beliefs
ward probability' of the evidence given the hypothesis is one); or
of scientists. Now it would be nice, to put it mildly, if the formal
when (b) the hypothesis is itself a statistical or stochastic hypoth-
part of rationality sufficed to guarantee good results. If we could
esis part of whose content is that the particular evidence
say that provided scientists make their observations carefully,
obtained should occur with a certain probability r. The difficulty
gather sufficient observations, and calculate degrees of support
in applying Bayes' theorem - a both
difficulty so serious that
according to Bayes' theorem, then eventually they will come into
philosophers and statisticians are deeply divided over the impor-
agreement even if they disagree at the beginning, owing to the
tance and usefulness of Bayes' theorem in the case of the confir-
difference in their subjective degrees of belief, then all would be
mation of theories — is the need for a prior probability metric, a
well. But there are two things wrong with this happy picture.
set of 'subjective degrees of belief, in the terminology of De The first thing wrong is that even if we could show that in the
Finetti and Savage.
long run the 'prior probability function' cancels out, or that sci-
Let us confine ourselves for the moment to hypotheses which
entists with different prior probability functions eventually come
are such that the 'forward probabilities' can really be computed.
into agreement provided they continue to gather more evidence
For hypotheses of kind the method just described is indeed
this
and to use Bayes' theorem, it would still be necessary that this
a purely formal method; that is we could program a computing
convergence be reasonably rapid. If scientists with different prior
machine to compute the degrees of support of the various hypoth-
probability functions will not come into agreement until the phe-
eses given the appropriate 'inputs'.
But the inputs would have
nomenon to be predicted has already taken place, or until mil-
to include not only the computable 'forward probabilities', but
lions of years have passed, then, in the short run, the fact that
also the prior probability metric in the given context. If we think
there is some mathematical guarantee of eventual convergence is
of this prior probability metric as representing the scientists'
useless; the trouble with long-run justifications is that the long
antecedent beliefs about the world, as the term 'subjective prob-
run may be much too long. In the famous words of John May-
ability function' suggests, then it looks as if one of the inputs to
nard Keynes, 'in the long run we'll all be dead'. The second thing
192 The impact of science The impact of science 193

wrong is that does turn out, as a matter of fact, that differences

it
human scientist would count as a 'method', there is no reason to

in the prior probability function can lead to violent differences think that a 'method' in this sense would be independent of judg-
in the actual degrees of support assigned to theories, and that ments about aesthetics, judgments about ethics, judgments
these differences can amount to what would ordinarily be con- about whatever you please. The whole reason for believing that
sidered as gross irrationalities. the scientific method would not apply to or presuppose beliefs
To put this last point in another way, a scientist will only about was the belief that the sci-
ethical, aesthetic, etc., matters

assign degrees of support to hypotheses that look 'reasonable' if entific method was a formal method, after all.
he starts out with a 'reasonable' prior probability function. If a My discussion has depended on assuming the correctness of
person only obeys the formal part of the description of ratio- one particular approach to formalizing the scientific method, the
nality, if he is logically consistent and assigns degrees of support so-called 'Bayesian' approach. But similar problems arise in each

in accordance with Bayes' theorem, but his prior probability of the other approaches that have been attempted. Even if one
function is extremely 'unreasonable', then his judgments of the tries to isolate some small part of the inductive method which

extent to which various hypotheses are supported by the evi- would not be as 'high-powered' as the confirmation of theories,
dence will be (as scientists and ordinary people actually judge and which would be more in line with what Bacon understood
these matters) wildly 'irrational'. Formal rationality, commit- by 'induction', that is, even if one tries to isolate a method for
ment to the formal part of the scientific method, does not guar- confirming simple generalizations by examining a sufficient
antee real and actual rationality. number of and 'projecting' the truth of the generaliza-
instances

The extent to which this is true is in fact rather shocking. tion, similar problems arise. Nelson Goodman 5 has shown that
Arthur Burks has in fact shown that there are even 'counter no purely formal rule for inductive projection can even be free
inductive prior probability functions'. That is, there is a certain from inconsistencies; before a formal rule can even hope to yield
logically possible prior probability metric such that if a scientist consistent results one has to have in advance segregated the

had that metric then as more evidence came in for a hypothesis predicates of the language into those one is willing to regard as
(using the term more evidence on the basis of our normal induc- 'projectable' and those which one will treat as non-pro jectable'. '

tive judgments) then the scientist would assign lower and lower The fact that even the most elementary part of induction turns
weight to the hypothesis for a very long time. out to have a part (namely the division of the vocabulary into a
One way out of the difficulty might be to try to supplement projectable and a non-projectable part) which is informal, again,

the present formal account of scientific method by a further set strongly supports the conclusion suggested by our discussion of

of formal rules which would determine which priors are reason- Bayes' theorem, that one cannot draw a sharp linebetween the
able (henceforth, I shall refer to a prior probability function sim- actual beliefs of scientists and the scientific method. What Good-
ply as a 'prior', in conformity with common statistical usage), man did was to invent a predicate 'grue' which applies to things
and which priors are unreasonable. But there does not seem to just in case they are observed prior to the year 2000 and green
be any good reason to think that there would be a set of rules or not observed prior to the year 2000 and blue. Prior to the
which could distinguish between reasonable and unreasonable year 2000, everything which is examined and seen to be green is
priors and which would be any simpler than a complete descrip- also examined and seen to be grue. Any formal rule of projection

tion of the total psychology of an ideally rational human being. which told us that when we've examined a certain number of
The hope for a formal method, capable of being isolated from things, say emeralds, with a property P then we are allowed to

actual human judgments about the content of science (that is, infer that 'all emeralds are P' would permit us to make the con-

about the nature of the world), and from human values seems to
have evaporated. And even if we widen the notion of a method s
See his Fact, Fiction and Forecast, 2nd ed., Hackett (1977), first published
so that a formalization of the psychology of an ideally rational in 1954.
194 The impact of science The impact of science 195

tradictory inferences that 'all emeralds are green' and 'all emer- yet been ruled out, or thinks in the more conventional way that
alds are grue'. And Goodman convincingly shows that all one is trying to compute degrees of support for hypotheses, the
attempts to rule out 'bizarre' predicates like 'grue' on purely for- need for an informal element corresponding to a Goodmanian
6 decision that certain predicates are projectable and others are
mal grounds cannot work.
There is actually a close connection between Goodman's dif- not, or corresponding to the acceptance of a Bayesian prior, is
ficulty in the case of Baconian induction and the need for a prior still necessary.
in connection with Bayes' theorem. Suppose the two hypotheses At this may wonder, if there is no such thing
point the reader
the scientist has to choose between (at some time prior to the as the scientificmethod, or if the method, in so far as it can be
year 2000) are 'all emeralds are green' and 'all emeralds are formalized, depends on inputs which are not formalizable, then
grue'. Let us suppose that the relevant evidence is that a great how do we account for the success of science? It is undeniable
many emeralds have been examined and all found to be green that science has been an astoundingly successful institution. We
(and hence all found to be grue as well). If the scientist computes tend to feel that the its success must have something
reason for
the degree of support of the two hypotheses using Bayes' theo- to do with the between the ways in which scientists
differences
rem then it turns out that he can either find a much higher degree proceed to gather knowledge and the way in which people tra-
of support for the normal hypothesis ('all emeralds are green') or ditionally proceeded to gather knowledge in the prescientific
a much higher degree of support for the abnormal hypothesis ages. Is this wholly wrong? The answer is that it is not. The
('all emeralds are grue') or an equal degree of support for both alternatives that we have to choose between are not that science
hypotheses, depending on his prior. If one's subjective probabil- succeeds because it follows some kind of rigorous formal algo-
ity metric assigns a much higher prior probability to 'all emer- rithm, on the one hand, and that science succeeds by pure luck.
alds are green' than to 'all emeralds are grue', then one will, in Starting in the fifteenth century, and reaching a kind of peak in
fact, behave as if one were projecting 'green' and not projecting the seventeenth century, scientists and philosophers began to put
'grue'. From a Bayesian point of view the need for a decision as forward a new set of methodological maxims. These maxims are
to which predicates are projectable and which are not before one not rigorous formal rules; they do require informal rationality,
can make an induction is just a special case of the need for a i.e. intelligence and common sense, to apply; but nevertheless
prior. they did and do shape scientific inquiry. In short, there is a sci-

