ESSAYS IN QUASI-REALISM

This page intentionally left blank

 ESSAYS IN
QUASI-REALISM

Simon Blackburn

New York Oxford

OXFORD UNIVERSITY PRESS
1993
 Oxford University Press
Oxford New York Toronto
Delhi Bombay Calcutta Madras Karachi
Kuala Lumpur Singapore Hong Kong Tokyo
Nairobi Dares Salaam Cape Town
Melbourne Auckland Madrid
and associated companies in
Berlin Ibadan

Copyright © 1993 by Oxford University Press

Published by Oxford University Press, Inc.
200 Madison Avenue, New York, New York 10016
Oxford is a registered trademark of Oxford University Press
All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any means,
electronic, mechanical, photocopying, recording, or otherwise,
without the prior permission of Oxford University Press

Library of Congress Cataloging-in-Publication Data

Blackburn, Simon.
Essays in quasi-realism / Simon Blackburn.
p. cm. Includes index.
ISBN 0-19-508041-6; ISBN 0-19-508224-9 (pbk)
1. Realism. I. Title.
B835..B35 1993 149'.2—dc20 92-37005

246897531
Printed in the United States of America
on acid-free paper
 Preface and Acknowledgements

The essays collected in this volume represent a large part of the work that I
have done over the years on the issues surrounding realism. Since my own
approach to this matter has been sufficiently different from that of others to
cause some comment, and since the articles appeared in anthologies as well as
orthodox journals, it seemed justifiable to collect them together. I have pref-
aced them with an Introduction, hoping to do something to justify my ap-
proach, and to locate it broadly among the very different philosophical atti-
tudes to these contested areas.
I am pleased to acknowledge the following places of first publication, and
to thank the respective editors and publishers for permission to reprint the
papers: 'Truth, Realism, and the Regulation of Theory' was first published in
Midwest Studies in Philosophy, Vol V, 1980. 'Knowledge, Truth, and Reliabil-
ity' was the Henrietta Hertz Lecture of the British Academy, 1984. 'Morals and
Modals' appeared in Fact, Science and Value, Essays in Honour of A.J. Ayer's
Language, Truth and Logic, edited by C. Wright and G. Macdonald (Oxford:
Blackwell, 1987). 'Opinions and Chances' appeared in Prospects for Pragma-
tism, ed. D.H. Mellor (Cambridge: Cambridge University Press, 1980).
'Hume and Thick Connexions' appeared in Philosophy and Phenomenologi-
cal Research (special half-centenary volume) 1990. 'Moral Realism' appeared
in Morality and Moral Reasoning, ed. John Casey (London: Methuen, 1973).
'Supervenience Revisited' appeared in Exercises in Analysis, ed. Ian Hacking
(Cambridge: Cambridge University Press, 1985). 'Errors and the Phenomenol-
ogy of Value' came out in Ethics and Objectivity, ed. T. Honderich (London:
Routledge & Kegan Paul, 1985). 'How to be and Ethical Anti-Realist' ap-
peared in Midwest Studies, Vol XII, 1988. 'Attitudes and Contents' appeared
in Ethics, 1988. 'Just Causes' was originally given at an Oberlin Colloquium,
and appeared in Philosophical Studies, 1990. 'The Individual Strikes Back'
appeared in Synthese 1985. 'Losing Your Mind: Physics, Identity, and Folk
Burglar Prevention' was part of a conference held at Greensboro, North
Carolina, and appeared in The Future of Folk Psychology, ed. John Green-
wood (New York: Cambridge University Press, 1991). 'Filling in Space' was
presented at the Analysis half-centenary conference in Cambridge, and ap-
peared in the supplementary volume of the journal, April 1990.
The articles are reprinted as they first appeared, with a few minor amend-
vi Preface and Acknowledgements

ments to style and grammar. They are organized not chronologically, but
thematically, and are divided into three main areas of metaphysics, ethics, and
mind and physics. The footnotes sometimes differ from those in the originals,
partly because of added cross-references to other essays in this collection. I
have resisted the temptation to add more footnotes to explain why I said what
I did, even in cases where I would prefer to have said something different. But
to some essays I have added an addendum, pushing things further in direc-
tions that now seem to me important, often in the light of particular reactions
from friends and critics.
I would particularly like to acknowledge my debt to the editors of volumes
and the organizers of talks and conferences whose invitations prompted me to
the painful business of writing. Among the many auditors and readers who
have made valuable comments I remember Paul Boghossian, Edward Craig,
Elizabeth Fricker, Allan Gibbard, Bob Hale, Paul Horwich, John Kenyon,
Robert Kraut, Saul Kripke, David Lewis, John McDowell, the late Ian
McFetridge, Philip Pettit, Huw Price, Geoff Sayre-McCord, Marianne Talbot,
Ralph Walker, Jonathan Westphal, and Crispin Wright.

Chapel Hill, North Carolina S. B.

December 1992
 Contents

Introduction, 3
I. Metaphysics
1. Truth, Realism, and the Regulation of Theory, 15
2. Knowledge, Truth, and Reliability, 35
3. Morals and Modals, 52
4. Opinions and Chances, 75
5. Hume and Thick Connexions, 94
II. Ethics
6. Moral Realism, 111
7. Supervenience Revisited, 130
8. Errors and the Phenomenology of Value, 149
9. How to Be an Ethical Anti-Realist, 166
10. Attitudes and Contents, 182
11. Just Causes, 198
III. Mind and Matter
12. The Individual Strikes Back, 213
13. Losing Your Mind: Physics, Identity, and Folk Burglar
Prevention, 229
14. Filling in Space, 255
Index, 259
This page intentionally left blank
ESSAYS IN QUASI-REALISM
This page intentionally left blank
 Introduction

In many of these essays the main protagonist is a figure I christened the 'quasi-
realist'. Two routes led to this persona. One is familiar to every student of
moral philosophy. There everyone learns of philosophers who take a 'non-
descriptive' or non-representational view of our commitments, seeing them
instead as serving some other function, such as expressing attitude, endorsing
prescriptions, or, in general, putting pressure on choice and action. Such a
view is thought of as 'anti-realist', and is easily contrasted with a realist, or
descriptive or representational, story that says that such commitments do
what they seem to do: describe what we take to be the ethical facts. But while
everybody learns of this contrast, nobody, it seemed to me, really knew how
to conduct the debate about such ideas (or about the inevitable second-order
question: whether the alleged point of contrast is really well drawn this way,
or is even intelligible). The images behind the opposing sides are powerful
enough, and find cavalier expression in thoughts like this: the realist thinks
there is truth and knowledge to be had, that ethical properties exist, that they
explain things, that they are independent of us, that they are objective. The
anti-realist, or here the expressivist, is presumed to deny these doctrines. This
is why such a view seemed to many philosophers to imply an 'error theory' of
everyday ethical thought, claiming that a substantial component of that
thought is indefensible. The reasoning is that everyday ethical thought em-
bodies a claim to truth, or to knowledge or objectivity, or to something which
'lies beyond' the opinions and sentiments that we endorse, and which those
opinions might distort or represent badly. However, if this claim to objectivity
is quite spurious, as it seems to be on the non-descriptive theory, then every-
day ethical thought involves a kind of self-deception or fraud. But is the claim
indeed spurious, on the non-descriptive story? Does the anti-realist indeed
have to accept the baggage forced upon him when the debate is conducted this
way?
The issue is a large one, for not only in ethics but in many spheres—law,
literary theory, history, even science—there exists the same kind of radical
threat. The threat is that once we see the disappearance of some favoured
conception of objectivity and truth, or once we see what these amount to, we
can no longer exercise judgement as before. A proper consciousness of the
activity of judgement would unmask and undermine the activity itself. But is it

3
4 Introduction

true that an ordinary practice in good standing, like ethics or literary criticism
or law or science, needs to depend upon some illusion about objectivity? The
question would hinge, I considered, on whether an anti-realist story can make
sense of several ideas: that truth is the aim of judgement; that our disciplines
make us better able to appreciate it, that it is, however, independent of us,
and that we are fallible in our grasp of it. My idea was to domesticate these
high-sounding thoughts. Brought down to earth, the question is whether the
anti-realist can make sense of thoughts like ' I would like to know whether
bullfighting is wrong', or ' I believe that bullfighting is wrong, but I might be
wrong about that', or 'Bullfighting would be wrong whatever I or anyone else
thought about it'—claims asserting our concern to get things right, our fallibil-
ity, and some independence of the ethical from what we actually feel. These
concerns and claims look metaphysical, and indeed many theorists are content
to define their meta-ethical theory in terms of them. But looked at another
way they are merely part of good ethical thought: someone incapable of them
would lack a becoming modesty, rather than a metaphysical insight. Can a
non-descriptive story make sense of them? That depends on what it can do
with so-called 'indirect contexts', or sentences in which an ethical sentence is
itself embedded, but not actually asserted. I dramatized the question of
whether a non-descriptive story can indeed understand the use we make of
such contexts by inventing the figure of the quasi-realist, or someone who
'starting from an anti-realist position finds himself progressively able to mimic
the thoughts and practices supposedly definitive of realism'. Quasi-realism is
not really another 'ism' in the sense of a position or an ideology in the same
space as realism or anti-realism; it represents more an attitude of exploration
of the reality of the boundaries that those 'isms' demand, and may issue in a
complication or modification of the debate, as old oppositions prove incapa-
ble of carrying its weight.
Thus far the idea is simple enough, and it would be natural to suppose that
any success the quasi-realist enjoys would amount to a defence of a non-
descriptive theory of ethics—a defence, at least, against the charge that it can
make no sense of central elements of our ethical thought and practice. It
would be a welcome barrier against having to be an error theorist. But the
matter quickly goes deeper than that. For if this is all that we conclude, then it
seems that the quasi-realist inhabits a familiar but highly suspect philosophical
world: one where we know what we mean by descriptive versus non-
descriptive theories, by objectivity, or by realism versus anti-realism. Not only
would many philosophers deny that we can make good sense of these opposi-
tions, but quasi-realism can itself offer them support. For if, from the non-
descriptive starting point, it is successful in capturing some propositional fea-
ture of our discourse, then that feature can no longer be used as a litmus test
such that if we allow it we are realists and if we do not then we are ex-
pressivists. If it can do this for all proposed features, then there is no point at
which our use of ethical language supports realism rather than expressivism
(or vice versa); if that is so, it is tempting to conclude that the debate is unreal,
since there is no methodology for conducting it.
 Introduction 5

One might then think that if the quasi-realist explains and justifies our
practice with words like 'truth', 'fact', 'independence', 'objectivity', 'knowl-
edge', and even 'description', then the position seems to have bitten its own
tail: it seems not so much a buttress to anti-realism as an instrument for
dismantling the entire debate. But this, I think, is premature. Suppose we
honour the first great projectivist by calling 'Humean Projection' the mecha-
nism whereby what starts life as a non-descriptive psychological state ends up
expressed, thought about, and considered in propositional form.1 Then there
is not only the interest of knowing how far Humean Projection gets us. There
is also a problem generated even if the mechanism gets us everywhere we
could want. If truth, knowledge, and the rest are a proper upshot of Humean
Projection, where is it legitimate to invoke that mechanism? Perhaps every-
where, drawing us to idealism, or nowhere, or just somewhere, such as the
theory of value or modality. The difficulty and importance of thinking about
this does not diminish if quasi-realism is successful, although that success will
dramatically reduce the number of actual arguments to hand when we think
about such issues.
It teaches us a great deal about representation and description to learn
that they are so cheap to purchase that even the Humean can have them,
along with truth, fact, knowledge, and the rest. This was in fact the position I
found myself drawn to by the other route into the area. This arose not from an
interest in value, but from realist versus anti-realist debates conducted on a
wider scale and apparently in a different key, in particular as they appeared in
the works of Hilary Putnam (around 1979) and Michael Dummett. Although
the forum was wider, and even global if we are asked to take one view or
another of all linguistic practice, it seemed to me that the same question of
how much is actually defensible on anti-realist grounds was being neglected.
Both Putnam and Dummett had, it seemed to me, proposed particular litmus
tests, and in each case it seemed fruitful to ask whether someone might think
of himself as an anti-realist, yet with a good conscience find himself involved
in the intellectual practices proposed as definitive of realism. In the case of
Dummett's proposal, this would have meant a defence, on anti-realist
grounds, of the logical right to unrestricted classical bivalence, and with Put-
nam at the time, a defence of a proper contrast between what we best believe
and what is true. I addressed each task in the first essay in this volume. Apart
from its direction the main innovation of that essay, I think, was a firmly
'operational' approach to the notion of truth. I was certain that the anti-realist
need not involve himself with the definitions of what it would be good to
believe in an ideal limit, as in Peirce or Dewey, in order to give due weight to
the independence of belief and truth, or to classical logical principles appar-
ently governing truth. This was the beginning of a dislike of the straightjacket
that demands for analysis, in this case analysis of the notion of truth, put on
1. I was interested to discover the same interest in Humean Projection in work by H. P.
Grice, published in The Conception of Value (Oxford: Oxford University Press, 1991), esp. pp.
146-51. Grice's construction of value delineates processes strikingly similar to those suggested by
my quasi-realist.
6 Introduction

philosophical enquiry. Throughout these essays I have found it easy to resist

the blunt demand for a 'theory' of truth, forcing me to come clean on whether
all commitments are true or false in the same sense. What conception of truth
we have is shown in the use we give the notion; that use is most interestingly
visible in our logic, and in our thoughts about such matters as the point of the
commitments, their independence, and our fallibility. If we can give a story
explaining or justifying those thoughts, then there is no last chapter about
truth waiting to be written. The same in my view is true of knowledge, and in
the second essay collected here I advocated what it would now be fashionable
to call a 'minimalist' theory of knowledge. One of its effects is that the issue of
realism versus anti-realism cannot usefully be framed as an issue of cog-
nitivism about a disputed area.2 Wherever there is truth and a discipline for
better or worse judgement, we can in principle talk about knowledge, and
such talk therefore falls as well into the quasi-realist net.
This second route into an interest in quasi-realism led me to sympathise
with the view that it can work to chip away venerable oppositions. Yet in the
local case of ethics I retain the other sympathy: that it actually helps to give
the verdict to an anti-realist direction of thought. This second view has un-
doubtedly exasperated bolder spirits who mined a global 'minimalism' from
these seams. Once we have the pure gold of Wittgenstein's philosophy of
thought and language, Humean Projection is supposed to be an outdated base
substitute: 'fool's Wittgenstein'. Wittgenstein is taken to have reconciled us to
an indefinite motley of different discourses or, as they used to be called,
'language games', each with its own coordinated, unpretentious version of
truth: moral truth, modal truth, arithmetical truth, and so on. There is no
external idiom in which to theorise about any of this, but only the occasional
need to remind ourselves that ethics is not arithmetic, modal truth is not the
description of middle-sized dry goods, and so on (although why we should
need reminding of this is not all that clear).
But my exploration is more thoroughly Wittgensteinian than this kind of
relaxed pluralism. Wittgenstein himself cannot be seen frolicking on the
streets of Paris. He is no minimalist. In all the areas that he considered in the
later philosophy he relies upon a contrast between descriptive and non-
descriptive uses of language. His main technique is to press a non-descriptive
function for paradigmatic assertions: philosophical assertions express gram-
matical rules; ethical assertions do not describe facts; mathematical assertions
characteristically lay down rules of description; assertions about oneself do
not describe, but have the function of avowals; and so on for religious lan-
guage, expressions of certainty, and indeed everything that he treated in any
detail at all.3 One reason for missing this is that Wittgenstein did appear to
hold a minimalist theory of truth (although even that simplifies the dialectical
2. A suggestion recently refurbished by Crispin Wright, Truth and Objectivity (Cambridge:
Harvard University Press, 1992), chapters 3-4.
3. I assembled the case for this interpretation in an article not included here: 'Wittgenstein's
Irrealism' in Wittgenstein: Eine Neuberwehrung, cds. J. Brandl and R. Haller (Vienna: Holdcr-
Richler-Temsky, 1990).
 Introduction 1

position in which he expresses such views). But relying on this fact ignores
that minimalism about truth is entirely compatible with holding a very un-
minimal theory of assertion, as is shown by the example, very familiar to
Wittgenstein, of F. P. Ramsey, who is both a minimalist about truth and a
quasi-realist about many categories of commitment. The point is that a thin
theory of truth can consort with a thick theory of judgement, ascribing a
variety of functional roles to the commitments that on the surface all get
expressed by equally well-behaved indicative sentences.
Authority apart, there is something strange about maintaining that all
questions are immanent, or, equivalently, that we should be minimalist about
every notion that would provide a concept in which to couch a metatheory of
any area. Consider, for instance, realism in David Lewis's sense, about possi-
ble worlds. In 'Morals and Modals' I begin to sketch a quasi-realist story
about the modal, and although in the first instance it seems set to work better
for operator treatments of modality than for possible-worlds treatments, I am
not convinced that there is any insuperable obstacle to grafting quantification
onto it. But put this aside, and suppose that the theory works. Then it seems
to me strange to interpret the result as showing that there was never an issue
about Lewis's realism. The quasi-realist will indeed be saying, with all the
sincerity that Lewis can give it, that there is a possible world in which there is
a talking donkey. He has his own story about why he can say that (and,
importantly, about why it is a useful thing to be saying). But does the fact that
he ends up saying it show that Lewis's realism is a friendly, minimal story, had
for free by anyone who makes modal distinctions? Hardly, I think. Nearly
everyone thought the position was incredible, and it would be strange to see it
as a notational variant on something perfectly credible, so that in the end the
quasi-realist's assent to modal judgement turns him into a Lewisian realist. I
sum up this message by insisting that in the philosophy of these things, it is not
what you end up saying, but how you get to say it, that defines your 'ism'.
What kind of evidence could there be for one story as opposed to another
about the function of judgement in an area? For a given local area, the answer
is plain: we garner evidence from its surroundings, or in other words from
other things we think, for example about description, explanation, and repre-
sentation of the world. Thus a realist must make us comfortable with a success-
ful activity of representing the area of reality he sees us as describing; he has
to make us comfortable too with the utility we find in describing it (both
aspects are highly visible when we think of a realist theory of modality).
Making us comfortable is not a task that is easily circumscribed, notably
because many realists are comfortable with things, such as advanced epistemo-
logical powers, that offend the rest of us. Equally, of course, anti-realism must
make good its explanations of the discourse and the role it plays in our lives,
while avoiding the view that it exists because it describes a genuine aspect of
reality.
Two features of this answer get some attention in these essays, but doubt-
less deserve more. First, even if it works for a local discourse, the wheels skid
more dramatically when we try to apply it to anything more global. We cannot
8 Introduction