Karl Popper has suggested that one should accept the most entific method; but it presupposes prior notions of rationality. 7
falsifiable of the alternative hypotheses; but it turns out that his It is not a method de novo which can serve as the be all and end
formal measures of falsifiability will yield different results all, the very definition of rationality.
depending on which predicates of the language one chooses to One
of the most important methodologists of the seventeenth
take as primitive. Whether one thinks of the scientist, as Popper century was the physicist Boyle. Prior to the seventeenth century,
does, as trying to find the most falsifiable hypothesis that has not physicists did not sharply distinguish between actually perform-
ing experiments and simply describing thought experiments
6
Goodman's own solution is to consider form plus the history of prior
which would confirm theories that they believed on more or less

projection of the predicates involved in the inference (along with certain a priori grounds. Moreover, physicists did not see the need to
related matters, e.g. 'entrenchment' and 'over-riding'). On Goodman's publish descriptions of experiments which failed. In short,
proposal it would follow that which had always projected such
a culture
experiments were conceived of largely as illustrations for doc-
'crazy' predicates as his celebrated predicate 'grue' would now be
perfectly justified in doing so - their inferences would now be induct-
7
ively valid'! Mill himself concedes this (in a remarkably grudging tone of voice) when
While I agree with Goodman with past practice is an impor-
that fit he writes that we cannot expect the inductive method to work 'if we
tant principle in science, Goodman's version of this principle is too suppose universal idiocy to be conjoined with it' (Utilitarianism, Chapter
simple and too relativistic. 2).
196 The impact of science The impact of science 197

trines believedon deductive and a priori grounds; not as evi- hypothesis as the one to go on for the time being, and repeat the
dence for and against theories. Boyle wrote manuals of experi- entire procedure. Since the elimination of all the theories but one
mental procedure, he emphasized the need for a sharp is made on deductive grounds - a theory is eliminated when it

distinction between thought experiments and actual experi- implies a prediction which is definitely falsified -no use of
ments, and he emphasized the need to give a complete descrip- Bayes' theorem is required, and no estimation of degrees of sup-
tion of all the experiments one performed, especially including port is involved, Popper claims.
the experiments that failed. Boyle was himself a disciple of the One problem with Popper's view is that it is not possible to

philosopher Francis Bacon, and Boyle was undoubtedly led to test all strongly falsifiable theories. For example, the theory that
an appreciation of the importance of these rules by Bacon's if Iput a flour sack on my head and rap the table 99 times a
inductive outlook; in fact, however, the specific instructions demon will appear is strongly falsifiable, but I am certainly not

given by Boyle may have been more important or as important going to bother to test it. Even if I were willing to test it I could
inshaping the course of physical inquiry as the more abstract think of 10 100 similar theories, and a human lifetime, or even the
and schematic defense of inductive procedure given by Bacon. lifetime of the human species, would not suffice to test them all.

Turning away from trying to establish theories a priori For logical reasons, then, it is necessary to select, on methodo-
towards trying to test theories by deriving testable conclusions logical grounds, a very small number of theories that we will
from them and performing experiments certainly was a meth- actually bother to test; and means that something like a
this

odological shift. As we have seen, however, we cannot simply prior selection is involved even in the Popperian method. As I
identify being rational with believing theories solely because they remarked above, even Popper's computations of degrees of fal-
are supported by carefully performed experiments. For one sifiability are sensitive to which predicates one considers as

thing, even in science it is not always possible to perform con- primitive in one's language, and in that sense even the notion of
trolled experiments. Sometimes one has to relyon passive obser- falsifiability requires a prior decision analogous to Goodman's
vation rather than on the kind of active intervention which is decision that certain predicates are 'projectable' and others are
implied by the term 'experiment'. And, as we saw before, even not. Let us waive these technical points, however, which are not
when one has carefully performed experiments for the purpose of interest to us in our present discussion, in any case. Even if
of choosing among alternative theories, the estimation of the the Popperian method is incomplete, and requires to be supple-
degree to which the experimental results support the various mented by a more intuitive method which we are not able to
alternative theories is still a wholly informal matter. formalize at the present time, could it not be that
it describes a

Against what we have been urging, Karl Popper has repeatedly necessary condition, if not a sufficient condition, for scientific
argued that there is a distinctive scientific method, it can be rationality? Could it not be, in short, that a necessary condition
stated, and we should rely only on it for discovering the nature for the acceptability of a scientific theory be that it have survived

of the world. a Popperian test? The Popperian test itself may involve a prior
Popper does think, however, that there are notions of ration- selection of theories to test which is itself informal and for which
alitywhich are wider than scientific rationality and which do we do not have an algorithm; the calculation of which theories
apply to the making of ethical decisions. are most strongly falsifiable may involve informal decisions for
In Popper's conception, set forth in his influential book, The which we do not have an algorithm; but we could still insist that
Logic of Scientific Inquiry (Logik der Forschung), and in subse- no theory be accepted unless a set of theories has first been
quent publications, Popper has argued that the scientific method selected all of which are intuitively 'highly falsifiable', and unless
consists in putting forward 'highly falsifiable' theories; theories all those theories except the one which we accept have been sub-

that imply risky predictions. We then proceed to test all of the sequently refuted by carefully performed experiments. In short,
theories until only one survives. We then accept the surviving could it not be that the advice we ought to give the scientist is:
198 The impact of science The impact of science 199

proceed as Popper advises you should proceed, and, where Pop- a legitimate form of scientific inference-drawing, even when the

on your 'best explanation' inferred to not strongly falsifiable in Pop-

is
per's methods are not capable of being formalized, rely
intuition as to how the Popperian maxims should be interpreted? per's sense. The scientific method has now become a tremen-
And might it not be the case that the Popperian method, vague
dously vague thing; 9 but this we expected to happen, anyway, in
and informal as it is in part, exhausts not only the notion of view of the formal results in inductive logic we described above.
scientific rationality, but all of cognitive rationality, that is,
Could it be that the 'scientific method', described this vaguely is
might it not be the case that a statement is warrantedly assert- now exhaustive? And could itbe that given even such a vague
ible, or rationally acceptable, if and only if it is implied by a
description, it is clearly the case that no value judgment is capa-
theory which can be accepted on the basis of a Popperian test? ble of being verified or confirmed by this method? The answer is
that if the scientific method is described simply as 'make experi-
The answer is that such a conception of rationality is too narrow
even for science. For one thing it would rule out the acceptance ments and observations as carefully as you can, and then make
of one of the most successful and widely admired of all scientific inferences to the best explanation and eliminate theories which

theories, Darwin's theory of evolution by natural selection. This

can be falsified by crucial experiments', then it is impossible to
is a consequence which Popper himself is willing to accept with
see what cannot be verified by a method so vaguely described.

equanimity, but certainly the scientific community is not. The Suppose, for example, I want to verify the statement 'John is a
theory of natural selection is not highly falsifiable; it does not bad man.' I might argue as follows: 'The following are observed
facts, that John is inconsiderate, that John is extremely selfish,
imply definite predictions such that if they come out wrong then
the theory is refuted. We accept the theory of natural selection and that John is a very cruel person. Someone who is incon-
siderate, selfish and cruel is prima facie a bad person; therefore
not because it has survived a Popperian test, but because it pro-
vides a plausible explanation of an enormous amount of data, John is a bad man.' There are two points at which a defender of
because it has been fruitful in suggesting new theories and in the view that 'value judgments' cannot be 'scientifically verified'

linking up with developments in genetics, molecular biology, might object to this argument. He might object to the last step;

etc., and because the alternative theories actually suggested have

that is the step horn John is John is inconsiderate, John is
cruel,

either been falsified or seem wholly implausible in terms of back- extremely selfish, to John is a bad man. Admittedly this is a con-

ground knowledge. In short, we accept the Darwinian theory of ceptual step; the claim is that there is a conceptual link between

evolution by natural selection as what Peirce called an 'abduc- being cruel, inconsiderate, and selfish and being morally bad. 10
tion', or what has recently been called an 'inference to the best
explanation'. This is exactly the kind of inference that Popper 9
we could restrict the term 'scientific method* to refer to the
Alternatively,

wanted to drive out of science; but scientists are not going to be conscious application of maxims of experimental procedure, as I
recommended in Meaning and the Moral Sciences, and just stop trying
persuaded by Popper that they should give up theories which are to make it so elastic that it can cover everything we call 'knowledge*.
10 The
not strongly falsifiable in cases where those theories provide fact that a truth or an inference is of the sort we call 'conceptual'

good explanations of vast quantities of data, and in cases where does not mean that it must be purely linguistic in character (i.e. true by
virtue of arbitrary linguistic conventions). Philosophers of many
no plausible alternative explanation is in the field. Indeed, as I different tendencies have seen that concepts, fact, and observations are
have pointed out in another publication, 8 Popper exaggerates interdependent. As we remarked in Chapter 6, concepts are shaped by

the extent to which even the theories of classical physics are 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
themselves strongly falsifiable.
exactly analogous to the following inference involving 'conscious*. 'John
We weaken our description of the scientific method still fur- is speaking intelligently, acting appropriately, and responding to what

ther, then, by allowing the 'inference to the best explanation' as 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
8
See 'The Corroboration of Theories', in my Mathematics, Matter and that inconsiderateness, selfishness, and cruelty are prima facie reasons

Method. for attributing moral badness.

200 The impact of science

Of course, if there are no conceptual links among the moral

predicates then this is an invalid step; why should one
but
believe that there are no conceptual links among the moral pred-
icates? Itmight 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
Values, facts and cognition
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-
I argued in Chapter 6 that 'every fact is value loaded and every
tational force that A on B can be described as caused by
exerts
(the mass of) A simply by virtue of understanding Newton's the- one of our values loads some fact'. The argument in a nutshell
ory. was that fact (or truth) and rationality are interdependent
Of course, if we describe the scientific method as consisting in notions. A fact is something that it is rational to believe, or, more

the drawing of 'inferences to the best explanation', or whatever, precisely, the notion of a fact (or a true statement) is an ideali-

from 'observational statements' which are themselves in value- zation of the notion of a statement that it is rational to believe.

neutral language, then we can rule out 'John is inconsiderate' 'Rationally acceptable' and 'true' are notions that take in each

and 'John is selfish' as 'observation statements' (although, in other's wash. And I argued that being rational involves having
particular cases, it might be easier to get agreement on these than criteria of relevance as well as criteria of rational acceptability,
on, say, whether an object is mauve). But such statements occur and that all of our values are involved in our criteria of rele-
constantly in the writings of, for example, historians. That his- vance. The decision that a picture of the world is true (or true by

tory, clinical psychology and ordinary language description can our present lights, or 'as true as anything is') and answers the
really avoid words like 'considerate' and 'selfish' altogether is relevant questions (as well as we are able to answer them) rests

doubtful (and where to draw the line would be an immense on and reveals our total system of value commitments. A being
problem: is 'stubborn' value-neutral? is 'angry' value-neu- with no values would have no facts either.
tral? for that matter, is 'twisted her arm savagely' value-neu- The way in which criteria of relevance involve values, at least
tral?). But, in any case, to identify rationality with scientific indirectly, may be seen by examining the simplest statement.
rationality so described would be to beg the question of the cog- Take the sentence 'the cat is on the mat'. If someone actually
nitive status of valuejudgments; it would be to say these judg- makes this judgment in a particular context, then he employs
ments are not rationally confirmable because they are value conceptual resources - the notions 'cat', 'on', and 'mat' - which
judgments, for rationality has been defined as consisting exclu- are provided by a particular culture, and whose presence and

sively of raw and neutral observation and the drawing of infer- ubiquity reveal something about the interests and values of that

ences from value-neutral premisses. But why should one accept culture, and of almost every culture. We have the category 'cat'

such a definition? 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'
202 Values, facts and cognition Values, facts and cognition 203

because we regard the division of inanimate things into artifacts Acceptability and relevance are interdependent in any real
and non-artifacts as significant, and we are further interested in context, however. Using any word -whether the word be
the purpose and nature a particular artifact has. It is relevant - involves one in a
'good', or 'conscious', or 'red', or 'magnetic'
that it is a mat that the cat is on and not just a something. We history, a tradition of observation, generalization, practice and
have thecategory 'on' becausewe are interested inspatial relations. theory.
Notice what we have: we took the most banal statement im- It one in the activity of interpreting that tradi-
also involves
aginable, 'the cat is on the mat', and we found that the presup- tion, and of adapting it to new contexts, extending and criticiz-
positions which make this statement a relevant one in certain ing it. One can interpret traditions variously, but one cannot
contexts include the significance of the categories animatel apply a word at all if one places oneself entirely outside of the
inanimate, purpose, and space. To a mind with no disposition tradition to which it belongs. And standing inside a tradition
to regard these ^relevant categories, 'the cat is on the mat' would certainly affects what one counts as 'rational acceptability'. If
be as irrational a remark as 'the number of hexagonal objects there were one method one could use to verify any statement at
in this room is 76' would be, uttered in the middle of a tete-a-tete all, no matter what concepts it contained, then the proposed sep-