mount a naturalistic explanation of science or common sense en bloc, for there

is nowhere else to stand. The prospect for framing a global debate, for exam-
ple on the issue of instrumentalism versus realism in science and common
sense, in these terms is not bright. In work still in progress I take up the issue
of how much of a loss this is. Second, there is a background problem: Which
aspects of reality are to be taken as genuine? What is a proper base from
which to conduct this theorising? Quasi-realism is often charged with 'scien-
tism' at this point, or in other words with confining genuine reality to an
ontology and a set of features delineated by some favoured fundamental
science, such as physics. A better basis, it is then suggested, will include much
more: Why should it not include the properties and features that, on the face
of it, we know about as we indulge the area in question—moral, aesthetic,
causal, and modal features, for example? Is not their exclusion just a scien-
tific, or possibly an old empiricist, prejudice?
Three of the essays collected here—'Just Causes', 'Losing Your Mind',
and 'Filling in Space'—explore the issue of when we can say that a property
exists, when we can invoke it in explanation, and what kind of basis explana-
tion has in commonsense, spatially extended nature. They illustrate my char-
acteristic liberality, for instance by allowing forms of explanation invoking
ethical properties, which will surprise both moral realists and their oppo-
nents, who frame the debate centrally in terms of the explanatory import of
those properties. Just as knowledge and truth can be earned, I claim, so can
explanation. Roughly: properties are the semantic shadows of predicates,
and a supervening judgement, whose use of predicates is protected by quasi-
realism, may be cited by way of explaining all kinds of things. Views like this
are of course popular in the philosophy of mind, where a similar structure
exists. But my liberality over explanation does not extend to all the purposes
for which explanation is required; in particular, when our project is to place
ethical or modal or probabilistic discourse, that is, to understand why we go
in for it and what its role is in our activities and lives, a simple explanation of
our judgements as responses to the facts is markedly inferior to the longer
story explaining their functional role, the reason we have them and give them
the importance we do. This inferiority, I argue, is particularly visible when
we think of supervenience itself: a phenomenon that requires the most care-
ful description and explanation, if it is not simply a device for concealing
problems.
I hope that these metaphysical essays exonerate me from the charge of
finding the basis for explanation obvious or simple. But I am not repentant
about finding it more obvious that a projective theory in a quasi-realist
armour is an attractive theory of value and modality, than that such-and-such
a selection of things and properties deserves a privileged explanatory priority.
It seems to me clear that the exercise of 'placing' our propensity for value, for
the normative, and for modality in a naturalistic world is philosophically
urgent whatever exactly we include in the natural. It may even be that for
different problems different bases in the natural can be accepted as given, so
 Introduction 9

that there is no one dimension of explanatory priority.4 For any problem there
may be one useful starting point, although there is no one useful starting point
for any problem.
As well as selecting some acceptable basis for explanation, a successful
projective plus quasi-realist story needs to start by identifying the cast of mind
voiced by the judgement under consideration, in order make out its initial
contrast with straightforward belief. We have to have a sense of what we are
doing in moralizing, other than describing the lie of the moral land, and
similarly for the other areas. Some critics have denied that this contrast can be
made out. Perhaps their weightiest point is that the cast of mind we voice is
inextricably linked to the propositional form in which we voice it—and discuss
it, learn it, ponder it and teach it. If we want to know what state of mind is
voiced by an ethical or modal remark, it is most natural to locate it as simple
belief, for instance that X is good, or Y impossible. To paraphrase William
James, it is as if we find the trail of the Fregean serpent over all. Again, my
reaction is to distinguish what we start with in the construction, and what we
finish with: however irreducible and culturally particular valuing something
has become, the Humean is still in business if its irreducibility and particular-
ity is visibly the upshot of a natural process of voicing and projecting the non-
representational states she starts with.
This irreducibility is also supposed to be consistent with an approach that
has gained popularity in the last decade. This exploits the analogy with secon-
dary qualities by giving a 'response dependent' or 'judgement dependent'
truth-condition for moral or modal commitments, equating them in some way
with remarks about us and our dispositions under some set of conditions. Such
a position advances a biconditional:

X is <o>= [some persons] are or would be disposed to [reaction]

under [circumstances]

Each bracketed term is capable of indefinite variation: the persons might be

me, or us, or some of us, or me or us or some of us as we actually are, or me or
us or some of us as we would be after a process of improvement or idealiza-
tion. The reaction might be one of stance or attitude or judgement or sensa-
tion, and in turn it might have to be identified by using the very term o,
acknowledging irreducibility—or it might not. The circumstances might be
usual, or normal, or idealized, or disjunctive. Finally the whole equation
might be presented as giving the meaning of the original, or giving its truth-
condition, or doing both if we equate these, or doing neither but still some-
how helping with understanding the role of the original judgement. So a
proper taxonomy of such positions would be a long business, and many objec-
tions to some versions will not apply to all.
Clearly, proponents of the approach have to navigate between the Scylla

4. This seems to have been Grice's view, op. cit., pp. 102-3.
10 Introduction

of getting the truth-conditions simply wrong and the Charybdis of getting

them right, but by importing into the right-hand side the kind of judgement
we were hoping to understand. An example of Scylla would be

X is good = we as we actually are would be disposed to desire X

when we see it.

since this forgets that we are often bad, and do not always approve of what we
desire, and often misapprehend things when we see them. A full rectification
is to hand, but has us falling into Charybdis:

X is good = perfect people would be disposed to value X under

whichever conditions it takes to understand X's value.

The reason this represents defeat is that it takes moralizing—the very activity
we are trying to understand—to make the right-hand-side judgement, so all
we have is an equivalence between a quick way of judging that something is
good and a lengthy way of doing the same. Obviously, perfect people in the
right conditions will value X = X is good, but this tells us nothing about what
it is to identify people as perfect, or an activity as that of valuing, or conditions
as right.
The response dependency theorist has to show us a useful course between
Scylla and Charybdis. Yet there is every reason to doubt that the straits have
to be navigated in the first place. There is no reason at all to expect that
judgement that is, on the surface, not about ourselves but rather about num-
bers or possibilities or rights or probabilities, is usefully identified at any level
as concealed judgement about ourselves. Of course we voice our own reac-
tions as we talk of these things, but we voice our own beliefs as we say
anything whatsoever, yet not all our sayings are about our own beliefs. 5 We do
not constantly turn our heads sideways to look at our own reactions. Appeal-
ing again to authority we could say that this approach is un-Wittgensteinian in
at least three ways. First, it seeks a kind of reduction, or a forcing of judge-
ments that have apparently quite different functions into the one surprising
mould of describing us (or me, or some of us, etc.). Second, it detects com-
plexity and choice at the bracketed points, supposing that the folk who go
about describing colours or possibilities or values have made extremely sub-
tle, and apparently pointless, decisions about, for example, whether to make
reference to ourselves rigid or not, when there is no evidence at all that such
decisions ever needed making or could have been made. (This offends against
the 'nothing is hidden' principle.) Third, it goes bullheaded at the issue in
terms of finding truth conditions, whereas from the point of view of these
papers, and from Wittgenstein's point of view equally, if you want to talk in
these terms, then the best thing to say about 'X'is in the cases considered is
5. This critical comparison is urged by Huw Price in his 'Two Paths to Pragmatism'
(Response-Dependent Concepts, cd. Peter Menzies, Research School of Social Sciences, Austra-
lian National University Publication, 1991), p. 49.
 Introduction 11

that its truth-condition is that X is —but this will not be the way to under-
stand matters.
Another, but less easily identified, reservation comes from philosophers
who want to oppose or at least soften the fact-value distinction, often in
favour of giving priority to the so-called 'thick' evaluative concepts. I have
views about that move that go beyond discussions in these papers.6 But it is at
this point that the famous 'rule following considerations' lurk. (It is a consider-
able irony of these debates that to some of us these provide a proper field for
the very kind of quasi-realistic investigation that, according to others, they
rule out.) A number of these essays gesture at this theme, but 'The Individual
Strikes Back' is a direct commentary on the sceptical paradox and its anti-
realist solution. Obviously, since quasi-realism has its home in the account it
gives of normativity, it is essential to see that it might well apply to the norms
governing meaning that get lost in Kripke's sceptical dialogue. Again, I be-
lieve that Wittgenstein's authority lies squarely behind this approach, con-
trary to the weight of commentators who seem to have been so baffled by Saul
Kripke's setting of the problem.
When I started writing essays on these themes, I had not been fully struck
by their sheer size. It seemed to me likely that fairly local debates, in the
theory of probability or ethics, could be tidied up by a quick application of old
ideas. But the introduction of the same projective theme in other areas
quickly beckoned, both because of its historical importance and because mo-
dality especially seems an almost irresistible projectivist playground. Yet mo-
dality is implicated in everything we understand about the occupancy of
space. The final essay in this collection confronts this problem, if only to say
that things are no better for everyone else. I hope it leaves the reader tanta-
lized, for I suppose if I have a meta-philosophical creed it is that there are too
many ideologies and too few works wondering about the foundations underly-
ing their differences.

6. I tackle the issue in 'Through Thick and Thin,' Proceedings of the Aristotelian Society,
Supplementary volume, 1992.
This page intentionally left blank
 I
METAPHYSICS
This page intentionally left blank
 1
Truth, Realism, and the Regulation of
Theory

I
In this essay I want to approach an area where our metaphysics and our
theory of knowledge are apt to become scrambled. Few would deny that a
general theory of what it is that marks a statement as true should have
implications for the theory of knowledge—of what is necessary, or sufficient,
to know such a statement. Equally a view of knowledge might carry with it a
picture of the kind of thing that makes true the statements said to be known.
My trouble is this: I begin to doubt whether familiar ways of characterizing
debates in the theory of truth—realism vs. instrumentalism, and so on—
actually succeed in marking out interesting areas of dispute. I doubt too
whether, if this is so, the theory of knowledge will provide a solution, so that
by looking at conflicting views of what knowledge is we might come to
understand genuine metaphysical oppositions. I shall pursue these doubts
through the figure of the 'quasi-realist,' a person who, starting from a recog-
nizably anti-realist position, finds himself progressively able to mimic the
intellectual practices supposedly definitive of realism. In effect, quasi-realism
is the program begun by Hume in his treatment of both causal and moral
belief. I take it that its success would be a measure of the difficulty of
defining a genuine debate between realism and its opponents. There are
certainly images and perhaps attitudes to our discourse within a field which
seem to be associated with realism, but I shall argue that unless these are
given some concrete employment they represent subjects not so much for
decision as for nostalgia. This concrete employment would, in effect, be a
practice that a quasi-realist cannot imitate.
Some of think of ourselves as realists about some things, but as more like
pragmatists, instrumentalists, idealists, or, in a word, anti-realists about oth-
ers. We are happy that the world contains certain kinds of states of affairs—
perhaps physical or phenomenal ones—but no more. We might not be happy
with additional realms of conditional or counterfactual or moral or mental or
semantic or social or mathematical states of affairs. If this is not where we
would draw the line, nevertheless, we might think, there is a line to be drawn;
or, even if we prefer not to draw a line, still, we think, we can understand the

15
16 Metaphysics

position of those who do. One possible response is to seek a reduction, so that
although there exist no further independent facts corresponding to truths in a
certain area, nevertheless there are familiar lesser states of affairs in which
their truth consists. Another response is to advocate that we abandon the area
of thought in question. It was pessimism about reduction, and about the
existence of a distinct area of fact, making truths about what people believe or
mean, which led Quine to denigrate psychological and semantic theory. 1 But
the interesting issue for quasi-realism arises if we remain anti-reductionist but
also are unwilling to abandon the way of thinking in question. There is no real
option to abandon conditional, moral, mathematical, etc., thought, even if we
become squeamish about the existence of distinct states of affairs correspond-
ing to our beliefs in these matters. We would like to continue behaving as
though there were facts, even if we feel the anti-realist pull.
But how are we to tell, from a society or a language or the conduct of an
intellectual discipline, or just from our own thought, when we are in the
presence of commitment to a realm of fact? One might try saying, with due
emphasis and seriousness, that one thinks that there is a fact of the matter
whether. . . . We are all used to debates, say, in the theory of morals, in which
one side asserts this and the other denies it, and they appear to themselves and
probably to us to have located an issue. But why? Two problems suggest
themselves. The first is that from the standpoint of the anti-realist the asser-
tion 'there really is a fact of the matter (often: an objective, independent, or
genuine, fact of the matter) whether . . .'is itself suspicious. For if his puzzle
arises from the thought that no kind of thing could possibly play the role of
(moral, conditional, etc.) states of affairs, then he ought to translate his doubt
into difficulty about what the realist means to be asserting when he claims that
there are such things. We face the familiar philosophical trouble that in failing
to imagine something we also fail to imagine what it would be to imagine it.
One may have a picture of what the realist means, but I shall argue below that
this is not enough.
The second problem is that it is not so clear that the anti-realist must take
issue over any sentence of the kind: 'there really is a fact . . .'. This is
because it may belong to a discipline to think such a thing, and it is no part of
the anti-realist's brief to tamper with the internal conduct of a discipline. The
point would be obvious enough if we accepted Ramsey's redundancy theory
of truth: serious assertion of some statement p transforms into 'it is true that
p' and 'it is really a fact that p' without any escalation of ontology or meta-
physics. But part of what is at stake in accepting Ramsey's theory is whether
there is a genuine issue over realism and anti-realism, and we cannot yet rely
on it. Nevertheless, the threat remains that the anti-realist blandly takes over
all the things the realist wanted to say, but retaining all the while his convic-
tion that he alone gives an unobjectionable, or ontologically pure, interpreta-
tion of them. Both these feelings can arise when reading apologies for objec-

1. W. Quinc, 'Reply to Chomsky', in Words and Objections, ed. D. Davidson and J.

Hintikka (Dordrecht: Reidel, 1969), p. 303.
 Truth, Realism, and the Regulation of Theory 17

tivity and fact in, say, moral philosophy. Often nothing is conveyed except
that their authors are very well-brought-up and serious people. In the face of
philosophical doubt, one cannot simply ladle out objectivity and facts.
Here is an example to illustrate both these points. I might embrace an anti-
realism about conditionals by thinking along these lines. Imagine a thin time
slice through the universe, capturing a momentary array of regions of space
manifesting properties, like a large three-dimensional photograph. Nothing in
it corresponds to a conditional fact. Imagine a succession of these: things
change, order appears, yet neither in a single slice nor in a succession of them
is there room for a conditional fact. But our universe is just such a succession.
So there are no conditional facts. Now, if we are persuaded by this Humean
picture (it is no part of my claim that we should be), we lose grip not only of
what a conditional fact could be but also of what somebody could mean to be
asserting, what point he could be making, when he counters that there are
indeed such things. The temptation is to rely upon simile: the realist thinks of
them as like a fixative or stabilizing glue somehow explaining order. But if
pictures and simile fail, as I shall argue they do, and if we cannot use them to
make sense of the realist's position, there is still nothing to stop us from using
conditionals in the practice of our thought, from asserting them, and, for all
we have yet shown, from emphasizing our commitment to them with notions
of truth and fact. We become quasi-realists in following out the discipline—
appearing in as many ways as possible like realists but believing that we have
saved our souls in the process.
Clearly, images of what the original realist versus anti-realist debate is
about do come to mind. The realist thinks of it as though the moral order is
there to be investigated like a piece of geography, that numbers are like
eternal objects spread over space, that the past and the future are laid out,
even now, in a state of deep-frozen existence, and so on. But the presence or
absence of such images cannot define an interesting issue. It must depend on
some other way of directly identifying bad realist (or anti-realist) practice, and
without such a way it is irrelevant. The only serious debate arises about the
propriety of using or succumbing to these images. But we do not know what it
is to use or succumb to them. We have seen the anti-realist fearing that the
realist gives an improper role to the realm of fact and believing that he alone
gives a defensible interpretation of his utterances by seeing them in the spirit
of a quasi-realist. But what is the role that is improper, and what are the
contrasting interpretations? The similes and images do nothing to answer this
question: in themselves, indeed, they are harmless, and until we can define a
harm that they do, to lack them is to miss nothing worse than a mildly poetic
sort of pleasure.
A direct description of the contentious role that a realist gives to his states
of affairs is therefore needed, unless the metadebate, about whether there
was ever a point at issue, is to give the verdict that there was not. Some
philosophers have so concluded. Yet it is not easy to believe that such a hardy
perennial has such infirm roots. I now turn to some proposals and try to relate
them to what we should want to think about knowledge.
18 Metaphysics