between young lovers. aration of the ability to verify statements from the mastery of a
Not only do very general about our value system show
facts relevant set of concepts might be tenable; but we have already
themselves in our categories name, term for a
(artifacts, species seen that there is no reason to accept the myth of the one
spatial relation) but, as we saw in Chapter 6, our more specific Method.
values (for example, sensitivity and compassion) also show up in
the use we make of specific classificatory words ('considerate',
The two-components theory
'selfish').To repeat, our criteria of relevance rest on and reveal
our whole system of values. Our present intuitions about rationality seem to be in conflict;
The relevance of this discussion of relevance to the question certainly no one philosophical theory seems to reconcile them
raised in the preceding chapter ('What is the value of rational- all. On the one hand, it is simply not true that we never judge
ity?') is immediate. If 'rationality' is an ability (or better, an inte- ends as rational or irrational; on the other hand, when we are
grated system of which enables the possessor to deter-
abilities) confronted with a case like that of the hypothetical 'rational
mine what questions are relevant questions to ask and what Nazi' we do not see how to justify criticizing such an intelligently
answers it is warranted to accept, then its value is on its sleeve. elaborated and considered system of ends as irrational even if we
But it needs no argument that such a conception of rationality is find it morally repellant.
as value loaded as the notion of relevance itself. One way in which it has been suggested we might resolve these
It be objected that I have lumped together factors that
may problems is the following: assuming a sharp fact/value dichot-
belong apart, however, and thereby masked a sort of sleight of omy, we can justify condemning the man who is only interested
hand. The very fact that I have spoken of two factors, rational in knowing the number of hairs on people's heads as irrational
acceptability and relevance, testifies, it may be claimed, to the on the ground that he has an inadequate perception of facts
persistence and permanence of something like the fact/value (what 'adequate' means is a problem, of course). The Nazi only
dichotomy. A rational person, on the conception the objector disagrees with us about values, which is why he is not irrational.
has in mind, would be one who could tell what was and what In between cases can be handled, perhaps, along the lines sug-

was not warrantedly assertible; what a person chose to regard as gested by Bernard Williams. In particular, then, our argument
interesting, or important, or relevant might have bearing in eval- against the method fetishist, that we cannot, without circularity,
uating his character or even his mental health, but not his cog- rule out of the 'observational evidence' such descriptive judg-
nitive rationality, the objector would say. ments as 'John is considerate', can be met by advancing the the-
204 Values, facts and cognition Values, facts and cognition 205

ory that the ordinary language moral-descriptive vocabulary has the truth of this statement are context dependent, interest rela-
two 'meaning components' simultaneously. One component is a tive, and vague. There is no reason to think that even 'in princi-
factual component; there are certain generally accepted stan- ple' there exists a finite expression in the language of physical
dards for considerateness, and 'John is considerate' conveys the theory which (in any physically possible world) is true of an X if

information that John meets those standards. 1 But there is also and only if that X is thinking about Vienna. Not only might it

an emotive meaning component: 'John is considerate' conveys a be false that a finite equivalent in physicalist language exists for
'pro-attitude' towards a certain aspect of John's conduct. What the ordinary language statement 'X is thinking about Vienna';
can be rational, it is claimed, is the acceptance of the factual even if such an equivalent does exist, the equivalence would be
component of the statement 'John is considerate'; acceptance of equivalence on the basis of an empirical theory or group of the-
the emotive meaning component, sharing the 'pro-attitude', is ories which are themselves not known (perhaps they are so com-
what is neither rational nor irrational. plicated human beings will never know them), and which are
The notion of the 'factual' here comes from the individual phi- certainly not part of the meaning of 'X is thinking about
losopher's preferred view of the Furniture of the World. For a 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-
f
materialist philosopher, the 'factual' component in the meaning
of any statement has to consist in a statement expressible in the fied) brain state.'

vocabulary of physical science. But a difficulty comparable to What holds for

f
X is thinking about Vienna' will hold for any
the difficulty that plagued phenomenalism at once arises. ordinary language predicate whose conditions of application do
not mesh well with those which govern physical concepts. X
e
Phenomenalism, we recall, was the doctrine that all meaning- is

ful statements are translatable without residue into statements considerate' — even 'X is brown', 'X is an earthquake', and 'X is

about sensations. 'Physicalism' (as the type of Materialism we a person' - are also not translatable into the language of 'physi-
are discussing came to be called) was the doctrine that the whole cal theory'. What this means is that, if there are two components
'factual' meaning of any statement can be translated without res- to the meaning of 'X is considerate', then the only description
idue into the language of physics. And, once again, the doctrine we can give of the 'factual meaning' of the statement is that it is

seems to be false. true if and only if X is considerate. And this trivializes the notion
To why, consider not a moral-descriptive statement but a
see of a 'factual component'.
psychological statement, say 'X is thinking about Vienna.' It is To say that the 'two components' theory collapses is not to
obvious that even if there are necessary and sufficient conditions deny that 'X is considerate' normally has a certain emotive force.
expressible in terms of brain states or whatever for an arbitrary But it does not always have it. As we pointed out in Chapter 6,
organism to be thinking about Vienna, it would take an we can use the statement 'X is considerate' for many purposes:
unimaginably perfected neurological (or, perhaps, functionalist- to evaluate, to describe, to explain, to predict, and so on. Distin-
psychological) theory to say what these are. The conditions for 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

a problem however. If the claim is

considerate.'
1
What 'standards' means here 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 itwould follow from this analysis of the 'factual Moore and the 'Naturalistic Fallacy'
component* that there cannot be a person whom a sensitive judge would
correctly classify as 'considerate' even though most speakers disagree. Weber's claim that 'value judgments' cannot be rationally con-
Such an account of the 'factual component* would simply amount to the firmed was the source of the present fact/value dichotomy; but
claim that all truths {at least about 'standards') must be 'public*, but
the dichotomy was reinforced by G. E. Moore (contrary to his
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.) own intentions). Writing at a time when Bertrand Russell and
206 Values, facts and cognition Values, facts and cognition 207

John Maynard Keynes, along with other future members of the ter 4, this would rule out such accepted scientific discoveries as

Bloomsbury group, were still young students, Moore argued for the discovery that the magnitude temperature is the same mag-
the thesis that Good was a 'non-natural' property, i.e. one totally nitude as the magnitudemean molecular kinetic energy. (One
outside the physicalistic ontology of natural science. His defense could use Moore's 'proof to show that temperature must be a
of non-naturalism backfired; his students may have been con- 'non-natural property', in fact. For one is not contradicting one-
vinced by Moore that there were such things as 'non-natural when one says x has temperature T but x does not have
self
(

properties' (although Russell, at least, was to lose the faith) but mean molecular kinetic energy E', where £ is the value of the
later philosophers of a naturalistic kind tended to feel that kinetic energy that corresponds to temperature T, even if the

Moore had provided a reductio ad absurdum of the idea that statement is always false as a matter of empirical fact. So, Moore
there are such things as value properties. In the 1930s Charles would have to conclude, Temperature is only correlated with

Stevenson and the Logical Positivists were to advance the 'emo- Mean Molecular Kinetic Energy; the two properties cannot lit-
tive theory of ethics', that is X is good' means 'I
the theory that
f
erally be identical.) In fact, Moore conflated properties and con-

approve of X, do so as well!', or something of that kind. Value cepts. There is a notion of property in which the fact that two

properties began to be rejected on epistemological grounds, but concepts are different (say 'temperature' and 'mean molecular
even more on ontological grounds; as John Mackie 2 has recently kinetic energy') does not at all settle the question whether the

expressed it, it is compatible with natural science that there corresponding properties are different. (And discovering how
should be such things as value attitudes but it is not compatible many fundamental physical properties there are is not discover-
with natural science that there should be such things as value ing something about concepts, but something about the world.)