II

The debate would be on again if we could see that the quasi-realist can imitate
the realist so far but no further, or alternatively that the realist must balk at
certain intellectual practices of his anti-realist opponent. If we can identify a
practice or thought that we can agree to be available to a realist but that even
a quasi-realist must avoid, then we have identified a contentious role for the
conception of fact, and in sorting it out we will be joining in what is revealed
as a genuine debate. I shall discuss four such thoughts. The first is the thought
that a theory to which one is committed could, even if only as a bare possibil-
ity, be false. This suggests a conception of facts as 'transcending' theory, which
the pragmatists or idealist might dislike. Anti-realist schools are often sup-
posed to accept that truth is 'theory-relative.' The second is the thought that
reality must be determinate—that at any point and whatever the infirmities of
theory, there is a truth of fact of the matter whether. . . . This too can appear
to be a thought that a quasi-realist must deny himself. The third, which is
closely connected, is that an anti-realist can, or must, make a certain equation
that to the rest of us with our realist propensities appears invalid. The equa-
tion I shall try to identify is made by denying a distinction between a regula-
tive and a constitutive status for principles, in Kant's sense.2 The Kantian
distinction between maxims of procedure in our intellectual inquiries, on the
one hand, and truths known about the world, on the other, may appear
tenable only if we are real realists about the world. The particular way of
taking certain principles (particularly that of bivalence, I shall argue), which is
forced upon us if we deny the distinction, seems to be that adopted by Ram-
sey.3 Eventually I shall conclude, although tentatively, that a competent quasi-
realist can make as much sense of the distinction as the rest of us, so that here
there is no distinct intellectual practice separating the quasi-realist from us.
The fourth and currently the most discussed view is that realism is the best
way of explaining our scientific success, that the existence of facts explains the
way in which our knowledge expands and progresses: here an explanatory
role seems to carry with it an ontological commitment which, again, is surely
problematic to the quasi-realist. All these suggestions are current in the litera-
ture. Should we accept any of them? I shall start with the first.
Putnam says
What does show that one understands the notion of truth realistically is
one's acceptance of such statements as:
(A) Venus might not have carbon dioxide in its atmosphere even though it
follows from our theory that Venus has carbon dioxide in its atmosphere,
and
2. I. Kant, The Critique of Pure Reason, the Appendix to the 'Transcendental Dialectic.' The
importance of this part of Kant in this connection was pointed out to me by E. J. Craig.
3. F. P. Ramsey, The Foundations of Mathematics (London: Routledge & Kegan Paul, 1931),
especially 'Facts and Propositions', p. 153, and 'General Propositions and Causality', pp. 240 ff.
 Truth, Realism, and the Regulation of Theory 19

(B) A statement can be false even though it follows from our theory (or
from our theory plus the set of true observation sentences).4
But as he himself remarks, the modal facts (A) and (B) are themselves
commonsense facts about the world. They are consequences of our beliefs
about the ways in which we gather knowledge or (to beg no questions) form
opinion. In other words, when we describe to ourselves what makes us form
opinion, we will see that (A) and (B) are true. But we cannot yet rely on anti-
realist views about how we gather knowledge (or opinion) being different from
anybody else's. It seems possible then that the quasi-realist will intone (A) and
(B) but not regard them as denoting a sellout to the opposition. And if this is
permissible, then they fail to provide the necessary litmus test. To come to
understand whether it is permissible, let us consider a quasi-realist attempt on
ethics and then on statements made with the subjunctive conditional.
I take an emotivist starting point: we see the meaning of moral utterances
as essentially exhausted by their role in expressing the speaker's attitude. I
have argued elsewhere that a surprising degree of quasi-realism is consistent
with that view.5 For example, it need not be surprising that moral utterances
characteristically take an indicative form. Nor need we be worried by, for
example, their appearance in the subordinate clauses of hypotheticals. 'If it is
wrong to allow secondary picketing, then the government has been negligent'
need not be regarded as hypothesizing the existence of a state of affairs, and
we do not immediately declare ourselves to be realists by using it. An
emotivist can perfectly well describe the role of such a conditional: it ex-
presses a conviction that if one attitude is to be held then so must the other be
held; this conviction may itself be moral, but there is nothing to prevent an
emotivist from holding attitudes to the interrelations of attitudes and to the
relation of attitude and belief. In fact, he must do so, for to hold this last kind
of attitude is what it is to hold a moral standard, and these too gain expression
in conditionals with the moral sentence on a subordinate clause: 'if lying
causes harm, then it is wrong'. For this reason we should not accept Geach's
test for whether we hold an indicative sentence to have a truth-condition, if
this is itself a commitment to realism.6 It fails to give the quasi-realist a proper
run. But Putnam's test seems more compelling. What could an emotivist be
up to in thinking: 'perhaps there is no such thing as a right to strike, even
though it follows from my moral theory that there is'? Must he deny himself
such thoughts? Is he to say that truth is 'theory-relative'?
4. H. Putnam, Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978),
p. 34.
5. See essay 7, 8, 9 and chapters 5 and 6 of Spreading the Word (Oxford: Clarendon Press,
1984). In recent years authority has been added to the position by Huw Price in Facts and the
Function of Truth (Oxford: Blackwell, 1990), by Allan Gibbard in Wise Choices, Apt Feelings
(Cambridge, Mass.: Harvard University Press, 1990), and by Paul Grice, whose routine for
construction that he calls Humean Projection is identical with the line I endorse. Sec The Concep-
tion of Value (Oxford: Clarendon Press, 1991), esp. pp. 107-111.
6. P. Geach, 'Assertion', in Philosophical Review, 1965, pp. 449-65. Dummett expresses
reservations about the test in Frege (London: Duckworth, 1973), pp. 348-53.
20 Metaphysics

An analogue of Putnam's theory about how our opinion is gathered would

be a view about the genesis of moral attitude. Let us suppose, as seems
possible, that prominent in such a story we would find a capacity to sympa-
thize, and an imaginative capacity to put oneself in another person's place or
to see what it is like from his point of view. There would no doubt be other
ingredients, but these illustrate my point. Suppose too that one recognizes
that these capacities vary from person to person and time to time. Suppose
finally that one admires those in whom they are boldly and finely developed.
All these thoughts and attitudes are perfectly accessible to the emotivist. But
they seem to give him all that is needed for a concept of an improvement or a
deterioration in his own moral stance. He can go beyond saying 'I might
change' to saying 'I might improve'. For he can back up that possibility by
drawing the change as one which from his own standpoint he would admire.
But now he has the concept of a possible flaw or failure in his present stan-
dards. A simple example would be the worry that a certain attitude, or the
way in which one holds a certain attitude, is not so much the outcome of a
proper use of imagination and sympathy, which one admires, but is the out-
come, say, of various traditions or fears which one does not. Hume, of course,
has been here before us:
A man with a fever would not insist on his palate as able to decide concern-
ing flavours; nor would one affected with the jaundice pretend to give a
verdict with regard to colours. In each creature there is a sound and a
defective state: and the former alone can be supposed to afford us a true
standard of taste and sentiment. If, in the sound state of the organ, there be
an entire or a considerable uniformity of sentiment among men, we may
thence derive an idea of the perfect beauty. . . .7
So the quasi-realist can certainly possess the concept of an improved stand-
point from which some attitude of his appears inept, and this, I suggest, is all
that is needed to explain his adherence to the acceptance of the apparently
realist claim: 'I might be wrong'. I think this view is confirmed if we ask: could
one not work oneself into a state of doubting whether the capacities generat-
ing moral attitudes are themselves so very admirable? The answer is that one
could, but that then the natural thing to say is that morality is all bunk and
that there is no pressure toward objectivity for the quasi-realist to explain. For
this kind of reason we should be skeptical of John Mackie's claim that we can
detect a mistake about objectivity in our ordinary moral thought. 8 For at least
as far as the modal test goes we can find ourselves holding fast to emotivism
yet perfectly imitating the allegedly realist thought.
Another example comes from the suggestion, again from Mackie, that
counterfactual claims are best regarded as condensed arguments. Once more
we might advance the Putnam test: surely I ought to recognize that it may be
false that if Mary had come to the party she would have gotten drunk, even

7. D. Hume, Collected Essays, 'Of the Standard of Taste'.

8. J. Mackie, Ethics (London: Penguin Books, 1977), chapter 1, section 7. See also essay 6 in
this volume.
 Truth, Realism, and the Regulation of Theory 21

though I know that by far the best arguments, given what we know and the
supposition that she came to the party, support the view that she got drunk?
Again, the suggestion is that this makes sense only if we have a realistic
conception of the truth or falsity of a counterfactual. But again the defense
will be that anyone can and should recognize the partial nature of any given
evidential basis of argument, including our own present one about Mary; that,
once we recognize that, we can permit ourselves the concept of an improved
standpoint from which our present argument takes account of only part of the
facts; and that this gives us a sufficient grasp of the argument's fallibility to
enable us, quasi-realistically, to say that it might be true that if she had come
she would have stayed sober.
When Putnam talks of his modal criterion he admits that it is possible for
an anti-realist to oppose truth-within-a-theory against truth as that which
would be agreed upon in the limit of investigation. Peirce and Sellars are
famous for this conception of truth, and in the remarks I have just made I may
appear to be approaching it. But this is not yet certain. It may be that the
notion of an improvement is sufficient to interpret remarks to the effect that
my favorite theory may be wrong, but not itself sufficient to justify a notion of
the limit of investigation; if these things are each so, then the notion of a limit
cannot be necessary to interpret the fear that my favorite theory is wrong.
There are bad reasons for thinking that we must possess and justify the
concept of a limit toward which improved investigation must converge. Per-
haps we are misled into thinking that the notion of a limit to which improved
opinion must converge is necessary if we are to believe that truth can be found
or finally settled, rather than be forever fugitive. But the notion of a limit is
not sufficient to remove that worry; it is itself forever fugitive. And it is not
necessary. The possibility of an improvement is perfectly compatible with the
existence of particular certainties. We can understand the concept of an im-
provement even if it on occasion remains a bare or notional possibility that
such an improvement should result in modification of a particular component
of our views—an attitude toward textbook examples of evil or the belief in the
shape of the solar system, for instance. Truth can be achieved, and in many
places it has been. The second motive to define a limit may be that it alone
gives value or point to the search for knowledge. But, again, it is insufficient
for that purpose: it would be no particular source of pride to know that my
opinion was one on which all judges would converge in the long run, unless I
had already attached value to the processes that would lead them to do so. For
instance, it would be by itself no merit for a house to be one on which all
modern architectural design would converge. This is the element that is un-
attractive in Peirce's definition of truth: unless there is a value in the processes
leading to convergence, there is no merit in an opinion's being true in his
sense.
We then get two different cases. It may be that the virtue in the processes
can only be described as their tendency to act as mid wives to the truth; that
this is why we appreciate, say, simplicity, elegance, the drive to unified expla-
nation, and so on. Or it may be, as in the moral case, that we can see the
77 Metaphysics

virtue in increases of sensitivity, sympathy, freedom from prejudice, and so on

regardless of any belief that they are instrumental in advancing a cause of
truth. If the latter is true, then we can accept Nietzsche's words:
Can we remove the idea of a goal from the process and then affirm the
process in spite of this?—This would be the case if something were attained
at every moment within the processes.9
Or, in other words, the mere idea of improvement is after all sufficient to give
content to our concepts of self-doubt and of fallibility and error.
But is it sufficient? One might object: it is all very well to say that an
attitude M could appear inept to one with an improved use of faculties generat-
ing such attitudes (call the improved use F*). But in itself it is consistent with
the possibility of a further improved use of those faculties F**, which re-
instates M, and so on for evermore. It is only if one is entitled to a concept of
convergence toward a limit, in which M is either vindicated or not, that the
modal claim makes sense. In short, it is objected that we need more than the
concept of an improved use of the relevant faculties. Perhaps we need
the concept of a perfect or ultimate use of them, and this is something to which
the emotivist has established no right. But is this argument cogent?
It is the final conclusion that is at fault. Suppose we are wondering
whether a particular view is true. The argument may establish that we need
the belief in a possible improvement of opinion to a point at which it is
asserted to be true, or to be false, and beyond which any admirable change in
theory merely serves to confirm that verdict. But this is a belief that a quasi-
realist may defend. And it is not a belief that demands that we understand the
concept of a perfect or ultimate use of the relevant faculties. It is of course
true that the quasi-realist may take a pessimistic view of the ease of finding
such points, or of our security in believing that we have ever found them. But
so may a realist. Or he may optimistically regard a large core of opinion as
already in that state, and express this by saying that certainty, or even knowl-
edge, can be had. And like anybody else, he can take into account the empiri-
cal question of whether the history of a discipline shows an increasing core of
solid fact, rendered more or less immune to successive revolutions of theory,
or whether it shows little or no such tendency.10
Two issues remain to trouble us. One, which I discuss in section IV, is the
possible claim that realism is the best explanation of such convergence in
opinion as we do find. The other, which I discuss in section V, is the specter of
indeterminacy, or the possibility of equally rational or admirable processes
generating inconsistent or incommensurable opinion even in the face of all

9. F. W. Nietzsche, The Will to Power, § 55. I owe the reference to Gordon Beam.
10. This notion has subsequently been explored under the title of 'superassertibility' by
Crispin Wright, chapter 9 of Realism, Meaning and Truth (Oxford: Blackwell, 1987), pp. 295-
302. I should have made it clearer that the notion of opinion, and of method, that I am using here
is highly abstract. In any actual situation we might do as much as is possible to improve an actual
opinion without ending up at the t r u t h , because of contingent limitations on our position or
resources of investigation.
 Truth, Realism, and the Regulation of Theory 23

possible evidence. This is an obvious threat to the realist, who must react
either by shrinking the area of fact to a point below that at which the possibil-
ity arises or by allowing that truth transcends all possible ways of knowing
about it. But it is equally a threat to the anti-realist. For I would summarize
the results of this section by saying that they justify the quasi-realist in a
conception of truth as a regulative ideal, or a focus imaginarius upon which
the progress of opinion is sighted. A direct result of the indeterminacy theses
would be that we are not entitled to that notion. I try to cope with this threat
at the end of the essay.
The net result so far is that apart from raising these two problems there is
nothing in the modal claims to distinguish realism. The quasi-realist need not
wallow in the unattractive or incoherent idea of truth being 'relative to
theory', and if we are to find what marks him off, we must turn elsewhere.

III

The second proposal for a distinguishing test for real as opposed to quasi-
realism is the thought that there must be a fact of the matter whether p or
whether not-p. This takes us into the area defined by Dummett, in which the
law of bivalence, that every statement is true or false, is seen as something to
which the anti-realist cannot owe any allegiance. The question must be
whether, in pursuit of a theory, we might quite naturally find ourselves need-
ing to be governed by bivalence regardless of any attitude toward the exis-
tence of facts or states of affairs that the statements of our theory describe.
Dummett himself talks as though 'true' and 'false' were in some sense
derivative from a fundamental division of utterances into those which it is
right to make and those which it is wrong to make; acceptable and unaccept-
able; correct and incorrect. This alone prompts the thought: Why should not
such an apparently normative division be made and rigidly applied to every
possible statement, regardless of any theory of what it is, if anything, that
makes a statement come down on one side or the other? We are investigating
the credentials of realism as such a theory, but the purposes of theory may be
quite sufficient to drive us to think in terms of the universal application of
such a division. To take an example (I shall take more later): the discipline of
moralizing may be subject to the regulative constraint that everything is either
permissible or impermissible, and this may be reflected as a commitment to
bivalence, which an anti-realist would therefore wish. He will happily say 'it is
either true that you did wrong or false that you did wrong' as a prelude to
discussion without in the least regarding himself as subject to any unfortunate
lapse.
The lapse would, perhaps, be unfortunate if it involved the suspect notion
of completing an essentially uncompletable task. Thus, if we need an idea
corresponding to God seeing the whole of the natural number series, or seeing
a completed totality of argument surrounding a very indeterminate counter-
factual, or seeing the whole of physical fact before him, and if without that
24 Metaphysics

idea bivalence is groundless, then we might agree that it rests on suspect

imagery. And the most forceful examples where we want to query it concern
infinities. It is not nearly so tempting, for example, to express anti-realism
about other people's sensations in terms of doubts about bivalence applied to
them, because it is manifestly true of you that you are in pain or false of you
that you are. (Wittgenstein's Investigations, §352 seems to me to be counsel-
ing us not to deny this but to regard it as simply irrelevant, as not containing
within itself the key to some understanding of the meaning of pain ascriptions.
Nor, therefore, is it a sensible target for suspicion, either.) But as we saw in
the last section, it is not generally true that to use a concept of truth, even to
describe the result of infinite or indefinite or open-ended investigation, re-
quires the paradoxical idea of a completion of such an investigation. It re-
quires only the concept of a point beyond which no improved perspective
demands a rejection of a given previous opinion. And it may well be that
serious theorizing about an opinion requires faith in the existence of such a
point: to agree that the revolutions of theory may endlessly bring in and out of
favor a particular opinion is just to agree that the theory is worthless or an
insufficient foundation for interest in that opinion. It would be equivalent to
abandoning the area of discussion.
I do not intend these remarks as more than a preliminary orientation.
They certainly do not remove the necessity to investigate the detailed argu-
ments in the theory of meaning which Dummett uses to support his position.11
Here I can only record my scepticism about those arguments: they all seem to
me to trade upon an insufficiently analysed notion of what it is to manifest
understanding of a proposition or, to put it another way, neglect of the intellec-
tual powers of the audience to whom one is manifesting that understanding.
This systematic neglect enables Dummett and his followers to generate the
astonishing conclusion that one cannot communicate what one cannot be
observed to communicate—as though one had to manifest understanding to
the most wooden or passive of possible interpreters.12 Nevertheless, I hope
my remarks are sufficient to raise a general doubt about the use of bivalence
as a litmus test.
Perhaps its most obvious application comes when we have the conception
of a certain realm of discourse as grounded in another, but where the ground-
ing can fail to support either a proposition or its negation. The clearest exam-
ple arises in fiction, where remarks about, say, Hamlet are true in virtue of
what is written in or supported by Shakespeare's texts. There may exist noth-
ing at all in those texts to support the proposition that Hamlet had a baritone
voice, nor that he had not. It would then seem wrong to suppose that such a
proposition must be true or false, even if it might also be wrong ever to be sure
that it will not turn out to be, as more subtle readings of the text are made. But
11. My own exploration of some relevant themes is 'Manifesting Realism' in Midwest Studies
in Philosophy (Notre Dame, Ind.: University of Notre Dame Press, 1989).
12. M. Dummett, especially in 'The Philosophical Basis of Intuitionistic Logic', Truth and
Other Enigmas (London: Duckworth, 1978), p. 217.
 Truth, Realism, and the Regulation of Theory 25

bivalence does not seem to be of so much use in discriminating the ontological

commitment within an assertion of truth of falsity. For instance, we may be
convinced that in some cases there may be nothing to make a counterfactual
either true or false. But we may also think that when a counterfactual is true, a
realist account of what makes it true — a realm of conditional facts, or powers
or dispositions, or even nested possible worlds — is in order and makes a
genuine opposition to anti-realism. The question is whether the quasi-realist
should accept commitment to bivalence as something forced upon him by the
discipline or practice of theorizing, and see no sellout to realism in so doing.
Since this strategy is crucial to quasi-realism, I shall follow it through in
some detail, first as though a regulative-constitutive distinction can indeed be
defended, and then in the light of skepticism about it. Corresponding to that
distinction we can define two senses of commitment to bivalence in an area of
discourse. Such a commitment may be expressed as:

(P1 ) In all cases that arise we should say: either this assertion is true, or
it is false.

or as:

(P2) We should say: in all cases either an assertion is true, or it is false.

is regulative, in a sense in which (P2) is apparently not, because we

might see that it should be adopted as a maxim governing practice, even if we
do not accept (P2). To take a particularly simple but illustrative example, it
may be that the practical requirements on a judge demand that he should
obey (P1) where the area that the assertion concerns is within his jurisdiction
(say: 'this contract is valid'). He must eventually find for one party and against
the other, and the requirement regulates his attitude toward the possibilities.
This is perfectly consistent with believing that there are likely to exist incom-
pletenesses in the law, or areas of discretion in which truth would not attach to
just one verdict. It will be part of the judge's discipline to think of truth as
attaching to just one side, which it is his task to find, but it does not follow that
in his reflections on legal reality he must suppose that this assumption has any
foundation in the facts. The open-ended nature of the discipline may mean
that it is never in practice possible to categorize a case as one in which it was
neither true that the contract was valid, nor false that it was, in which case the
maxim has an imperative force equivalent to 'keep looking for reasons'; its
status is exactly like the drive to a complete and unitary explanation, which
Kant saw as governing science.
Using the distinction between (P1) and (P2) , we can thus give an account of
the pressure that leads Dummett to distinguish between rejecting the law of
bivalence, which is something he thinks we should do, and rejecting the
principle of tertium non datur, which is something he thinks we should not do.
In other words, he thinks we should reject (P2), but that we should not hold
26 Metaphysics

that there are propositions that are neither true nor false. This way of putting
it depends upon an intuitionist view of the quantifiers. But without that view,
we can see the attraction of a position that is not committed to (P2) in an area
but that is committed to practice in any case as though (P2) were true and that
on principle refuses to consider an independent third status for an assertion.
This position is just one that thinks that on principle we should not be satisfied
with less than one of the polar verdicts and that we cannot ever regard our-
selves as in a position to assert that something other than one of those is
correct. And that seems compatible with a general doubt about whether there
will always exist facts in virtue of which just one side is correct, and even with
doubt about whether there must be the points we have defined, beyond which
improvement always confirms one side or confirms the other.
Although it is a digression from my main theme, I should mention that this
makes it quite wrong to argue, as Professor Dworkin sometimes seems to do,
as though the practice of a bivalent or polar legal system implies confidence in
its completeness, in the sense of its being sufficiently grounded to provide a
right answer to each question.13 It is particularly obvious that the constraints
on legal decisions make the regulative principle necessary. The fact that their
practice is so regulated licenses no inference to a need to use a group's
political theory, nor for that matter to use a determinate moral reality, to fill
the gaps in their law that may exist after we consider all the other actions they
may have taken to create that law.
The quasi-realist seems, then, to be able to accept the regulative principle,
(P1), provided that the purposes of making judgement in the area in question
can be seen to demand polar verdicts. But at this point the plot suddenly
thickens. For given that this is so, given that his thought is to be governed on
each occasion by the application of bivalence, what is to prevent the quasi-
realist from voicing acceptance of (P2)? If it is on principle to be his practice to
think 'either this is true, or it is false', what is to stop him from thinking 'in all
cases truth falls on one side, and falsity on the other'? It is quite natural to
argue, as we have done, that something more is involved in accepting (P2), but
it might appear at this point that it is only a realist who should accept the
distinction. A principle may be constitutive as opposed to regulative only if
there is an area of fact whose constitution it purports to describe. We cannot
just accept this contrast; it is the very one whose credentials we are querying.
Thus we may think there is a distinction between accepting the maxim of
inquiring into nature as though every event has a cause, on the one hand, and
believing it to be the case that every event has a cause, on the other. But
doesn't this very distinction stamp us as realists about cause? Once one has a
view of a causal (or legal, conditional, moral, etc.) fact as a thing with a
distinct ontological standing, perhaps one has a notion of it simply failing to
exist in a given case, and this possibility would be there even if reason com-
mands us to comport ourselves as though there were a truth of the matter to
13. R. Dworkin, 'No Right Answer', in Law, Morality, and Society, cd. P. Hacker and J. Raz
(Oxford: Clarendon Press, 1977).
 Truth, Realism, and the Regulation of Theory 27

be found. But what can an anti-realist make of all this? Surely it would be
natural for him to take it that the commitment to the maxim is all that can gain
expression in the universal generalization ([P2], every event has a cause, etc.),
which can therefore be bought quite cheaply.
It is interesting in this connection that Kant himself, according to Bennett,
can be interpreted as sometimes betraying his own distinction.14 For although
his official position is that the antinomies arise only through mistaking a
regulative for a constitutive principle, he seems to have no certain view of the
distinction, nor of what it means to make this mistake. One view he may have
held, again according to Bennett, is that given the sort of general principles
that regulative principles are, there is simply no distinction between 'accept-
ing the advice which (one of them) embodies, and believing that it is true as a
matter of fact'. This is to equate acceptance of (P1 and (P2). If Kant did hold
this, then the diagnosis of the antinomies becomes epistemological. We are
supposed to be in danger of mistaking what it is that entitles us to accept one
of them: it is not an insight into reality but the 'speculative interest of reason'.
However, this is an unsatisfactory contrast for the reasons we have given.
There are indeed images associated with the difference between things we
take from the world and things we read into it. But we do not know what it is
like to employ or misemploy these contrasting images. Kant, it is true, must
have thought he could explain this, for only if we think that a principle is
constitutive, which on this account means that it represents knowledge of the
world, can we find ourselves involved in the antinomies. But since the steps in
this diagnosis are so unclear, if we want to find a distinct employment for the
different ways of taking a principle, such as bivalence, we cannot rely upon
Kant as having paved the way.
Ramsey is probably the best known anti-realist to have equated such pairs
as (P1) and (P2). I think we can also regard Dummett as seeing no significant
difference between them. It explains why if we start with doubts about (P2), as
he tends to do, they become doubts about the policy adopted in (P1, and
abandoning that policy prevents us from using bivalence in the course of a
logical proof and demands the intuitionist modification of logical practice.
The curiosity in that route is, as I have already remarked, that starting by
considering (P2) and finding doubts about its truth seems appropriate to real-
ists, who, thinking they have a distinct conception of the existence of states of
affairs corresponding to the truth or falsity of a judgement, also fear a point
beyond which the facts simply do not exist. But one would expect the anti-
realist to take the reverse direction and, as it were, defuse bivalence rather
than reject it. This point is supported if we favor the kind of explanations of
meaning that Dummett adopts. If we learn what truth is in, for example,
moral or legal contexts by observing and understanding the relevant practices
and if we are kept on a very tight rein in any attempt to read more into those
concepts than appears on the surface, then the concept of truth that emerges
is at least governed by the commitment to (P1). This is why Dummett has
14. J. Bennett, Kant's Dialectic (Cambridge: Cambridge University Press, 1974), esp. p. 276.
28 Metaphysics

always seemed vulnerable (as he himself recognizes) to the counter that classi-
cal mathematical practice is part of what does give a real meaning to, in
particular, the concept of an unsurveyable infinite totality. I do not think such
a move is met by the reply that it involves a 'holism' that threatens to make
any logical practice immune to criticism.15 In areas where the whole question
is what our understanding consists in and whether it is legitimate in any case,
classical logical practice may play a part in locating it, even if in other areas we
have a firm enough grasp of the truth conditions of various propositions for a
given practice of inference to be vulnerable to criticism. In any case, and even
if I am wrong about Dummett, it appears that hesitation about (P2) is more a
mark of the realist than of his alleged opponent.
If this is so, the prospects for a litmus test look bright, even if not in the
way that might have been anticipated. For the quasi-realist becomes an embar-
rassingly enthusiastic mimic of traditional realist sentiments, and in his very
zeal we might expect him to differ from a real realist. 'In every case of legal
[moral, counterfactual, etc.] dispute', we hear him say, 'there is a truth of the
matter falling on one and only one side'. The very fact that for him there is
nothing more to accepting (P2) than there is to accepting (P1) entitles him to
say such a thing; the fact that for us there is some doubt about (P2) in spite of
our defense of (P1 shows that we are different from him. This seems to be the
position if our quasi-realist accepts the regulative-constitutive equation, and,
as we have seen, it is plausible to believe that the only motive for refusing to
accept it is the belief in a real status for facts or states of affairs.
Let us keep with the legal example. Perhaps after all it was not fair to
lumber the quasi-realist with Ramsey's equation. Clearly the question is
whether the only motive for assenting to (P,) yet not to (P2) is a belief about
the status of legal fact which marks a realist. If it is not, then the quasi-realist
may be able to imitate whatever hesitancy about their relationship we our-
selves feel. In the legal case the hesitancy about (P2) arises from our theory of
grounding: the belief that a group may not have got up to a question such as
whether a particular kind of contract is valid, and that none of its lawmaking
activities or beliefs covers the ease or determines a verdict. Now there should
be nothing in that theory that offends an anti-realist. He can also see judg-
ments of legal fact as answering entirely to questions about what the group has
arranged (or to questions about what the group thinks is right). It would seem
strange then if he cannot hesitate over (P2) just as much as we do, since he
shares the theory of grounding which we express by saying that there can be
cases in which there is no fact of the matter, but which could be decided either
way with equal propriety. Now in spite of the attraction of Ramsey's view, I
think the quasi-realist can admit this, even if he knows that on principle we
must proceed as though a decision one way or the other were always to be
found. The escape is to remember the strategy used in the face of Putnam's
test. There we removed the apparently metaphysical implications of the
modal fact by seeing it as arising from a natural view of the nature of our
15. Dummett. Truth and Other Enigmas, pp. 218 ff'.
 Truth, Realism, and the Regulation of Theory 29

moral capacities. The parallel here is to remove the similar implications of the
thought that 'even if we must all proceed as though bivalence held, it is still
not true that on every legal issue there must be a right answer' by seeing it as a
natural reaction to the contingent and unproblematic belief that a society is
unlikely to have thought of every eventuality. As with Putnam's test, the
commitment supposed to identify realism can be seen as a reflection of a piece
of knowledge available to anyone, whatever his philosophy. If this is right,
then it is the regulative-constitutive equation that was at fault, but the legal
quasi-realist can comfortably conform to ordinary thought.
Although I think this suggests the right solution in the legal case, others
may be more complex. For often, as with the example of determinism or with
moral or counterfactual reasoning, the theory of grounding is not so easy to
sketch. I shall follow through the case of practical reasoning. Here there is no
relatively straightforward theory of grounding leading one to any particular
view on bivalence. Ramsey's equation seems to be more attractive. Given that
on principle we must, once we are considering the question, suppose that
either there is, for example, a right to strike or there is not, what else could we
be doubting if we ask whether on every moral issue truth falls on one side and
not the other? Must we not see this as a natural propositional reflection of our
principle of procedure? What difference could there be between obeying the
maxim and believing the generalization? The threat is that by seeing no differ-
ence the anti-realist purchases commitment to the generalization too cheaply,
so that his very complacency serves to distinguish him from someone with a
lively conception of moral fact and hence of its possible absence. But again
the matter can be, as it were, naturalized. The brief remarks I made about the
genesis of moral opinion suggest that there might exist persons in whom
capacities of imagination and sympathy are equally well developed but to
whom parts of our morality, or the structure of our practical reasoning, seem
as unattractive as theirs might appear to us. The concept of improvement that
we isolated gives us then no basis from which to claim that they are wrong and
we are right. The concept of truth that did allow the anti-realist to think
'perhaps I am wrong' does not allow him, in the face of such a possibility, any
confidence that they are wrong. (This is why the suspicion that a foreign
culture is an actual example of this naturally and properly leads us to think
that we are not entitled to interfere with it. All that is needed is the admirable
[moral] opinion that we have a right to interfere only if the people's sensibili-
ties are inferior to ours; lacking this justification, it is appropriate to live and
let live, or perhaps live and learn.) This is Hume's case where 'there is such
diversity in the internal frame or external situation as is entirely blameless on
both sides'.16
Thus, a reflection of proper awareness of this possibility might be: 'on
some moral issues, there may be no truth and no falsity of the matter'. Then
the quasi-realist can permit himself this, even while admitting that we will, in
practice, be committed to bivalence, and while admitting that there is never a
16. Hume, op. cit.
30 Metaphysics

point at which we can be sure that there is no improved perspective, from

which at most one side will appear to be in the right. As I showed above, this
is consistent with holding that sometimes there is certainty: we are entitled to
certainty that no improved perspective or equally admirable perspective could
much differ from us in our most fundamental attitudes. It seems then that a
quasi-realist can have just the right attitude to the general principle of biva-
lence, which he can agonize over with the best. His attitude does not betray
him in the way Ramsey's test suggested it should.
I shall not pursue the other example where Ramsey's equation or the test
of adherence to bivalence might seem to be promising methods for discriminat-
ing real from quasi-realism. There is no case in which they seem to me to
work. The devices I have gone through in the examples of law and morals
seem to me to be equally successful in other cases, such as those of counter-
factuals, conditionals, and cause. The strategy is always to see bivalence as a
natural commitment within a discipline, for more or less obvious reasons, and
to see the hesitation which it may be natural to feel about indeterminate cases
as available to anyone who reflects either on the theory of grounding of the
facts in question or on facts about the capacities enabling us to form an
opinion of them. So far, then, we have no distinctive range of thought or
argument that marks off either side in the original debate. We still face the
problem that the issue of whether there is a casus belli goes against both
realists and their opponents.

IV
The final suggestion differs from the other three in a fundamental respect.
Each of them seized upon a relatively formal aspect of our intellectual
practice—our assent to a modal claim, to a logical principle, to a form in
inference. This one seizes upon an attitude we may have toward our own
opinion or knowledge. Of course, our problem all along has been to deter-
mine whether a distinctive attitude really is associated with realism, and so far
we have nothing but images and similes to gesture toward it. And we know
that these are not enough. The purpose of the fourth suggestion is to put, in
place of these, a distinct explanatory view which, it is suggested, is available
only to a real realist.
The explanatory view that may be adopted by the real realist, but not by
the quasi-realist, is this:

(RR) It is because opinion is caused, perhaps indirectly, by the fact that

p, that it converges upon p.

A quasi-realist will either deny convergence or seek some other explanation

for it. Notice that he need not deny or worry over science's propensity to
depart as little as possible from immediately preceding theory or to seek to
incorporate as much as possible of that theory into new views. Such a propcn-
 Truth, Realism, and the Regulation of Theory 31

sity is quite naturally explicable in other ways and requires no mention of the
states of affairs making a theory true. It does not even imply that those who
are in this way conservative themselves believe in convergence.
Using (RR) to define a realist versus anti-realist debate accords with many
intuitions. A quasi-realist can mimic our formal practice with the concept of
truth or fact. But surely he cannot give the facts any role in explaining our
practice. To do so is to embrace their real distinct existence, or so it might
seem. Again, the most famous examples of plausible anti-realism can be seen
exactly as attempts to deny (RR) for the areas in question. Hume explains our
opinion of the existence of causal connections not by our exposure to the fact
of such connections but only by our exposure to regular successions of events.
His main point is not to propound an 'analysis' in terms of regular succession,
for he can consistently suppose the mind to inject an ingredient into its concep-
tion of causation, which is lacking from the mere idea of regular succession.
His point is that there is no explanatory role for any fact beyond that of
regular succession in accounting for our opinion; exposure to such succession,
given natural propensities in the mind, is sufficient. Moral realism is refuted in
the same way. Here, indeed, it is not just that explanation of our opinion and
practice can be had simply in terms of natural perceptions, given our desires
and needs; it is also that there are obstacles to any other explanation. A
distinct explanatory mechanism, starting with a distinct moral fact, would not
be reconcilable with the requirement that its output be logically supervenient
on other facts, known in other ways.17 This contrasts with all other explana-
tory mechanisms; our knowledge of color, for instance, is explained in ways
that leave it logically possible that everything else should remain as it is, but
color change.
In spite of these attractive examples, the test has one flaw, which suggests
that although (RR) may be something that a realist must say, it may be
possible to say it without being a realist. It would make a necessary but not
sufficient condition of realism. The problem arises if a test is made to deter-
mine whether our attitude is one of real realism or not, when the opinion or
theory that we are testing itself makes causal and explanatory claims. For
example, suppose that upon exposure to a given experimental result a physi-
cist comes to hold that it shows the decay of a radioactive atom: he holds that
it is the decay that causes the result. To deny that is simply to abandon the
physics. But now we have what Putnam calls internal realism as something
integral to the theoretical practice, and there is simply no issue of dissent from
(RR). There is no option to 'speak with the vulgar' by assenting to the theory,
but to 'think with the learned' by denying its own causal and explanatory
claims about the genesis of our opinion. If thinking with the learned is an
option, it is not identified by adherence to (RR), and we are back on the
depressing project of finding another way to determine whether there would
be such a thing as assenting to (RR) but with that special attitude that marks
one's assent as realistic or not.

17. See essay 5 for more on Hume, and essay 6 for more on the supervenience argument.
32 Metaphysics

I do not think (as Putnam seems to) that this makes (RR) metaphysically use-
less. It maintains its role as a necessary commitment of a realist. What it does
suggest is that it has no part to play in detecting a realistic attitude toward the
most interesting theories: those such as our view of the external world, or the
existence of the past or other minds, where to hold the theory is ipso facto to
hold a certain explanation of our opinions. Not all theories have this kind of in-
volvement with their own metatheory; first-order morals does not, for instance.
Yet taking (RR) as even a necessary commitment of a realist has one
unattractive consequence. This is that states of affairs are only allowed if they
play a part in causing opinion. Now, it seems at least possible that one would
wish to be a realist about mathematics yet deny that mathematical reality is a
cause of anything. Or one might wish to have the same attitude toward the
reality of both the future and the past yet believe that only past states of
affairs play a causal role in generating opinion. This is the obverse of the
preceding trouble: (RR) is too generous to states of affairs that enter causal
theories and too dismissive of the others. But it is, so far, our only prospect for
the required hurdle. We might indeed try to do without mention of causation.
We could offer the test of whether a theorist agrees that it is the existence of
the fact that p that explains convergence of opinion upon p, leaving it un-
settled whether the explanation is causal. May I say that it is the fact that two
plus two equals four that explains the convergence of mathematical opinion
upon the belief that it does? Or that it is the fact that people do have the right
to strike that explains convergence of moral opinion upon the belief that they
do? One imagines that the quasi-realist may take license to say such a thing; it
will not be clear why he should not, even if it seems a particularly bold
appropriation of realist vocabulary. What may be clearer is that a quasi-realist
must think that there is in principle a better or more illuminating explanation
of this convergence, if he remains true to his anti-realist starting point. A
recognizably Humean picture of causation may go so far as to say 'it is the fact
that A causes B that explains convergence of opinion that it does'. Once the
mind has 'spread itself on the world', it also regards itself as reading things off
the world it has projected.18 But a Humean also holds that there is a more
illuminating and economical account of the convergence. The suggestion
arises that a realist, as opposed to a quasi-realist, must hold that the best,
most illuminating, and economical explanation of our opinion that p must cite
the state of affairs that p. Yet put like this we lose touch with the original
instincts behind the positions. Might I not believe in the real, distinct exis-
tence of the external world, yet accept that the most illuminating and economi-
cal explanation of my believing in it mentions only the coherence of my
experience? Or believe in your mental states, yet accept that my opinions are
explained merely by exposure to your behavior? What counts as the best
explanation depends on our purposes in grading explanations. It is not obvi-
ous why realism must take a very strong stand on that.
18. This is why any reliance on the mere feeling of objectivity, to confirm realism about, e.g.,
values or obligations, is pointless. See also essay 7.
 Truth, Realism, and the Regulation of Theory 33

It seems, then, that the best use to make of convergence is not as a phenome-
non of which a realist has a superior explanation but rather as one in which he
alone has faith. If we use our faculties properly, and reality is determinate,
then surely opinion must converge. Equally, if it is decided that even in the
long run opinion need not converge, then to many people this is a powerful
argument—perhaps the powerful argument—for denying that there is a real-
ity which that opinion is purporting to describe. Hence the importance of
indeterminacy theses: it is pessimism about convergence even in the long run,
even in the face of reasoning admirably from all 'actual and possible
experience',—the pessimism that is expressed in indeterminacy theses—that
is the most potent enemy of belief in a determinate world of moral, or psycho-
logical, or semantic, or even physical states of affairs. The lemma is that
scepticism must be intolerable: truth cannot be transcendental, incapable of
being achieved by any extension of reason and experience; to imagine a truth
is to have some grasp of what it would be to know it. If this is accepted, then
the response to indeterminacy is to deny realism. Yet indeterminacy also
threatens the existence of the points that on any suggestion give the anti-
realist a decent surrogate for truth.
But the contemporary confidence placed in such indeterminacy theses is
quite improper. For such a thesis cannot be an argument against pursuing our
moral, psychological, semantic, or physical theory as best we can. And then
reason does not beg but commands us to treat a conflict of theories as a
sufficient proof that the truth has not yet become known; reason is not free to
order a pursuit of one systematic truth and at the same time tell us that many
such unities may be equally in conformity with the world; it cannot in such a
way 'run counter to its own vocation'.19 To theorize, to assert, at all is to
disbelieve indeterminacy theses. Nor is reason here at war with itself, telling
us to have confidence on the one hand in our correctness, or at least in our
progress toward correctness, and on the other hand in the diversity of equally
good opinion that could exist. For it is not observation and reason that tell a
theorist in a proper discipline that he is unlikely to be effectively pursuing the
truth. It is pessimism or loss of faith among philosophers in the value of
science or particular sciences; the loss of the will to believe. Indeterminacy
theses are a symptom of our disenchantments, not the products of our reason-
ing powers: here at least it is true that the owl of Minerva takes wing only with
the coming of the night. In saying this, of course, I do not deny that a history
composed mainly of worthless and persistent dispute may be a good reason
for abandoning a particular discipline or domain of inquiry. But neither phys-
ics, nor commonsense psychology, nor the science of interpretation and trans-
lation, nor even perhaps ethics, reveals such a history. And in a case that
does, induction from that history rather than the bare possibility declared in
19. Kant, Critique, A 651/B 679.
34 Metaphysics

an indeterminacy thesis would motivate us to reject the inquiry. The boot is on

the other foot: rejection of an inquiry gains expression in indeterminacy
theses. It is topsy-turvy to use fear of indeterminacy to motive a modification
of an otherwise rational theoretical practice.
At last, then, can we say that the quasi-realist achieves his goal? No talk of
knowledge or certainty, reason or truth, no belief in the convergence of opin-
ion; no proper contrasts between theory and reality nor between regulative
and constitutive seem beyond his grasp. The realist never intended to say
more than this, nor should the anti-realist have settled for less. It is true that
in getting to this point our sympathies were engaged mainly on behalf of the
anti-realist: I would be inclined to say that realism, in the disputed cases of
morals, conditionals, counterfactuals, or mathematics, can be worth defend-
ing only in an interpretation that makes it uncontroversial. Could this be
retorted upon anti-realism, from a standpoint of differing initial sympathies?
If this is so, philosophy indeed leaves everything as it is. Yet there may be
reason to sympathize more with the anti-realist. He has earned the concepts
associated with objectivity, while his opponent merely stole them; he has
founded our practices on known facts about human capacities, while his oppo-
nent invents more. Economy matters. And perhaps we feel less guilt about
what we have earned—we are less vulnerable to scepticism. Or could it be
that, after all, the old pictures and metaphors are important to the disputes, it
mattering to us not just that people talk, behave, and practise intellectually in
the same way that we do but also that they are haunted by the same ghosts?
 2
Knowledge, Truth, and Reliability

The philosophy of knowledge and truth is dominated by two metaphors: that of

a system of elements that correspond with particular facts in the world, and
that of a raft or boat of interconnected judgements, where no element corre-
sponds to anything external. The choice is between realism and a leaning
towards a correspondence theory of truth, on the one hand, and holism, verg-
ing towards idealism, with a coherence theory of truth, on the other: the
pyramid and the raft. 1 These images have complementary attractions, but
neither provides a solid, stable metaphor whereby we can understand how
truth connects us to the world and its facts. The first prompts the charge of
foundationalism, or of the myth that we can step outside our best beliefs to
estimate how well they correspond with the facts. And the second seems to
disconnect the web of belief from proper control by the world, so falling into
idealism or relativism. In spite of rearguard actions it is the second image that
dominates philosophy today.2 I shall have little to say to oppose that movement
of opinion. But if we follow it, does it leave our concept of knowledge where
we would like it to be, or does it demand, as some have maintained, abandon-
ing the concept as the remnant of a classical but outmoded self-image?
Suppose we come to sympathize with the image of the raft through
mistrust of 'the given', or through a Humean or Wittgensteinian naturalism.
These make it possible to argue that we must retreat to a coherence theory
of truth, and that this in turn gives a ready answer to scepticism. It brings
truth down to the natural earth. We learn, on this account, that knowledge
and truth are concepts that we use, in our world. Philosophies may have
falsely promised us a real correspondence, in which facts impinged upon us
unfiltered by our own concepts, or our own ways of classifying, or our own
perception of similarities. But by seeing how false that promise is, we learn
not to fall into scepticism when it is unfulfilled. We learn to feel comfortable
claiming truth and knowledge within our own terms, and not to respect an
alleged demand—which could be met only by an a priori argument—that
those terms be given any foundation beyond the fact that we find them
natural. I hesitate to ascribe this position to any one writer. Its characteristic
1. Ernest Sosa, The Raft and The Pyramid', Midwest Studies in Philosophy (Minneapolis:
University of Minnesota Press, 1980).
2. The image seems to me common to Quine, Goodman, Putnam, Rorty, and many others.

35
36 Metaphysics

combination of empirical (or internal) realism and transcendental idealism

(or conventionalism) is, I should have thought, almost orthodox, and writers
who oppose it are self-consciously fighting not just one author, but a whole
tide of thought.
But somehow it doesn't seem to work as it should. For it is at least equally
easy to feel that the combination destroys any right to regard ourselves as
knowing—really knowing—what the world is like. This is obvious if the tran-
scendental part—the part gestured at by thought experiments involving bent
classifications, Goodman's predicates, or what are generally called the 'rule-
following considerations'—issues in a kind of conventionalism. But it is nearly
as bad if it issues only in a kind of naturalism. The fear that nature, whatever
it may be, has grown as not so much with a mirror as with a veil or a distorting
lens is not easy to exorcise. So who is right: those who find a comfortable
answer to scepticism in the combination, or those who fear that it plays into
the hands of the sceptic? Or are we to suppose that part of the package is a
new, appropriate, concept of knowledge that itself supersedes any that per-
mits scepticism to remain a real challenge?
My object in this essay is to approach these problems using a natural,
everyday requirement of reliability. I shall start by placing that in the rela-
tively pedestrian context of the problems with the analysis of knowledge, on
which there have been so many recent assaults. The position I arrive at is a
version of 'reliabilism', but one that ought not to be opposed by theories of
knowledge that insist upon justification. The reason for this combination
emerges in due course. The position affords an argument against scepticism,
but it would be idle to pretend that it 'refutes' it: indeed, it offers a diagnosis
of the permanent appeal of sceptical thoughts. This diagnosis does not depend
upon a profound internal versus external reading of the sceptical concept of
knowledge. 3 It sees the sceptic, certainly, as introducing a new context of
inquiry, but it offers no straightforward way of dismissing that context either
as illegitimate, or as involving concepts that are new and different from any in
everyday use.
So I start with some observations about the concept—our concept—of
knowledge. These ought to help us to see just what that concept involves, and
therefore to understand how much survives the drift towards idealism.

The classical problem is to find the condition that adds to

p is true
and
x believes p
3. Barry Stroud, The Significance of Philosophical Scepticism (Oxford: Clarendon Press,
1984), especially chapters 3, 4, and 5. The distinction, of course, derives from Kant and Carnap.
 Knowledge, Truth, and Reliability 37

to give sufficient conditions for: x knows p. The standard suggestions include

refinements of the requirement that x be justified, refinements of the require-
ment that x be situated reliably with respect to the fact that p, and versions of
the requirement that x's belief be not defeasible (meaning that further evi-
dence ought merely to confirm the belief that p).4 The chase for more-
accurate versions of these conditions, and the rivalry that can develop be-
tween them, has been called Gettier's salt mine, and it can enslave us against
our will. So we might start by asking why we should need an extra condition in
the first place.
If the epistemic concepts earn an honest living, they must form a natural
intellectual kind. Even if some multipart analysis accurately matched our
judgements in difficult cases, we would still need to ask why we are interested
in just that set of conditions. (A similar question arises when we propose
complex psychological conditions for meaning, and in many other areas.) To
see ourselves and each other as knowing things is to be important. But how
can it be important to organize our lives around one complex of conditions
rather than another? We need a role for the epistemic concepts, and the role
that seems most natural is that of ranking and selecting titles to respect. We
have to pick up our beliefs about the world from our senses and from each
other. So we need a vocabulary to settle whether our sources are ones that
themselves properly indicate the truth. This is a natural need, and it gives us
the natural intellectual kind in which to place our epistemic verdicts.
So consider a subject who believes correctly that something is so. His
being right gives him one title to correctness. Why isn't this enough? Because
his position may not deserve respect as the kind of position from which one
may safely accept information. A subject may believe truly by exercising
defective propensities to form belief on occasions on which, by luck, he is
right, or he may believe truly by exercising proper propensities, but when it is
an admixture of luck, not those proper propensities, which is responsible for
his being right. Given this reply, the two concepts that are anathema to each
other are knowing and being in an unreliable, defective state, or using an
unreliable propensity to form belief (the close analogy, of course, is with the
agent who does the best thing by accident, but has not exercised virtue in
doing so). It is natural to detect two components in a subject's epistemic
virtue on an occasion. There is the amount of information at his disposal,
which may be more or less adequate, and there is what he makes of it, which
may involve more or less rationality, or more or less reliable propensities to
use information to deliver belief. These two components need not march in
step, of course. But for the moment the difficulties this could cause, and
indeed the difficulties of effecting the division in any accurate way, need not
concern us.
Let us say that someone in a certain state of information, and exhibiting
some disposition to form belief, also possesses a degree of soundness, or
4. K. Lehrer and T. Paxson, 'Knowledge: Undefeated Justified True Belief, in Essays on
Knowledge and Justification, ed. G. Pappas and M. Swain (Ithaca: Cornell University Press,
1978).
38 Metaphysics

solidity as a source of information. We can call this a degree of value to a

would-be information receiver, or IRV (information-receiver value). If this is
the normative dimension, as it were, in which to place knowledge, then we
would expect to be able to put the following principle down:

The Mirv/Pirv principle: If two subjects each believe truly that p, the one
cannot know, when the other does not, unless the former is in a position
with at least as much IRV as the latter.

Since the role of the epistemic concepts is to rank sources of information, then
if one source knows when another does not, it cannot be that the belief of the
knowing subject is unsafe in ways that give him less IRV than the subject who
does not know. This is a principle concerning belief. So there is a caveat to
enter in the use of this principle: we might call it the Matilda caveat. Matilda
'told such dreadful lies it made one Gasp and Stretch One's Eyes'. When she
eventually shouted that the house was on fire, 'they only answered "Little
Liar" '.5 Her effective IRV had disappeared with her credibility, but for all
that, she knew that the house was on fire. To use the Mirv/Pirv principle
properly, we must say that Matilda's report actually had IRV, because the
belief to which it gave voice was solid, even if rational hearers might have
doubted it. Ultimately, of course, we are to be concerned with problems of
our own reliability, and problems of insincerity in report, or of difficulties of
interpretation of our own langauge, do not arise.
The Mirv/Pirv principle comes initially as a constraint upon the missing
clause in the proposals for defining knowledge, and I suggest that it guides
many verdicts in contested cases. To give a simple example, consider the
subject who forms a true belief well enough, but who should have done
something else as well, albeit the extra thing would in fact have misled him (he
believes, rightly and reliably, that the president has been assassinated, but
others who did believe this have by now read the usually reliable morning
papers, which deny it . . .). If we are reluctant to describe him as the only
person who knows that the president has been assassinated, this is because
someone who has done the extra thing has done the kind of thing that makes
them a better source of information on this kind of issue.
The principle serves to rule out even powerful and plausible attempts to
analyse knowledge. More important, I suggest that it explains our unease with
these attempts: our sense that somewhere things are going to go wrong for
them. Consider, for example, the conditional analysis of Dretske and Nozick.6
This finds the missing clause in the two conditionals:

If p, then x believes that p

If -p, then it is not the case that x would believe that p.
5. Hilaire Belloc, Cautionary Tales for Children, Selected Cautionary Verses (London, Pen-
guin Books, 1964), p. 32.
6. F. Dretske, 'Conclusive Reason', Australasian Journal of Philosophy, 1971; R. Nozick,
Philosophical Explanations (Now York: Oxford University Press. 1981), chapter 3.
 Knowledge, Truth, and Reliability 39

The idea is that x's believing should be sensitive to the truth, so that x should
be what Nozick felicitously calls 'tracking' the truth. This idea is a good one:
sensitivity to truth is indeed the kind of solidity we are looking for. But its
realization in the two conditionals is not so good. For a little thought will show
that a person could satisfy them through possession of a defect, compared
with someone else who does not satisfy them, and that, for some audiences,
this defect could make him a worse informant on the kind of case in point. I
shall illustrate this by a case, but it is the principle that matters.
Two freshmen, Mirv and Pirv, see the Professor in a car. They each be-
lieve, truly, that the Professor is in his own car, and they are each good at
telling, in general, when propositions like this are true. Usually, for instance,
when the Professor is not in his own car he drives very insecurely, and each
freshman would judge that he was in an unfamiliar car. On this occasion,
however, the Professor might easily not have been in his car, which was due
for a service, and had he not been, the garage would have lent him a model of
the same type, which would thus have been familiar to him. The only differ-
ence is that the garage model has a sticker of Mickey Mouse on the back, but
the Professor wouldn't have minded that—indeed, he used to have such a
sticker himself, let us say. However, Pirv comes from a puritanical and
benighted part of the country, and could not bring himself to believe that
anyone as distinguished as a Professor would ever own a vehicle with such a
sticker. Mirv knows more about the world. But on this occasion his knowledge
stands him in bad stead, by Nozick's lights. For through it he fails to satisfy the
fourth condition: had the Professor not been in his own car, Mirv would have
continued to believe that he was. Whereas Pirv, through ignorance and misin-
formation, ends up satisfying the fourth condition: had the Professor not been
in his own car, Pirv would not have believed that he was. So Dretske-Nozick
would have us saying that Pirv knew the Professor to be in his own car,
whereas Mirv did not. This flouts the principle, for Mirv is a better source of
information about such things than Pirv. He is a better-tuned car-ownership
detector, using the right parts of a better system of belief about such matters.
To say that Mirv is more solid on this kind of issue raises the question,
noticed by Goldman,7 of how we should classify 'kinds' of proposition in order
to evaluate the reliability in informants (for reliability is inevitably reliability
in a kind of circumstance). And it raises the question of the antecedent
position of the receiver of information. Someone who knows much more
about a situation may rightly take information from a source who is generally
worse, or who on an occasion is behaving quite irrationally, just because he
knows that for particular reasons obtaining in this case, the irrationality is not
involved in the informant's situation.
Compare acting as a second to a careless rock-climber, who ties the belay-
ing knots in such a way that they might be safe, or they might not, and does
not check the difference. Was the weekend safe? God might have said so:

7. Alvin Goldman, 'What is Justified Belief, in Justification and Knowledge, ed. G. Pappas
(Dordrecht: Reidel, 1979).
40 Metaphysics

perhaps all the knots the leader tied that weekend were luckily secure. His
defect made no difference to your actual security. Swayed by this, we could
say that, ontologically (as it were), the weekend was really safe. But you
mightn't think so: there is a good sense in which you cannot ever be safe
behind such a person: you oughtn't to feel safe just because you don't know,
any more than the leader does, what he is actually doing (suppose his defect
only comes to light after the weekend is over: breaking out in a cold sweat,
you correctly say 'what a dreadful risk I took!').
The issues here are close to those that arise in any application of statistical
or dispositional facts to the singular case. Is it safe to bet on Fred surviving to
the age of eighty? He is a sedentary, bran-eating, slender academic . . . We
rapidly come to the narrowest class with weighty statistics, yet there is no end
to Fred's peculiar combination of properties. Suppose he does survive: it does
not follow that it was safe to bet on it. It may have been safe for God to bet on
it, just as it is safe for him to follow the unsafe leader, or safe for him to ignore
an exercise of irrationality on occasions when it is not in fact affecting the
truth of his belief. But it would not have been safe for us.
In the case of chance, we suppose that the weightier the reference class,
the better: we say that when we know more, so narrowing the kind in which to
put the single case, we have a better estimate, or are nearer to the 'true'
probability. This is easy to explain in pragmatic terms: someone using the
fuller information wins when betting with someone who can use only the
lesser. But because the standard epistemic position is not one in which
the receiver is the more knowledgeable party, we do not tailor the epistemic
verdict to cases in the same way. We are not, as it were, concerned with how
God might pick up information from a source: we are concerned with how we
might. Thus we take into account causal factors that render a source more or
less sound. But if flaws are involved then it is not the weightier position, which
happens to know that the flaws are not responsible for the truth of the infor-
mant's belief, that counts. There are cases in the literature which, in effect,
trade on this problem. Suppose, for instance, that Pirv is told by the President
of the Royal Society that the dark room he is about to enter contains a perfect
holographic illusion of a vase. Suppose that, irrationally, he takes no notice of
that information and believes because of a cursory glance that there is a vase
there anyhow. He doesn't know that there is a vase there, even if there was
(for the President was lying, deceived, or just failing to remember that the
machine was off). God, or anyone knowing this much more, could safely
accept Pirv's word that there is a vase there, because they know enough to
discount the exercise of irrationality in the way he came by his belief on this
occasion. But someone knowing no more and no less than Pirv does could not
accept his word. After the operations described, Pirv is not a solid source on
the matter for that person. Since the normal epistemic circumstance—the one
that makes channels of information important—is that of wanting to know
whether to accept information from a better-placed source, it is not these
ontologically superior positions that count. Pirv did not know that there was a
vase on this occasion, because there is a kind of thing he is doing—forming
 Knowledge, Truth, and Reliability 41

beliefs irrationally—and it is dangerous to accept beliefs when this kind of

thing is done. Someone knowing more can say that there is a narrower kind of
thing that Pirv is doing—forming beliefs irrationally when the facts are such
that the irrationality does not matter—and that it is safe to accept beliefs
when this is true. But this superior epistemic position does not dictate our
verdict: the concept of knowledge would lose its utility if it did. (There is a
telling temptation to go soft on this: if someone's stoutly maintained, but
irrational, belief turns out to be true, we sometimes let this success alone
dictate the epistemic verdict: 'funny how Beryl knew all along that Fred
was . . .'. The success makes us think that there must have been a reliable
kind of belief formation involved, even if it would have taken a superior being
to know what it was. Compare: 'so it was a safe bet after all', which is usually
said when it wasn't.)
Puzzle cases and disputes arise because there are different ways of classing
the kinds of case in which someone is reliable: there is the question of whether
the informant is in some normal causal relationship to the facts; whether he is
reliable over similar kinds of case; whether he would be justified or rational in
believing himself to be reliable; and finally whether his background beliefs
(which in turn may be rational or not) affect his standing as an informant. And
all these can come apart. If any of them fails, then there will be a way of
regarding the informant that makes him into a dangerous source of belief:
there will be a kind of case over which he does badly. But it would be optimis-
tic to expect principles to settle verdicts in such cases. Because of the 'holism'
of belief, there is no principled limit to the flaws that may result in our being in
kinds of states that are unreliable, and disqualify us from acceptability as a
source. In particular, we will always be vulnerable, if we try to isolate some
natural relation that a subject has to the fact that p, to cases in which he is
nevertheless playing Pirv to someone else's Mirv, although the other person
does not bear this particular relation to the fact. This explains the progress
found constantly in the literature: someone proposes a natural relation to the
facts sufficient for a subject to know that p, and someone comes along in a
generally better state (he has read the newspapers, etc.) but who through the
extra virtue, making him more solid on some kind of case, misses the title.

II

Solidity as an authority is a matter of degree. But knowledge, on the face of it,

is not. So how much solidity do we want: can knowledge tolerate chances of
being wrong, or even the bare possibility of being wrong? To put the question
in a closely related way, if a situation leaves it as much as barely or logically
possible that one is playing Pirv to a non-knowing Mirv, does that destroy
one's title? The most important initial division in the continuum of possible
improvements comes where a subject is sufficiently solid to be an authority,
and where any improvement in his state, or dispositions, would simply serve
to sustain his belief. We could relativize this, if the possibility of different
42 Metaphysics

recipients with different standpoints is worrying, and say that a recipient

should allow someone to know something only when anything that from his
standpoint counts as an improvement merely tends to confirm the original
belief. This suggestion is of course close to the familiar nondefeasibility condi-
tion on knowledge. It differs only in that I put the notion of an improvement
to the fore: it does not go without saying that increases in true belief, even
when reasonably used, count as improvements on a particular kind of case.
There also is a caution implied in putting the question as one of whether
improvements would sustain the verdict. It will not be to the point to go in for
thought experiments in which improvements that could overthrow the belief
could happen, but in the actual world wouldn't. This kind of stability is
sometimes easily achieved. Suppose I recognize my friend by a glance at his
face. I know who he is, not because weightier investigations could not be
made, but because they would simply confirm what I already know. Of
course, anyone whose position is as solid as this cannot play the role of Pirv to
someone else's Mirv: anyone in a genuinely improved position will also know.
If we used this as a cutoff point beyond which there is knowledge, we
would be importing what Armstrong called an 'external' element into the
notion of knowledge. 8 I could be in an informational state, and using disposi-
tions sufficiently well, yet not know something because, as a matter of fact,
the world does afford further evidence which would undermine proper confi-
dence in the belief. And I could be in the same state and using the same
intellectual dispositions, when, on the contrary, any improvement would con-
firm my belief; in that case, on this proposal, I know. People are uneasy with
this for several reasons: notably, it seems to cut the concept off from any
problems of objective justification, and it affords altogether too cheap a vic-
tory over scepticism (provided the way we are plugged into the world is all
right, then we know, and the sceptic cannot show that it is not all right).
These worries may lead us to divide the continuum higher up. At the
highest point, it is logically impossible that the subject should be playing Pirv
to another's Mirv. The gap between the subject's informational state and the
fact believed to obtain is to be closed altogether: it is to be logically impossible
that the state should exist, yet the fact not obtain. This exorcises all external
elements with a vengeance: it tries to ensure that there is no element of luck,
or even contingency, in the true believer's title to knowledge. Traditionally it
requires that we shrink the area of fact known, potentially down to an entirely
immediate given, just as the parallel motivation in the theory of ethics shrinks
the exercise of real virtue down from the chancy, external world where good
intentions can go wrong, to the safe realm of acts of will.
Alternatively, we might close the gap by expanding our conception of the
state the believer has got himself into. The states we get into, and because of
which we form beliefs, would be ones that we could not (logically) have been
in had there not been a spatially extended, temporally ordered world, contain-
8. D. M. Armstrong, Belief, Truth and Knowledge (Cambridge: Cambridge University Press,
1973), p. 157.
 Knowledge, Truth, and Reliability 43

ing the other minds, numbers, possibilities, values, and so forth, in which we
all believe.
What then is our best conception of the informational states whereby we
come to believe things, or to know them? Let us say that informational
states, in virtue of which we form beliefs, divide into two. There are those
which, as a matter of necessity, could not have existed had not the beliefs
formed in the light of them been true. We can call these guaranteeing states.
And there are those that do not meet this strong condition. Call these indica-
tive states. The question in front of us is whether only guaranteeing states
sustain knowledge, or whether indicative states, provided the external circum-
stances are right, can also do so. If we can happily see ourselves as largely
possessed of guaranteeing states, the looming problems of scepticism might
be thought to disappear. But can we? The 'informational state' in virtue of
which a system is disposed to absorb something new can include any part of
the deposit of previous times, as well as anything that could at all be thought
of in terms of the impact of the immediate environment. We think of our-
selves, of course, as getting into such states as a result of our physical posi-
tions and surroundings, the operations of our senses, and the use of con-
cepts, beliefs, and expectations which, be it because of reason (unlikely) or
nature (most likely) or convention (let us hope not), we find ourselves form-
ing. These banalities do nothing to support a 'guaranteeing' conception of
informational states. On the contrary, they conjure up painful images of the
ways in which the world responsible for our states might not conform to the
way we end up taking it to be.
A guaranteeing conception of our epistemic positions is given spurious
support by a spatial metaphor. (Kant charged that it was a mistake of Locke to
sensualize the understanding.9 I think it is at least as important a mistake, and
symptomatic of the same error, to spatialize it.) Thus we are often asked to
pronounce upon what is manifest, disclosed, given, embraced, internal to our
subjectivity, or accessible in our experience, or to settle issues of what we
really confront or access, or what we can penetrate to, or what is transparent
or open to us. The glassy blob of the mind reaches out to encompass (em-
brace, contain) facts, and knowledge stops at its boundaries. But then the
blob cannot stop short of embracing all kinds of states of the world, for if it did
it would be confined to embracing mental proxies of them, and these would so
intervene that it could never know the world, nor even understand a vocabu-
lary purporting to describe it. Whole issues, such as the realist/anti-realist
confrontation as it is framed by Dummett and his commentators, are impor-
tantly distorted (or sustained) by this spatial metaphor.10
9. I. Kant, The Critique of Pure Reason, A271/B327.
10. J. McDowell, 'Criteria, Defeasibility and Knowledge', Proceedings of the British Acad-
emy, 1982, pp. 455-79. The damage of the spatial metaphor is seen explicitly in McDowell's
argument against his opponents: he believes that if the mind does not embrace past states of
affairs, the sensations of others, and so on, then it must embrace only proxies of them 'interpos-
ing' between us and them (pp. 472-74), giving rise to insuperable problems of understanding and
knowledge. In the theory of thought this is the analogy of the position that we either see physical
44 Metaphysics

To escape this error we might query whether the very notion of a 'state'
plays us false here. Because of the interpenetration of theory and experience,
and because of the temporal growth of the system of belief, it can seem
artificial to analyze a response to new experience by thinking of informational
states at all. Certainly, if we do, there is little better to say about them than
that our state is one of being in an external world, surrounded by other minds,
possessing a long past, and so on. The promise of a quick victory over the
sceptic appears again, for our basic characterization of ourselves still entails
that we live in the kind of world that he finds it possible to doubt. Unfortu-
nately, we cannot retreat into dogmatism so comfortably. It will always seem a
fragile response to scepticism to refuse to set the problem up in the first
place—little better than announcing that the mind embraces the relevant facts
after all. So perhaps the best thing is to get away from the spatial image as
radically as possible, and this includes avoiding the protean notion of an
informational state. Although I sympathize with this, the notion of a state
does not have to be taken spatially, and there is no better general term to sum
up the fact that at given times we are in positions (states) in which we form
beliefs, and that the ways in which we do this, and their strengths and weak-
nesses, merit investigation.
The lowest place, as it were, at which we could claim knowledge was
where our state was what I called 'sufficiently' authoritative—meaning that it
made a reliable source on the kind of matter in question—and where it was
actually stable. This is a possible resting point: it depends, I think, on whether
we read the condition as strong enough to mean that there is no chance, or
virtually no chance, or only a chance that can be dismissed, of our being
wrong. Read without that understanding, the condition that belief be true,
authoritative, and stable would be far too weak for knowledge. For it could
coexist with a good chance of being wrong. But where we have a good chance
of being wrong, we are in a kind of state that makes us unsafe sources of
information. So we must read the condition so that it excludes any significant
chance of being wrong. And this is the point on which I want to focus in what
follows.
If we said that knowledge can exist provided there is no significant chance,
or real chance, of error, then we can defend a title to knowledge in the face of
an open, acknowledged possibility that the world might not be as we have
come to take it to be. The sceptic is apt to complain that when this is all we
have, then for all we know, things are not as we take them to be. But this is
wrong, for the whole issue is whether on the contrary we can know something
through being in a state that is indicative, although not guaranteeing, pro-
vided the external condition is satisfied: that we are authoritative, and the

objects 'directly', or we see proxies of them—sense data. Austin attacks this dichotomy at the
beginning of Sense and Sensibility: 'In philosophy it is often a good policy, when one member of a
putative pair falls under suspicion, to view the more innocuous seeming party suspiciously as well"
(p. 4). Dummett's 'challenge' to realists, to explain how things that are not 'manifest' can be
understood, which McDowell meets by the strategy of making more and more of the world
'manifest', seems to me to be much better met by entirely refusing the terms of discussion.
 Knowledge, Truth, and Reliability 45

state is stable. My suggestion is that the sceptic gets away with this bare citing
of possibility because of the normal implicature that we cite a possibility only
if we also give it some chance of being realized. It is normally only to the point
to cite possibilities that are 'relevant', and this is exactly what relevance is. So
it can seem that mere possibility left open defeats knowledge, whereas in fact
it may be that it doesn't, but that only real chance of error does. The exter-
nalist has it that we know because we are right, and because any improvement
in our position would just confirm that we are, and because we exercise
sufficient soundness to be a proper source of information on such a matter.
Once this is so, the sceptical possibility can, as we naturally say, be ignored. It
is the relation of this position to scepticism which I now wish to expose. For I
hold that, although it may seem to cheapen knowledge, in fact it does consider-
able justice to sceptical doubt: it offers an explanation of the deep roots of
those doubts, and it may enable us to place them even within the context of a
general sympathy with 'anti-realism' or 'internalism'.

III

Reliabilists and justificationists think of themselves as forming two different

camps. Now, one element in the view I have been defending supports each of
them. Reliabilists appear right, in so far as the soundness we require of
informants need not imply any self-consciousness on their part. They could be
like good instruments, and be deemed to know things just by being rightly
tuned to the truth. But although an informant need not have views about his
own reliability, we need to do so. It is always a weakness to have no account of
why an informant should be thought to be yielding the truth. It generates a
bad kind of state to be in. And when our own title is in doubt, externalism
does not help us unless we can properly see ourselves as reliable. To put the
matter in terms of section I, when we are unable to see ourselves as forming
belief reliably but nevertheless form it all the same, we are doing a kind of
thing that destroys our title as authorities. We cannot suppose that the mere
fact of our being right removes this taint, any more than Pirv in the Royal
Society case escapes the charge of irrationality or gains the title of knowledge,
just because on that occasion his belief was true.
The power of scepticism is quite underrated if it is seen as merely a forlorn
attempt to shake confidence by invoking possibilities that can normally be
ignored. Its real power comes with the absence of any sense of our own
reliability. Crucially, we have a sense of there being a large number of possible
worlds that appear as ours has done, but that contain scientific realities unlike
ours, skew distributions of other minds, large elements of counterinductive
truth, and so on. We might try to say, blankly, that we know that these
possibilities are not realized. But can we regard ourselves as reliable on just
this kind of point? How could I have a better than chance propensity to tell
when sceptical possibilities are realized? I can do nothing more than rehearse
the very considerations governing belief; if they leave open a space of possibili-
46 Metaphysics

ties, then there is nothing more to say about which possibility is realized, and
nothing more than chance to determine whether I am right. Suppose that
there is, for example, only one kind of world in which other minds distribute
as I naturally take them to do, but many where they distribute in other,
partial, ways (no other minds, ones attaching only to . . . , etc.). Suppose that
there is only one kind of world that is well behaved with respect to my
inductive regularities, but many that deviate in their different ways. If evi-
dence leaves the possibility of such mavericks, how can I be better than
chance at telling when they are realized?
I think it is wrong, or at least misleading, to suggest, as Barry Stroud does,
that scepticism here involves taking an 'external' view of our knowledge, as
opposed to an 'internal' one in which such questions do not arise.11 At least,
this is wrong if it leaves open a ready way of suggesting that the external
standpoint is optional, or even that it makes no sense. And the usual meta-
phor of externality is dangerous in this respect. All that really happens is that
normal ('internal') assessments of knowledge go on against a background of
assumptions of general reliability—the generally truth-yielding nature of our
procedures. But the same demand that there is no chance of error can be
made of the procedures, even if it is only in philosophical moments that we
think of raising it. Thus when Stroud diagnoses Moore as unable to hear the
philosophical sceptic's question in the intended way, it is unnecessary, on my
view, to suppose that there is a special, transcendental inquiry or context that
Moore cannot enter. Rather, there is a univocal query—about the chance of
being wrong—that is normally answered against a background of common-
sense theory (and is so answered by Moore), but that can equally be raised
about the procedures used in creating and sustaining that theory, or the princi-
ples upon which it seems to depend (induction, trust in the senses, etc.). I
suggest that this better explains Moore's peculiarity, which is that he seems
blind to the point at which any grasp of our own reliability fails us. It also gives
us reason to be cautious with the metaphor of externality, for it is not as if the
philosophical undertaking demands quite different tools or perspectives from
the everyday assessments of chance: it just has a different topic. Similarly, our
everyday financial standing may be settled by considering the credit we have
at the bank; this does not rule out a sensible query about the financial stand-
ing of the bank itself.
To show the query to be improper, in the philosophical case, we would
need to show that the relevant notion of 'chance' is inapplicable, when we
consider what we call the 'chance' of the sceptical possibility being realized.
Unless this is done, an airy assertion that there is no chance of things being
like that will sound quite unsupportable, and the sceptic wins. There is only
one way that I can see of respecting the possibilities but avoiding scepticism,
and that is to improve the theory of truth, for modal assertions and for
assertions of chance. We have to say that although there is a real space of
possibilities, as the sceptic maintains, there is also no chance that any of those
11. Stroud, op. cit., chapter 4.
 Knowledge, Truth, and Reliability 47

possibilities are realized: it is known that we are not unreliable. Are there
doctrines in the theory of truth that enable us to say this?