properties. Value properties, Mackie claims, are 'ontologically The concept 'good' may not be synonymous with any physical-
queer' - i.e. they are funny mysterious properties in whose exis- istic concept (after all, moral-descriptive language and physical-
tence scientifically enlightened people should not continue to isticlanguage are extremely different 'versions', in Goodman's
believe. sense), but it does not follow that being good is not the same

Moore's argument that Good cannot be a physicalistic prop- property as being P, for some suitable physicalistic (or, better,

erty (a 'natural' property) was that if 'Good' is the same property functionalistic) P. In general, an ostensively learned term for a
as 'conducive to maximizing total utility' (or whatever natural
property (e.g. 'has high temperature') is not synonymous with a
property, physical or functional, you care to substitute), then theoretical definition of that property; it takes empirical and the-
oretical research, not linguistic analysis, to find out what tem-
(1) 'this action is not good even though it is conducive to perature some philosopher might suggest, what good-
is (and,
maximizing total utility' ness is), not on meanings.
just reflection
An idea which came into philosophy of language a few years
is a self-contradictory statement (not just a false one).
after I introduced the 'synthetic identity of properties', and
But even a Utilitarian would not claim (1) is self-contradic-
which enlarges and illuminates the point I have been making, is
tory. And this shows, Moore claims, that although being Good
Saul Kripke's idea of 'metaphysically necessary' truths which
and being conducive to maximizing total utility might be corre-
have to be learned empirically, 'epistemically contingent neces-
lated properties, they could not be the same identical property.
sary truths'. 3 Kripke's observation, applied to the
Moore's argument turns on assumptions that I and many
temperature/kinetic energy case, is that, if someone describes a
other philosophers of language would reject today, however.
logically possible world in which people have sensations of hot
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-
Kripke's Naming and Necessity, Harvard, 1980 (lectures originally given
2
Ethics, Inventing Right and Wrong, Penguin, 1977. in Princeton in 1970).
208 Values, facts and cognition Values, facts and cognition 209

and and objects that feel cold,

cold, there are objects that feel hot What is wrong with the argument, Ruth Anna Putnam points
and which these sensations of hot and cold are explained by
in out, is that many descriptive predicates naturally acquire an

a different mechanism than mean molecular kinetic energy, then emotive force. In our culture, 'slobbers his food all over his shirt'
we do not say that he has described a possible world in which has strong negative emotive force although the phrase is literally
temperature is not mean molecular energy. Rather we say that a description. Any word that stands for something people in a
he has described a world in which some mechanism other than culture value (or disvalue) will tend to acquire emotive force.
temperature makes certain objects feel hot and cold. Once we The word 'good', in its moral sense, is applied to many things.
have accepted the 'synthetic identity statement' that temperature Some of these - good states of mind, for example - may be nat-
is mean molecular kinetic energy (in the actual world), nothing urally valued: it may be part of the content of that very state of
counts as a possible world in which temperature is not mean mind that one values being in it. I don't mean to suggest the
molecular kinetic energy. converse: that any state of mind that is naturally valued in this
A statement which is world is tradition-
true in every possible sense is good; that would be clearly false. Suppose 'good' were
ally called 'necessary'. A property which something has in every defined so that things which are naturally valued and are such
possible world is traditionally called 'essential'. In this tradi- that there is no good reason to disvalue them (as there is reason
tional terminology, Kripke is saying that 'temperature is mean to disvalue certain drug- induced states which are naturally val-
molecular kinetic energy' is a necessary truth even though we ued) count as 'good'.
can't know it a priori. The statement is empirical but necessary. Then one would expectthe statement that something is good
Or, to say the same thing in different words, being mean molec- to have positive 'emotional force' because of the nature of the
ular kinetic energy is an essential property of temperature. We property. Even if one is not a Consequentialist (i.e. one who
have discovered the essence of temperature by empirical investi- thinks everything with sufficiently good consequences is good),
gation. These ideas of Kripke's have had widespread impact on there is no doubt that the most common reason for calling an
philosophy of language, metaphysics, and philosophy of mathe- action good is that it has good consequences, among which
matics; applied to Moore's argument they are devastating. might be that it ultimately promotes states or situations which
Moore argued from the fact that (1) can only be false contin- are naturally valued; again, the very nature of the property
gently, that being P (for some suitable natural property P) could explains why the description comes to have 'pro' emotive force.
not be an essential property of goodness; this is just what the Mackie defends his claim that goodness is ontologically
new theory of necessity blocks. All that one can validly infer 'queer' by introducing as a premiss the assumption that one can-
from the fact that (1) is not self-con tradictory is that 'good' is not not know that something is good without having a 'pro' attitude
synonymous with 'conducive to maximizing utility' (not towards that something. This amounts to assuming emotivism
synonymous with P, for any term P in the physicalistic version of in order to prove emotivism. The devils in hell are frequently

the world). From this nonsynonymy of words nothing follows depicted as using 'good' with a negative emotive force ('He has
about non-identity of properties. Nothing follows about the es- a deplorable tendency to moral goodness' one of them might
sence of goodness. say); contrary to Mackie, I do not find such uses to be linguisti-
Ruth Anna Putnam has pointed out that another common cally improper or to involve any contradiction. And do we not
argument that goodness cannot be a natural property does not hear people say, 'I know that's a bad thing to do but I don't
work. 4 This is the argument that 'X is good' has 'emotive force', care'? As Philippa Foot has pointed out, one can even rebuff
'expresses a pro-attitude', and so forth. appeals to morality by saying, 'I'm not out to be a good man.'

4
What the possibility of these utterances shows is that while there
'Remarks on Wittgenstein's Lecture on Ethics', in Haller et al (eds.),
Language, Logic, and Philosophy, Proceedings of the 4th Intern. isindeed a difference between the describing use of language and
Wittgenstein Symposium, Vienna, 1980. the prescribing use or the commending use, this difference in
210 Values, facts and cognition Values, facts and cognition 211

uses is not a simple function of vocabulary. 'Descriptive' words guiding predicates seem 'ontologically queer' to a committed
can be used to praise or blame ('He slobbers food all over his physicalist.

tie') and 'evaluative' words can be used to describe and explain.

In Chapter 2, 1 argued that reference itself should seem 'onto-
(Consider the following dialogue: 'John must have been an logically queer' to a committed physicalist. If only physicalistic

exceptionally good man to do such a thing.' 'No, he had never properties and relations really exist, then reference, to exist,

been a moral paragon, if anything the contrary; but he must must be a physicalistic relation -
but then the problem, as we
have had a capacity for self-sacrifice we never suspected.' Here saw, is an overabundance of 'candidates'. There are an infinite
moral language is being used in an explanatory function.) To number of admissible reference relations (and all are physicalis-

repeat Ruth Anna Putnam's point again, nothing about good- tic,or at least naturalistic, if we count set theory as part of the
ness not being a property follows from the fact that 'is good' is naturalistic version of the world). If one of these were the rela-

used to commend. tion of reference, then that fact would itself be an ultimate meta-
Professor Putnam points out that there is, nevertheless, some- physical fact of a very strange kind.
thing right about Mackie's argument. Some moral expressions What would make such a fact strange is that we have built a

undeniably do have a built-in towards action.