IV
The sovereign proposal is to think of truth as some kind of construct out of
our conception of the virtues of methods of inquiry and the consequences to
which they lead. 'Realism', in at least one good use of the term in this
connexion, thinks that we can explain the virtues of method by certifying that
they are midwives to truth; 'anti-realism' sees truth as that which ought to be
established, or would be established, by the best use of the best methods. The
one philosophy sees the virtues of right reasoning as a precipitate from an
antecedent notion of truth, and the other reverses the priority.
It is often suggested that the anti-realist direction makes for an easy dis-
missal of scepticism.12 For instance, it is supposed that on the philosophy of
the later Wittgenstein, our procedures and practices of ascribing pain to other
people, together perhaps with the consequences we attach to such a descrip-
tion, determine what we mean by it. This leaves it open, it is supposed, that
such ascriptions are defeasible, so that any finite evidence for the ascription
can lead us to be wrong. But it is then supposed that the priority of as-
sertibility conditions forbids us from making sense of the sceptical possibility
that the world contains no consciousnesses but mine (or those of some
favoured subgroup including me): stoicism, pretence, and so forth can exist
only against a background of general correctness, and this correctness is sup-
posedly guaranteed by the criterial, practice-governed conception of meaning.
I find this obscure. The practice of attributing mental states to others leaves
open the possibility of error in the face of finite behavioural evidence. If it
leaves this possibility in each case, then even if it does not follow that it may
do so in every case, still we must ask why it does not do so in the conjunction
of individual cases—that is, as regards the world in general. The concept of
virtue attached to such ascriptions may leave us quite unable to reject this
bare possibility. Rejecting possibilities of error may be no part of the practice,
and not entailed by the virtues or ways of reasoning that are integral to the
practice.
There is another way of raising this problem. Once more, suppose we
sympathize with the anti-realist priorities. Then I might be confident that the
best possible system of belief about other people, formed by the most virtuous
dispositions, should contain the belief that others see colours as I do. But it
might also contain the proposition that it is possible, in spite of any of the
evidence I have or could have, that they do not. It would contain this proposi-
tion if the idealized increases in information or virtue do not rule it out, or, in
other words, if even a supremely virtuous cognitive agent, using information
12. For a typical assessment see Colin McGinn, 'An A Priori Argument for Realism',
Journal of Philosophy, 1979. But see K. Winkler, 'Scepticism and Anti-Realism', Mind, 1985.
48 Metaphysics

as admirably as possible, should still allow or respect that possibility. And

perhaps he should, for perhaps none of the increases in virtue that we can
imagine to ourselves would ever lead to a proper refusal to countenance it.
And in that case, it will be correct to allow it—on the anti-realist's own
construction, it exists. In the same way it would be correct to allow Horatio's
possibility, that there may be truths which would lie forever outside our com-
prehension. It would be the part of virtue to admit as much. Even God would
be right to doubt the guaranteed nature of what is, nevertheless, his own
knowledge; it would be part of the 'final best science' (this is a truth that the
final best science could not acknowledge, disproving that definition of truth,
or alternatively disproving the existence of both a final best science and
truth).
It will be evident then that I differ from both Putnam and Dummett in
holding that the question of priority does not coincide with the issue of
whether we can understand 'verification-transcendent' truth-conditions, or in
other words, allow sceptical possibilities. To that, both my sides reply that we
do so in so far as our practices contain as a legitimate element an enterprise of
wondering whether, in the largest respects, the world is as we take it to be.
This result accords with what I call 'quasi-realism', for it is another respect in
which someone who approves of the anti-realist instinct over the priority of
truth or virtue still ends up with the very thoughts that the realist took for his
own. In turn, the suggestion casts doubt on whether we really have an issue
between global anti-realism and global realism, for if each side ought to end
up following the same practice, there may be nothing to dispute over. But I
would urge that sometimes, in local areas, we can make sense of the diver-
gence of priorities and even award the victory to one side. For instance, I
believe that in the theory of morality, or modality, or chance, there is an
advantage to the side that starts off by regarding method as fundamental.
Moral and modal propositions and, most notably for present purposes, those
about chance, gain their identity, and the identity of the concept of truth to
associate with them, from their place in a two-sided practice—that of coming
to them, on the one hand, and of using them to guide the conduct of life and
thought, on the other. It is therefore particularly attractive not to try to
explain their role by postulating an antecedent notion of truth to which they
answer—a layer of facts about distributions of possible worlds, or of chances
defined over them—but rather to explain what is to count as truth in their case
by thinking of what it is for them to perform their role successfully. One might
try to say that this is always the case, so that if these propositions are better
understood this way, any proposition would be. But this does not follow.
These propositions may find their place in steering us around the facts, as it
were, rather than in describing new layers of fact, but it would not follow that
it is possible for all propositions to be like that. And these propositions (and
those of mathematics) share peculiarities that contrast them with others from
the outset, and that make the notion of truth so problematic in their case.
They have no recognized epistemology, and their truth is not the starting point
 Knowledge, Truth, and Reliability 49

of any serious explanatory theory of our experience. So perhaps they do not

serve as an attractive model for a general debate.13
So far, things look even better for scepticism (again, remember that this is
in spite of the relatively weak account of knowledge I am offering). The
sceptic is not silenced by the highly abstract changeover from, say, 'metaphysi-
cal' to 'internal' realism (anyone who has ever taken the problem of induction
seriously may wonder why he should have been thought to be put out by such
a charge). But we are owed an account of the relevant assessment of chances,
and here there is scope for a response. Suppose we put some fledgling, anti-
realist thoughts about the truth about chance alongside the position in which
we now find scepticism. Hume denied, rightly, that any durable good can
come of extreme, or Cartesian, scepticism. But that is not the same as saying
that durable good comes from disallowing the sceptical possibility. We have
already urged that virtue may involve respecting general possibilities of error.
But we have not yet seen that the virtuous method of forming belief about our
reliability should leave us any sense of a real chance that those possibilites are
realized. Can we hold the line against scepticism at just this point?
Scepticism invites us to 'stand back' and think of a logical space of possibili-
ties, many of which accord with our evidence, but only one of which accords
with the way we take actuality to be. When we do this we are apt to think that
there is a real probability measure, meaning that some such possibilities have
a better chance of realization than others. On an anti-realist line about
chances, matters are the other way round: proper confidence itself determines
what we are to say about any such measure. The ordinary considerations in
favour of induction, other minds, and so on, have to give us a title to say that
there is simply no chance (or dismissably small chance) of things not being as
we take them to be.
Saying that there is no such chance will sound like saying that we know
that a particular ticket will not win a lottery—something that is usually false,
since we can have no authority on whether such an event will happen. But it
is not like that. For there we have a kind of thing that does happen (individ-
ual tickets win) and our reliability over whether it is going to happen in a
particular case cannot be better than chance. Here we know the chances—
they supervene upon natural facts in our world. But nobody has any right to
say that massive undetectable facts according with sceptical suggestions are
the kinds of thing that happen: we are not flying in the face of an actual
empirical kind with a given frequency of realization when we deny them any
chance at all.
There is now an opening for the sceptic to ascend a level. What I am
doing, he will say, is denying that there is a real 'trans-world' probability
metric giving his possibilities a real chance of being actual. I am saying that
chances are properly to be evaluated in the actual world, in which, I suppose,

13. Of course, essay 1 explores the possibility of generalizing. Essays 3 and 4 go into more
detail about modal judgement and judgements of chance, respectively.
50 Metaphysics

things like massive undetectable failures in the mentality of others, or failures

of induction or memory, do not happen. But, he will complain, this is not
sufficient. Suppose that the idea of trans-world chances is incoherent. Perhaps
if we knew enough about the world, we could also say that there was no
chance of bizarre possibilities being realized: contingently the chance of his
possibilities is actually zero. But his claim is precisely that we do not know this
much about the world: for all we know about contingent reality, the chances
of (say) there being no other minds, or of the world conforming to Good-
manesque bent predictions, is quite high. In other words, denying a trans-
world metric on chance is not enough, for it still leaves us ignorant of the
distribution of chances at our actual world. And, the sceptic continues, the
chances may be pretty unfavourable.
I think this admits of no refutation, for it depends entirely upon who bears
the onus of proof. If the sceptic's task were to prove that there is a chance of
massive falsity in commonsense beliefs, then he fails. We can maintain, in one
breath, both our normal beliefs and the corresponding title to knowledge,
since there is no chance of their being false. The sceptic cannot dislodge us,
since he cannot prove the existence of the disturbing, knowledge-defeating
chances. On the other hand, we cannot prove against him that the relevant
chances are zero—not without helping ourselves to the very contingent knowl-
edge'that he wishes to deny us.
(Perhaps there is more to be said here along these lines. I have been
emphasizing the continuity between discussion of scepticism and everyday
discussion of chance, albeit that the latter is heavily constrained, empirically,
in a way that the former is not. Now someone might urge that this continuity is
spurious: it hides the crucial difference that in the ordinary case we have
procedures for assessing chances, but all of these, as we have seen, can be
disallowed by the sceptic. Can this be used to urge that the position and the
debate are indeed taking words out of the contexts in which alone they make
sense, so restoring suspicion of the whole issue? I do not think so. For exam-
ple, wrestling inductive procedures into an anti-sceptical shape is a perfectly
recognizable activity, bound by the same rules of chance, the same kinds of
arguments, that hold sway elsewhere. The trouble is that it tends not to issue
in anything that impresses the sceptic.)
So all we are left with to urge against him is a refusal to accept the onus.
Allowing chances, like allowing possibilities, is something that we do. Some-
times, when the actual world affords stable frequencies, doing it well is heav-
ily constrained by natural facts. But when we think of the chance of the world
being as sceptical thoughts suggest it might be, we are not so constrained.
Then, we are only aiming to express the proper, best way in which confidence
is to follow on from the use of the ordinary evidence in favour of common-
sense beliefs. So we can properly allow the bare sceptical possibility and
properly disallow any chance whatsoever that it is realized. Durable good
comes from both policies. Neither flies in the face of a real, trans-world
distribution of possibilities or chances, for there is no such thing.
 Knowledge, Truth, and Reliability 51

V
This may not be a particularly glorious victory over the sceptic. I do not mind
this—indeed, like Stroud, I would mind more if the victory had been gained
by the kind of dismissal that refuses to acknowledge the deep legitimacy of
sceptical worries. The deep roots of scepticism lie in the need to see ourselves
as reliable over as many matters as possible. It is not the use of inappropriate
standards, nor shifting to a different and doubtful external point of view, nor
yet accepting an unbearably strong cutoff point for knowledge, which leads us
to focus upon our most general methods and ask for their reliability. It is, as
we might put it, not trying to hover above our boat with a new and unfamiliar
set of a priori instruments of inspection; it is just using the same instruments
on the same boat, but on a little-visited and basic part of its structure. Unless
this is seen, scepticism will not have had an adequate answer. For that very
reason the problem of knowledge, as we have inherited it, or the very subject
of epistemology, should not be seen as the parochial, historical outcome of
mistaken conceptions of mind or experience. It may produce stunted side-
shoots because of such mistakes, but when they are lopped off the problem of
knowledge remains. For, given problems such as those of observation and
induction, we have no stable way of imagining a body of knowledge that
protects all the exposed surfaces where the title to reliability is vulnerable and
needs questioning, protecting, reconceiving.
Wittgenstein imagined that the philosopher was like a therapist whose task
was to put problems finally to rest and to cure us of being bewitched by them.
So we are told to stop, to shut off lines of inquiry, not to find things puzzling
nor to seek explanations. This is intellectual suicide. If the philosopher is
indeed like a therapist, then his task is to insist upon constant exercise: the
inspection of the bindings, the exposed surfaces, the possibilities and chances,
the dangerous places where a sense of our own reliability takes no place in the
rest of our scheme of things.
 3
Morals and Modals

I. Introduction
Conclusions properly drawn must be true when the premises are; events must
unfold in accordance with natural law; people must obey the moral laws. Why
do we find it so tricky to give a satisfactory philosophy of these necessities? In
the first part of this essay, I suggest that it is because we have a rooted, and
inadequate, conception of what is needed to establish such an understanding.
This conception dominates the philosophy of modality, just as it does other
areas, but it makes a genuine advance in understanding impossible. The diag-
nosis here is quite simple, but it is not so simple to disentangle ourselves from
its influence, and to become practised with tools that are better suited to the
problem.
What would a philosophical theory of logical, natural, or moral necessity
be? By making judgements of necessity we say things, and these things are
true or false. Perplexity arises because we think there must therefore be
something which makes them so, but we cannot quite imagine or understand
what this is. Nor do we understand how we know about whatever this is: we
do not understand our own must-detecting faculty. Elucidating the truth-
condition, and our access to it, is the goal of philosophy, to which its tech-
niques and controversies are essentially directed. Not only is this so, but
surely it has to be so, for the philosophical itch is that of finding the nature of
the facts strange and incomprehensible, of failing to imagine what could make
true the relevant judgements. The problem is that of the fugitive fact, and the
solution is to capture the nature of the fact in an intelligible way. This answer
would tell us what such truths consist in: the answer would be obtained by
establishing the truth-conditions for such judgements. It would give us an
'account' of the states of affairs in which their truth consists, or of what it is
that makes them true. The account would have an explanatory role as well:
fully established, it would explain why it is necessary that twice two is four, or
how it can be that natural laws exist, or why we must be nice to one another.
The most direct technique would be analysis, showing, it might be hoped, that
the judgements are made true by some state of affairs relatively familiar and
unproblematic (by whichever standards prompted the perplexity). Another
technique would be more aggressive: to suggest that the concepts involved in

52
 Morals and Modals 53

the judgements are defective and due for replacement, so that the fugitive
'facts' were not really such, not really worth chasing after all.
Within this conception of the philosopher's quest, there is room for dis-
agreement over detail—for instance, whether the description of the state of
affairs finally fixed upon as making true the original modal judgement has to
be synonymous with that judgement; whether one range of arguments or
another succeeds in showing some concepts to be defective, or over what
would count as an admissible reduction class for the modal claims. It is to the
twists of this detail that we naturally turn when faced with the embarrassment
that the head-on search for truth-conditions for modal assertions has turned
up nothing at all promising. Where else is there to turn? For rejecting the
problem is too much like ignoring the itch.
The modal concepts need a theory. But I do not think that they need or
could possibly get a theory described, however remotely, in the terms sug-
gested so far. In other words, I think that we have completely misinterpreted
the kind of solution the philosophical problem needs. This may seem surpris-
ing, for I posed the problem and the kind of solution in terms deliberately
bland—the kind of terms that would go quite unremarked as a preface to
discussions. But I shall argue that they mislead us, and that a better way to
approach the matter exists.