orientation certain neutrality, a certain mindlessness, into our very notion of

'Should', 'ought', 'right', and 'must' are prime examples of such Nature. Nature is supposed to have no interests, intentions, or
'action guiding' words. The 'is/ought' problem is not the same as point of view. Given that this is right, then how could one admis-

the 'fact/value' problem, as she points out. 'I am not out to do sible reference-relation be metaphysically singled-out?
what I should' sounds much odder than 'I am not out to be a It is same mindlessness of Nature that makes the action-
this

good man' (and 'I am not out to do what I must' sounds crazy). guiding predicates 'is right' and 'is a justified belief seem 'queer'.
Mackie points out that no physical property has a built-in If one physicalistic property ? were identical with moral light-

connection to action (or to approval of an action), and concludes ness or with epistemological justification, that would be
that 'being the right thing to do', etc., are 'ontologically queer'. 'queer' - queer for precisely the same reason that it would be

But (besides depending on the assumption that the physicalist 'queer' if reference were a physicalistic relation. It would be as if

version of the world is the One True Theory), this argument Nature itself had values, in the moral case, or referential inten-
proves too much. For some epistemic predicates (e.g. 'rationally tions, in the semantical case.

acceptable', 'justified belief) are also action-guiding (taking For this reason, I think that Moore was right (even if his argu-
'action' in a wide sense, so that accepting a statement counts as ments are not acceptable) in holding that 'good', 'right' (and also
an action). One can say, good thing to do' and 'There is
'X is a 'justified belief, 'refers', and 'true') are not identical with phys-

a good deal of evidence that and not be committed to doing

Y* icalistic properties and relations. What this shows is not that

(or prescribing) X or to accepting Y; but if one says 'X is the goodness, rightness, epistemic justification, reference, and truth
right action to perform in this situation', or 'Believing that Y is do not exist, but that monistic naturalism (or 'physicalism') is an
completely justified', then one is oriented to doing (or prescrib- inadequate philosophy.
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 'rational Nazi' again
the moral sense does.
If we now mimic Mackie's argument and conclude that there What troubled us earlier was that we did not see how to argue

is no such property as being justified, but only 'justification atti- that the hypothetical 'perfectly rational Nazi' had irrational

tudes', then we land ourselves in total relativism. Before going ends. Perhaps the problem is this: that we identified too simply

so catastrophically far, we should pause to see why action- the question of the rationality of the Nazi (as someone who has
212 Values, facts and cognition Values, facts and cognition 213

a world view or views) with the rationality of the Nazi's ends. If are familiar with and able to use the medieval notion of 'chiv-

there is no end 'in' the Nazi to which we can appeal, then it does alry'), still these (our present moral descriptive notions such as
seem odd to diagnose the situation by saying 'Karl has irrational 'considerate', 'compassionate', 'just', 'fair') will not be notions
goals.' Even if this is part of what we conclude in the end, surely that he employs in living his life: they will not really figure in his
the first thing we want to say is that Karl has monstrous goals, construction of the world.
not that he has irrational ones. Again, wish to emphasize that I am not saying that what is
I

But the question to look at, if we are going to discuss Karl's bad about being a Nazi is that it leads one to have warped and
rationality at all, is the irrationality of his beliefs and arguments, irrational beliefs. What is bad about being a Nazi is what it leads
not his goals. you to do. The Nazi is evil and he also has an irrational view of
Suppose, first, that Karl claims Nazi goals are morally right the world. These two facts about the Nazi are connected and
and good (as Nazis, in fact if not in philosophers' examples, gen- interrelated; but that does not mean the Nazi is evil primarily
erally did). Then, in fact, he will talk rubbish. He will assert all because he has an irrational view of the world in the sense that
kinds of false 'factual' propositions, e.g. that the democracies are the irrationality of his world view constitutes the evil. Neverthe-
run by a 'Jewish conspiracy'; and he will advance moral propo- less, there is a sense in which we may speak of goals being

if one is an 'Aryan', one has a duty to subjugate

sitions (e.g. that, rational or irrational here, it seems to me: goals which are such
non- Aryan races to the 'master race') for which he has no good that, if one accepts them and pursues them then one will either
arguments. The notion of a 'good argument' I am appealing to be led to offer crazy and false arguments for them (if one accepts
isinternal to ordinary moral discourse; but that is the appropri- the task of justifying them within our normal conceptual
ate notion, if the Nazi tries to justify himself within ordinary scheme), or else one will be led to adopt an alternative scheme
moral discourse. for representing ordinary moral-descriptive facts (e.g. that
Suppose, on the other hand, that the Nazi repudiates ordinary someone is compassionate) which is irrational, have a right to be
moral notions altogether (as our hypothetical Super-Benthamite called 'irrational goals'. There is a connection, after all, between
did). I argued that a culture which repudiated ordinary moral employing a rational conceptual scheme in describing and per-
notions, or substituted notions derived from a different ideology ceiving morally relevant facts and having certain general types
and moral outlook for them, would lose the ability to describe of goals as opposed to others.
ordinary interpersonal relations, social events and political 'But what if the Nazi gives no reason for being a Nazi except
events adequately and perspicuously by our present lights. Of "that's how I feel like acting"?' This is a natural question, but
course, if the different ideology and moral outlook are superior here surely the natural answer is also the right one: in such a
to our present moral system then this substitution may be good case the Nazi's conduct, besides being evil, would also be com-
and wise; but if the different ideology and moral outlook are pletely arbitrary. Notice that 'arbitrary' is one of the words I
bad, especially if they are warped and monstrous, then the result have been calling 'moral-descriptive', i.e. a word which can be
will be simply an inadequate, unperspicuous, repulsive represen- used, without change of denotation, to evaluate (in this case to
tation of interpersonal and social facts. Of course, 'inadequate, blame), to describe ('John quite arbitrarily decided to change
unperspicuous, repulsive' reflect value judgments; but I have jobs'), to explain (or to indicate that no explanation of a certain

argued that the choice of a conceptual scheme necessarily reflects kind can be given), etc. Indeed, when I just said that Karl's deci-

value judgments, and the choice of a conceptual scheme is what sion to be a Nazi (in the case described) would be completely
cognitive rationality is all about. arbitrary, I was primarily describing, not evaluating. Many
Even if the individual Nazi does not lose the ability to use our things I do are, quite literally, arbitrary - e.g. choosing one path
present moral descriptive vocabulary, even if he retains the old across the campus rather than another; but this does not mean
notions somewhere in his head (as some scholars, perhaps, still there is anything wrong about these actions. (The matters are
214 Values, facts and cognition Values, facts and cognition 215

simply too trivial.) Even if I do something important 'arbitrar- larly, the statement that preferring poetry to pushpin is a preju-
ily' -say, change jobs -if I don't have family responsibilities, dice is literally false.) It is being suggested that it is somehow
etc., this may simply be my right But if the action is one that ontologically legitimate to admit that there are such things as
requires justification, then performing it arbitrarily and with no satisfactions, but not ontologically legitimate to admit such
expose one to legitimate blame. Making a deci-
justification will things as enlarged sensibilities, enlarged repertoires of meaning
sion which adversely affects the lives of others (and perhaps and metaphor, modes of expression and self-realization, and so
one's own life) to a great extent with no justification, just as an on. The idea that values are not part of the Furniture of the
arbitrary and willful (another of those moral descriptive words!) World and the idea that 'value judgments' are expressions of
act, is a paradigmatic example of irrationality, and not just irra- 'prejudice' are two sides of the same coin.
tionality but perverseness. We have investigated the question whether 'value judgments'
can be rationally supported. We have seen that various negative
We started our discussion in Chapter 7 by looking at Ben- answers rest on dubious philosophical assumptions: that ration-
tham's claim that 'prejudice aside' the game of pushpin (an ality itself is only good for 'prediction', or only good for getting
ancient children's game similar to tiddlywinks) is just as good as 'consensus', or that there is only one Method for arriving at
'the arts and sciences of music and of poetry'. In Bentham's view truth (where, sometimes, the only criteria for 'truth' are said to
the only reason poetry is better than pushpin, ultimately, is the be prediction and consensus), or that value judgments have 'two
brute fact that poetry gives greater satisfaction than pushpin (or meaning components', or that value properties are 'ontologically
gives satisfaction to more people, or both). There are, basically, queer'. The position I have defended is that any choice of a con-
two things wrong with this view: one thing wrong is that 'satis- ceptual scheme presupposes values, and the choice of a scheme
cannot be an aim of any being
faction' (or 'self-interest') itself for describing ordinary interpersonal relations and social facts,
who does not have other aims. If I had no aim other than 'my not to mention thinking about one's own life plan, involves,
welfare', then my 'welfare' would be a meaningless notion, a among other things, one's moral values. One cannot choose a
point which goes back to Bishop Butler. More important, some scheme which simply 'copies' the facts, because no conceptual
satisfactions are betterand 'nobler' than others, and one can give scheme is a mere 'copy' of the world. The notion of truth itself
reasons why. Poetry and music give solace, they enlarge our sen- depends for its content on our standards of rational acceptabil-
sibilities, they provide important modes of self-expression to ity, and these in turn rest on and presuppose our values. Put
many people, including many of the most gifted people the schematically and too briefly, I am saying that theory of truth
human race has produced. Calling these reasons for valuing cer- presupposes theory of rationality which in turn presupposes our
tain satisfactions above others 'prejudice' is actually closely con- theory of the good.
nected with both the 'two components' theory and the idea that 'Theory of the good', however, is not only programmatic, but
value properties are 'ontologically queer'. Bentham is operating dependent upon assumptions about human nature, about
is itself

with the model of 'neutral facts' and arbitrary 'prejudices'. society, about the universe (including theological and metaphys-
Indeed, calling the preference for poetry a 'prejudice' is just Ben- ical assumptions). We have had to revise our theory of the good
tham's way of suggesting that the fact that poetry gives greater (such as it is) again and again as our knowledge has increased
satisfaction than pushpin is the only consideration that is not and our world-view has changed.
'arbitrary' incomparing the two; any preference for one kind of It has become clear that in the conception I am defending there
satisfaction over another (it is suggested) is arbitrary. But this is is no such thing as a 'foundation'. And at this point people
simply false, given the actual place in our conceptual scheme of become worried: are we not close to the view that there is no
the notion of an 'arbitrary' preference, and meaningless if 'arbi- difference between 'justified' and 'justified by our lights' (relativ-
trary' is wrenched out of the scheme to which it belongs. (Simi- ism) or even 'justified by my lights' (a species of solipsism)?
216 Values, facts and cognition

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 Appendix
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 Here is the Theorem referred to in Chapter 2.

so on, in the case of morality); but this is not at all to say that all Theorem Let L be a language with predicates F U F 2 , . . . JF fc

is entirely reasonable and well with the conceptions we now (not necessarily monadic). Let / be an interpretation, in the sense

have. We are not trapped in individual solipsistic hells, but of an assignment of an intension to every predicate of L. Then if
invited to engage in a truly human dialogue; one which com- J is non-trivial in the sense that at least one predicate has an

bines collectivity with individual responsibility. extension which is neither empty nor universal in at least one
Does this dialogue have an ideal terminus? Is there a true con- possible world, there exists a second interpretation / which dis-
ception of rationality, a true morality, even if all we ever have agrees with I, but which makes the same sentences true in every
are our conceptions of these? Here philosophers divide, like possible world as I does.
5
everyone else. Richard Rorty, in his Presidential Address to the Proof Let W W ly 2, . ,.be all the possible worlds, in some
.

American Philosophical Association, opted strongly for the view well-ordering, and let U be the set of possible individuals which
t

that there is only the dialogue; no ideal end can be posited or exist in the world W f
. Let Ry be the set which is the extension of
should be needed. But how does the assertion that 'there is only the predicate F in the possible world Wj according to I (if Fy is
t

the dialogue' differ from the self-refuting relativism we discussed non-monadic, then R y will be a set of w r tuples, where n is the {

in Chapter 5? The very fact that we speak of our different con- number of argument places of Ft). The structure (V^u (* = 1>2,
ceptions as different conceptions of rationality posits a Grenz- . .jk)) is the 'intended model' of L in the world W, relative to
.

begriff a limit-concept of die ideal truth. J (i.e. Uj is the universe of discourse of L in the world Wj, and
(for i = 1,2, . . . ,&) Ru is the extension of the predicate F { in W jt
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, If at least F u has an extension R uS which is
one predicate, say, ,

1979. neither empty nor j9

permutation P5 of Uj such
all of U select a
that Pj(R uj )?Ruj. Otherwise, let Pj be the identity. Since P } is
a permutation, the structure (Uj;Pj(Ru) (' = 1,2, . . . ,fc)> is iso-

morphic to (Ui;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 s). W
Let/ be the interpretation of L which assigns to the predicate
F (* = 1,2,
{ ,&) the following intension: the function fi{W)
. . .
218 Appendix

whose value any possible world

at s
W
is P/R^). In other words,

the extension of F in each t s

under W
the interpretation / is
defined to be Pj(R u ). Since (UtfARis) (i = 1A ...,*)> is a
model for the same set of sentences as (UyyRa (i = 1,2, k)) . .
. ,

(by the isomorphism), the same sentences are true in each possi-
ble world under/ as under /, and/ differs from J in every world
in which at least one predicate has a non-trivial extension, q.e.d.

Index
Comment: If, in a given world W jy there are two disjoint sets

which are extensions of predicates of L in 5

W
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 P 5 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
Althusser, L., 158-60 consciousness, 85-102; see mind-body
will ensure that under/ the extension of the first predicate - the anthropology {relativism in), 161-2 problem
one whose extension under I is the set of cats - is a set of dogs Apel, K. O., 175, 178 correlation (mind-body), 80-1; see
a priori (and identity theory), 82-4 mind— body problem
under / in Wj, or the extension of the second predicate -the
Aristotle, 57-8, 135, 148, 177 correspondence, 38-41; theory of
one whose extension under J is the set of dogs - is a set of cats truth, 51, 56-69, 72-4
under / in Wj. Bacon, Francis, 96
Baker, J., 105 Darwin, 109, 198
bats (sensations of), 92-3 Davidson, D., 116, 124
Second Comment: If there are objects - say, 'sensations' -
Bayes* theorem, 190-4 Da Vinci, Leonardo, 76
which one wishes not to be permuted, because one regards pred- Bayesian school, 189-95 De Finetti, B., 190
Bentham, 151-2, 168-73, 214, see Dennett, D., 28, 89, 91
icates of those objects as 'absolute* in some sense, one simply
super-Benthamites Descartes, 57, 75-7
stipulates that the permutations Pj are to be the identity on those Dewey, J., 162, 167-8
Benthamite psychology, 168-73
objects. This will have the effect of making the restriction of any Berkeley, 58-60, 64, 180 Diderot, 78
and Block, N., 78, 92 Divine Right of Kings, 156-7
predicate of L to those privileged objects the same under J
Boyd, R., 166 Dummett, M., 56-7
under / in each world. Boyle, 195-6
brains in a vat, 1-21, 130-5 Eccles, J., 91
Burks, A., 192 Einstein, 124
Third Comment: Since sentences receive logically equivalent
empiricism, 64-71, 124, 180-8
truth conditions under J and under/, it follows that on the stan- Carnap, R., 26, 88, 106, 112, 125, emotive force of ethical utterances,
dard 'possible worlds semantics', counterfactual conditionals are 136, 163, 181-4, 189 208-11
causal chain of the appropriate type ethics, 'inverted pyramid* picture,
also preserved.
(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
110
Cavell, S., 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
220 Index Index 221

extension, 18, 25-9; received view of Kant, x, 31, 56, 57, 60-4, 74, 83, 116, Montague, R., 26 and perspicuousness, 137; and
how fixed, 32ff; and non-standard 121, 128 Moore, G. E., 205-11 perception, 137-8; and interper-
interpretations, 29-35, 217-18; see Keynes, J. M., 191,205 Moses, 158 sonal situations, 138-9; and value
reference Kohler, W., 152 Murdoch, I., 139, 154, 167 terms, 138-41; see rationality
externalist view, 49ff Kolers, P., 68 rationality, 103-5, 163-4, 174-200;
Kripke, S., 46-7, 207-8 Nagel, T., 92
and reasonableness, 107; criterial
fact-value dichotomy, 127-49, Kuhn, T., ix, 38, 113-19, 126, 150 205-11 conceptions of, 105-13; and
naturalistic fallacy,
201-16; not to be drawn on basis of 46 philosophy, 113; ordinary language
natural kind terms, 22-5,
vocabulary, 138-9; and subjectivism Leibniz, 75 philosophers* view of, 110; scien-
Neurath, O., 106
about goodness, 141-7 Lenin, 124 163 tism and, 124-6; modern instru-
Newman, John (Cardinal), 136,
falsifiability, 194-8 Lewis, C. S., 147 Newton, 58, 73, 75, 76, 200, 216 mental notion, 168-9; of pig-men,
Feyerabend, P., ix, 113-19, 126 Locke, 47, 57, 61, 180 Nietzsche, 157, 216 171-2; modern versus ancient views,
Field, H., 45-6 logical positivism, see Carnap, Steven- non-realist semantics, 56; see inter-
173; and technological success,
Foot, P., 209 son, empiricism 175-6; and 'instrumentalism',
nalist view
Foucault, M., 121, 126, 155-62 178-80; and empiricism, 180-8; and
ix, Nozick, R., 36, 122, 164-6
Frege, G., 27, 124-5 Mach, 124 tradition, 203; and majority agree-

Freud, 157 Mackie, 206-11 ment, 177-8; and method fetishism,

J., Ockham's razor, 133
functionalism, 78-82; see mind—body majoritarianism, 177-8 188-200, 203; and inductive logic,
ontological queerness, 206-11
problem Malament, D., 90 189-94; and falsifiability, 194-8;
organic unity, 152
Marx, 157 and theory of the good, 215-16; and
Garfinkel, A., 119-20 Marxism (of Althusser), 158-60 solipsism, 215-16; as Grenzbegriff
parallelism (mind-body), 76-7 (limit conception), 216; see rational
Glymour, C., 90 meaning, 29; see extension, index, in-
Peirce, C. S., 30, 74, 198 acceptability
Goodman, N., 68-9, 74, 79, 98, 123, tension, interpretation, intentions,
phantasm, 57 rational Nazi, 168-71, 211-14
125, 146, 193-4 intentionality, notional world,
phenomenalism, 180-5, 187 realism, see correspondence theory of
Grice, P., 105 reference, truth, two-components
philosophical discussion (compared truth, externalist view, internalist
Griffin, D., 92 theory
with political), 164-6 view, Dummett, reference, qualia,
'The Meaning of "Meaning" ', 22-5
Plato, 75, 120-1, 124, 155 metaphysical realism, causal
Harre, R., 109 (summary of the theory)
Platonism, 69 realism, causal chain of the appro-
Hegel, xi, 158 mentalism, 79
Popper, K., 166, 181-4, 195-9 priate type
history, x, 155-8 metaphysically necessary truths, 46-7,
207-8
primary qualities, 57-61 reason, see rational acceptability, ra-
Hume, 107, 124, 180
Private Language Argument, 64-72; tionality
Husserl, E., 28 metaphysically unexplainable facts (on
further interpreted, 121-4 reductionism, 56-7
hydra-headed robot, 96-7 physicalist theory of reference), 46-8
projection, 141-7 reference, magical theories of, 3-5, 16,
metaphysical realism, 134, 143-7; see
properties (synthetic identity of), 84-5, 51; and use, 8-12, 17ff; Turing Test
identity theory, 77-9; functionalism, correspondence theory of truth, ex-
78-82; and synthetic identity of
207 for reference, 9; causal theories of,
ternalist view, internalist view,
Putnam, R. A., 208-10 14, 45-8, 51ff; not in the head,
properties, 84-5; and split brains, non-realist semantics, reference
85-92; and a priori, 82-4; and con- qualia 22-5; operational and theoretical
sciousness, 85-102 method fetishism, 188-200, 203 qualia, 75-102; subjective color, constraints on, 29-32; of natural
incommensurability, 113-19 Mill, John Stuart, 180, 189 79-81, 86-94; realism about, 85, kind terms, 22-5; and Self-

index (in semantics), 26 mind-body problem, 75-102; and 88-9, 94-6, 99-102 Identifying Objects, 51, 53-4; inter-
inductive logic, 125-6, 189-94 parallelism, 76-7; and interac- Quine, W., 30, 33, 35, 44-5, 82-3, nalistview of, 52; causal chain of
'instrumentalism', 178-80 tionism, 76-7; and identity theory, 113, 116, 124 the appropriate type and, 14, 51-4,
intension, 25-9; and meaning, 27; 77-9; and mentalism, 79; role of 65-6, 70; similitude, 70-1, 56ff; ex-
and non-standard rational acceptability, 103-26; logical ternalist view of, 49ff ; see extension,
and Sinn, 27; in- physics in, 75-6; functionalism,
terpretations, 29-35, 217-18 78-82; correlation, 80-1; and positivist conception of, 105-13; truth
intentionality, 2, 17ff synthetic identity of properties, 'anarchist* conception (Feyerabend), relativism, 54, 119-24, 151-62; in

intentions, 41-3 84-5; and split brains, 85-92; and 113-19; and relativism, 119-24; and anthropology, 161-62; false rel-
interactionism, 76-7 consciousness, 85-102; and subjec- inductive logic, 125; and scientism, ativism, 166; objective relativism,
internalist view, 49ff; and Kant, 60ff tive color, 79-81, 86-94; and 126; and science, 134; and empiri- 167-8; B. A. O. Williams' discus-
internal realism, see internalist view realism about qualia, 85, 88-9, cal world, 134; and optimal sion of, 168-73
interpretation, 29-35, 217-18 94-6, 99-102; and a priori, 82-4 speculative intelligence, 134; and relativityof perception, 59-60
intrinsic properties, 36-8 Monod, 109 'real world', 135; role of adequacy relevance, 201-3
J.,
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-
iimilitude (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