II. The Quasi-Realist Alternative

Let us call the direct approach the truth-conditions approach. Here is a di-
lemma that attends it, and that I shall exhibit quite generally for moral,
natural, or logical necessity. If we ask what makes it so that A must be the
case, we may be given a local proof, a proof of A from B. This is satisfactory if
we already understand why B must be so (if our topic is logical necessity, there
is also the status of the proof to consider). But if our concern is with the whole
area, then we turn to scrutinize that understanding. Attention shifts to why B
must be the case, for our philosophical concern is with necessity in general,
not with A in particular. Suppose an eventual answer cites some truth F, and
so takes the form: ' A because F'. ('Because' here is taken to include consti-
tutive variants: the truth that A consists in F, is made so by F, etc.)
Now, either F will claim just that something is so, or it will claim that
something must be so. If the latter, there is no problem about the form of the
explanation, for one necessity can well explain another. But, as we have seen,
there will be the same bad residual 'must': the advance will be representable
as 'if we see why this must be so, we can now see why that must be as well'.
And there is no escape from the overall problem that way. Suppose instead
that F just cites that something is so. If whatever it is does not have to be so,
then there is strong pressure to feel that the original necessity has not been
explained or identified, so much as undermined. For example, suppose a
theorist claims that twice two must be four because of a linguistic convention,
or that particles must attract each other thus because of some ongoing cosmic
54 Metaphysics

setup, or that we must be nice to one another because that is what God wants.
Suppose it is denied that there is any residual necessity, that we must make
just those conventions, that laws determine the consequences and continua-
tion of the cosmic setup, or that God's wants ought to be heeded. Then in
each case there is a principled difficulty about seeing how the kind of fact cited
could institute or be responsible for the necessity. This is because the explana-
tion, if good, would undermine the original modal status: if that's all there is
to it, then twice two does not have to be four, particles don't have to attract
each other, and we don't have to be nice to each other, even if it would be
unwise not to. This is, of course, a generalization of the famous Euthyphro
dilemma. Either the explanandum shares the modal status of the original, and
leaves us dissatisfied, or it does not, and leaves us equally dissatisfied.
So why is the truth-conditional approach so dominant—why is this di-
lemma not universally recognized? Partly at least because it leaves room for
work. The circle can be virtuous and explanatory. In other words, there is no
embargo on finding theories of the form ' p because F' where F stays within
the modal sphere in question—' p because in all possible worlds p'; ' p
because there is a relation of necessitation between certain universals', or ' p
because ~ p is impermissible', for example. Such theories can and do uncover
important aspects of our thought: making the logic of modality intelligible, for
instance. But from the standpoint that prompts the original problem—the
dissatisfaction with the fugitive fact—by staying within the family in question,
the analyses cannot do more than postpone things. Of course, at one level this
is perfectly well known, for everyone agrees that it is one thing to have a
possible-worlds approach to modality, for example, and quite another to have
a theory of the metaphysics or epistemology of the things we say about possi-
ble worlds.
The poor prospects of the truth-conditional approach would be easier to
tolerate if there were another approach. Fortunately, there is. The truth-
conditional approach looks for another way of characterizing the 'layer of
reality' that makes true modal utterances. The alternative starts (and, I shall
urge, ends) with our making of those utterances: the thing we intend by
insisting upon a necessity or allowing a possibility. We could call it a 'concep-
tual role' or even a 'use' approach, but neither title is quite happy, for neither
makes plain the contrast with truth-conditional approaches that is needed.
The conceptual role of use of a modal idiom might be just that of expressing
belief in the fugitive layer of fact! If the best that can be said about our
commitments is that they are those of people who believe in particular distri-
butions of possibilities—logical, natural, or moral—then we are silenced
again. But this may not be the best that there is to say: we can approach the
commitments differently.
This alternative is familiar under the heading of projectivism (or some-
times, which is worse, 'noncognitivism') in ethics: this is why in setting the
scene I have includes moral musts. It has been pioneered in the philosophy of
natural law by Ramsey and Ayer, and my aim is to made it a recognized
option in the metaphysics of modality.
 Morals and Modals 55

Notice that this is not the alternative of saying that 'there are no laws of
nature' (or no possible worlds), any more than projective theory of ethics
involves the 'eccentric' view that there are no obligations. Instead, this ap-
proach gives its own account of what it is to say that there are, and, if the
commitments are valuable, why it is correct to do so. The account has two
stages. It starts with a theory of the mental state expressed by commitments in
the area in question: the habits, dispositions, attitudes they serve to express.
It is these that are voiced when we express such commitments in the ordinary
mode: when we say that there exists this possibility, that necessity, this obliga-
tion. The second stage (which I called quasi-realism) explains on this basis the
propositional behaviour of the commitments—the reason why they become
objects of doubt or knowledge, probability, truth, or falsity. The aim is to see
these propositions as constructions that stand at a needed point in our cogni-
tive lives—they are the objects to be discussed, rejected, or improved upon
when the habits, dispositions, or attitudes need discussion, rejection, or im-
provement. Their truth corresponds to correctness in these mental states, by
whichever standards they have to meet. Such a theory only collapses back into
realism if we are reduced to saying that correctness in modal or moral judge-
ment is simply representing the modal or moral facts as they are. But accord-
ing to my direction of theorizing, we can do better than that, and what we can
do involves no irreducible appeal to a moral or modal reality. It is here that
the opposition to realism lies, although I shall try to make it plain that the
interest of the approach remains even if, as I also believe, there is no very
coherent realism for it to be 'anti'.
It is tempting to characterize this anti-realism as an 'as-if philosophy: we
talk as if there exist moral or modal facts, when in fact there are none. This
makes it sound as though, according to this approach, some error of expres-
sion or thought is involved in such talk—for we talk as if p, when in fact p is
false. This consequence of an as-if characterization is especially tempting when
we remember other areas in philosophy where such projections are supposed
to be responsible for mistakes we make—pathetic fallacies, for instance. Spi-
noza, for example, believed that what we take to be contingency in the world is
merely a reflection of our ignorance, and this diagnoses a mistaken belief that
we have.1 Most writers on projective theories of morals and modals mention
Hume, of course, and then continue with some version of this:
Hume's view is that we then make a mistake: we project something essen-
tially 'inner' onto the external world, and come to the mistaken belief that
the concept of necessity we have applies to propositions in virtue of the
objective properties of ideas and, as a consequence of this, we mistakenly
believe that modal judgements can be true or false.2
There is excuse for the interpretation, for Hume is not as clear as one might
wish. The first passage in which he appeals to the metaphor of the mind
1. Spinoza, The Ethics, Part II, Prop. XL1V. I owe the reference to Al McKay.
2. Graeme Forbes, The Metaphysics of Modality (Oxford: Oxford University Press, 1985),
p. 218.
56 Metaphysics

spreading itself on external objects is in the context of diagnosing a mistake—

the 'contrary bias' that leads people to ridicule his philosophy of causation, to
suppose that, by making the 'efficacy of causes lie in the determination of the
mind', Hume is reversing the order of nature.3 But this does not show Hume
admitting that, by talking of causes (or obligations or necessary relations of
ideas) as we do, we make any mistake. The theorist may misinterpret the
nature of our judgements, their origins, and the standards that justify them.
But the first-order user of the vocabulary makes no mistake: there is decisive
evidence that Hume thought he made none. This is clearest in the moral case,
of course, for Hume's philosophy of natural belief is infected by the back-
ground problem that our belief in the external world in any case involves a
mistake—natural and inevitable propensities of the mind that must lead us to
falsehood. But there is no further mistake involved in 'causalizing'—in find-
ing causal order in the world we take ourselves to inhabit—any more than
there is in moralizing as a reaction to characters and actions.
Hume's position is best explained by separating two different applications
of the notion of projection. In the one use (which I prefer) we 'project' when
we use the ordinary prepositional expressions of our commitments, saying
that there is this causal relation, that natural law, this other obligation. In the
other we project only when we adopt, as philosophers, a particular 'realist'
explanation of the sphere in question. This is a quite different thing, and it is
what gave the contrary bias of which Hume is indeed complaining. The space
between the two uses is easily missed, especially by philosophers coming with
a realist bias in the first place. For they will be only too apt to suppose that the
ordinary use has, as it were, done their work for them, so that a realist
ontology is the only possible explanation of the first-order usage. But this, in
Hume's view and mine, is not so. And this view must be given a hearing.
How can a projective theory accompany the view that no mistake is made
in talking as we do? We would only make a mistake in saying that things ought
to be done, or have to be so, if these judgements have a false content. But if
their content arises as the projectivist + quasi-realist story maintains, they do
not. No error occurs in moralizing or modalizing, even if philosophers have
mistaken the kind of content these judgements have. Error exists only if there
is a real mismatch between the truth about the nature of the claims, and their
content, or what we make them do in our theories of things. But no mismatch
exists in the thought that '1 + 1 = 2', that bees cause stings, and so on.
Quite apart from the implication that we make some kind of mistake, an
as-if description of the theory makes it appear inadequate to the depth of our
commitments. It looks refutable by a kind of phenomenological reminder of
the strength of our belief that there really are possibilities, necessities, etc.
Don't you believe that there really are natural laws, iron proofs, genuine
duties? It is not just that we talk as if there are such things! But a quasi-realist
will properly say: it is not simply that we think and behave as if there are
3. David Hume, A Treatise on Human Nature, Book 1, Pt. III, Sec. XIV, ed. L. A. Selby
Bigge (Oxford: Oxford University Press), p. 167.
 Morals and Modals 57

necessities: there are. And we are right to think that there are. The commit-
ment, and its correct expression, should not be in question.
What then is the mistake in describing such a philosophy as holding that
'we talk as if there are necessities when really there are none"? It is the failure
to notice that the quasi-realist need allow no sense to what follows the 'as if
except one in which it is true. And conversely, he need allow no sense to the
contrasting proposition in which it in turn is true. He no more need allow such
sense than (say) one holding Locke's theory of colour need accept the view
that we talk as if there are colours, when there are actually none. This is
doubly incorrect, because nothing in the Lockean view forces us to allow any
sense to 'there are colours' except one in which it is true; conversely, neither
need it permit a sense to 'there are actually none' in which that is true.
Theorists may construct such senses: for instance, a sense in which 'there are
colours' implies that colours do some work of physical explanation, or could
be apprehended by more than one sense, and of course the Lockean will deny
anything implying such a thing. But if the words retain an uncorrupted, En-
glish, sense, then the Lockean, and similarly the quasi-realist, holds not just
that we talk and think as if there are . . . , but that there are.4
Then the objection might be rephrased: according to the quasi-realist, we
think and talk as if there were real moral and modalfacts, but there are none.
However, this too, although it points in a better direction, invites misunder-
standing. It cannot stand as an accurate diagnosis of a position, for the word
'fact' also has an uncorrupted English sense: it is a fact that there are colours,
and there are many facts about them. Certainly, there is a sense in which the
quasi-realist is opposed to giving an ontological status to moral and modal
facts, but according to him you cannot read off this status just from the nature
of our commitments, their modes of expression, or their genuine place in our
thinking, even if that thinking goes on invoking talk of facts. The appearance
tempts philosophers to ontological quests, puzzles, and errors, but the mis-
take lies with the theorist who succumbs to the temptation.
Of what then is the quasi-realist suspicious? We can see now how the
problem of characterizing either realism or anti-realism becomes acute. Sup-
pose, for instance, we are satisfied with a quasi-realist construction of modal-
ity: we see what we are doing when we modalize, and why talking of possibili-
ties or possible worlds is a legitimate form for these commitments to take. So
when a writer such as Lewis maintains the irreducible nature of the modal
idiom and expresses his commitments in that idiom, he is doing no more than
a quasi-realist allows. What more does he intend by deeming himself a realist?
How is there to be space, as it were, for some extra content in any such claim?
One might see illegitimate content: if a theorist held that alternative possibili-
ties are real in the sense that we can find them in space or hold them responsi-
ble for causing various results, or if he took comfort in the thought that he
could model apprehension of possibilities upon sensory apprehension. But
4. I do not have a fixed opinion on what Locke himself thought about the existence of
colour. See P.A. Boghossian and ID. Velleman, 'Colour as a Secondary Quality', Mind, 1989.
58 Metaphysics

theorists, including Lewis, call themselves modal realists without accepting

any such theses. It begins to look as if there is no way of framing an ontologi-
cal or metaphysical opposition. Saying 'I believe in possible worlds, and I am
(or: I am not) a realist about them' would amount to no more than accepting
irreducible modal idioms, and in either form the last conjunct is quite idle.
Universal harmony is desirable, but it does not come quite so cheaply. The
difficulty of characterizing the dispute shows that it is up to anyone who takes
pride in announcing himself in this style to make sure that the last conjunct
has a content. And in my view, many philosophers who take pleasure in
calling themselves 'moral realists' have failed badly in this obligation. They
have either been content to pour cold water on revisionary anti-realism of
John Mackie's kind, or content to insist on the surface appearances, or con-
tent to generalize what is mistakenly seen as a late Wittgensteinian lesson, to
the effect that every indicative sentence shares the same role—that of describ-
ing an aspect of the world ('our world'). The existence of the kind of theory I
am describing should undercut this. But there is still room for disagreement,
specifically about what in the commitments needs explaining, and about the
kind of explaining modal and moral facts can themselves do.
Realist theorizing is apt to pay too little attention to the first and to make
too much of the second. It worries too little about the curious place that moral
and modal commitments have, about what notion of truth can be appropriate
to them, about why it matters, and about how the commitments blend with
others that we have. It worries less about these issues because if these commit-
ments are beliefs, then their aim is simple truth, and this is proper depiction of
the modal or moral realm. This is an application of the second tendency: to
make much of the explanatory powers of the moral or modal states of affairs.
A realist may betray himself, for instance, by relying upon metaphors of
perception or vision to explain how we become acquainted with moral or
modal fact, or by entering false theses about the creation or destruction of
such facts and their dependence on others, or by supposing that the existence
of such facts explains other genuine states of affairs, in the way in which one
state of affairs can explain another. To suppose, for instance, that the world
exists as it does because it ought to do so might be the privilege of the moral
realist. To suppose that the world exists because God made it is the privilege
of the theological realist. If this kind of belief is intrinsic to first-order theoriz-
ing (as in the theological case), then the kind of diagnosis of the commitments
offered by a projectivist will indeed find error in the everyday practice, as well
as in various philosophical interpretations of it; this is why a 'Wittgensteinian'
protection of religious belief is a kind of cheat. Ordinary religious belief,
thought of in an expressive way, involves the mismatch referred to above. This
is also why there is very doubtfully any space for a genuine realist verses anti-
realist debate about explanatory physics. But first-order theories are notably
silent about the explanatory role of possible worlds or moral duties; it is left to
the philosophers to inject good or bad views about that. 5
5. Obviously, in this paragraph I ignore the possibility of the generalized quasi-realist move
introduced in essay 1: the move that allows even the use of a concept in explanatory roles, but still
 Morals and Modals 59

Once the explanations are agreed, not much is left in the words. So the
universal harmony is better approached in a case like that of colour, where we
feel reasonably confident of the underlying facts and the way they relate to
colour perception. And then indeed it is no great matter whether we say that
there are colours (and I am a realist about them) or that there are (and I am
not). The space for dispute has shrunk away and can only be resurrected if
false implications are read into the parenthetical remarks. It is no great trick
to announce oneself in either style; the work comes in earning a right to do so.
But to achieve this harmony in the modal case involves the hard work of
showing how to explain modalizing in the first place, and this remains to be
done.
At the risk of appearing moralistic, I shall close this section by illustrating
how truth-conditional theorizing dominates our philosophical imaginations.
One of the clearest expressive approaches to commitment to natural law is
that of Ramsey and Ayer. Here is Ayer:
In short I propose to explain the distinction between generalizations of law
and generalizations for fact, and thereby to give some account of what a law
of nature is, by the indirect method of analysing the distinction between
treating a generalization as a statement of law and treating it as a statement
of fact.6

It is, however, a little unclear from this way of setting it up quite how Ayer
conceives the step from a theory of what it is to treat something as a law of
nature to giving 'some account of what a law of nature is'—the ontological
overtone of this suggests that the truth-conditional theory is not quite exor-
cized. For if the expressive theory is successful, there is no last chapter to
write on what a modal fact or state of affairs is. We would know what we do
and why we are correct to do it when we commit ourselves to necessities of
logic, nature or action, and that would be the end. Ayer's nod towards truth-
conditional hankerings is wholesale prostration in other writers. A recent
example is David Armstrong. After observing that inference from the ob-
served to the unobserved is central to our whole life as human beings, and that
if there were no laws those inferences would be unreliable, he continues:
'hence the notion of law is, or should be, a central concept for epistemology. If
so we will also want to enquire into its ontology. We will want to know what a
law of nature is'.7 The grip of the truth-conditional approach appears when
Armstrong considers the alternative to this, which he identifies as the 'truly
eccentric view . . . which denies that there are any Laws'.8

defends an anti-realist construction of it. It is not that I changed my mind between the two papers,
or between then and now, but that for the purposes of this paper it is the different direction of a
quasi-realist story that is important. Even if explanatory contexts eventually fall within the quasi-
realist net, it is not right to start with them.
6. A. J. Ayer, 'What Is a Law of Nature?', The Concept of Person (London: Macmillan,
1963), p. 231.
7. D. Armstrong, What is a Law of Nature? (Cambridge: Cambridge University Press, 1983)
p. 5.
8. Ibid., p. 5.
60 Metaphysics

Even writers as cautious as Edward Craig and Crispin Wright find it

straightforward to agree on the point that, in effect, closes off projectivism +
quasi-realism. The context here is that Craig had demonstrated decisively the
imaginative block that faces us when we try to conceive, in proper detail, of a
counterarithmetical reality. The projectivist is then poised to see this imagina-
tive block as something expressed when we insist upon the necessity of arith-
metic. But Wright commented, 'If as Craig makes plausible, we are unable to
conceive of how any alternative determination might be viable, then that is
how things are with us; it is a further, tendentious step to inflate our imagina-
tive limitations into a metaphysical discovery'.9 And Craig, acknowledging
that he and Wright agree that we should not ask the imagination to do too
much, concedes immediately: 'It certainly is a further step'.10 Is it so clear that
there is this further step? Only if claims of necessity are 'metaphysical discover-
ies', and this the projectivist will query. Again, the position is clear if we
revert to the moral case: a projectivist will see commitment to an obligation as
a distinctive mental state—call it a sentiment—but he will not accept any
charge that we tendentiously inflate our sentiments into metaphysical discover-
ies (discoveries about the independent structure of the world of obligations),
precisely because he denies that in our awareness of duty and obligation we
are in fact making any such discoveries. (I return later to Craig's reasons,
which were good, for thinking there is a further step—only it is not this one.)
There are aspects of the work of making quasi-realism attractive that I
shall not repeat in this paper. These include its distinction from naive sub-
jectivism, its moves to accommodate the prepositional nature of ethical
claims, its explanation of the syntax and semantics that go with that, and the
basis for constructing a working notion of truth. My concern here is to see
how this shape of theory fares with one of the other two 'musts': that of logic.

III. Policies versus Needs

We allow possibilities, rule out impossibilities, and insist upon necessities.
This is not describing anything. As in Wittgenstein, attributing necessity to a
proposition is not making a true or false claim about it—or at least is not to be
understood that way.11 It is more like adopting a norm, or a policy or a rule
that a thesis be put 'in the archives', above the hurly-burly of empirical deter-
mination. The decision dictates how we shall treat recalcitrant evidence. This
accords with the parallel with morals. The one kind of rule makes courses of
thought intellectually obligatory; the other makes courses of action so. But
9. Crispin Wright, Wittgenstein on the Foundations of Mathematics (London: Duckworth,
1980), p. 439.
10. E. J. Craig, 'Arithmetic and Fact', in Exercises in Analysis ed. Ian Hacking (Cambridge:
Cambridge University Press, 1985), p. 90.
11. For example, Remarks on the Foundations of Mathematics (Oxford: Basil Blackwell,
1956), V. 6, 4-5, p. 163. Wittgenstein constantly appeals to hidden difference of role underlying
superficial descriptive appearances. See also essay 8, note 13.
 Morals and Modals 61

there is a major problem: to identify any space for this rule-making. Modal-
izing, like moralizing, does not feel optional: it feels as though we regard '1 +
1 = 2' as necessary simply because we must do so, not because we have chosen
to do so. Its status is more naturally seen as a product of our inability to
conceive otherwise, or to do anything with a counterarithmetical judgement.
If the necessity of propositions is in any sense conferred by us, it is still
unnatural to see it as reflecting anything in which we had a choice. So notwith-
standing Wittgenstein, a projectivist will be wise to look for the mental set
that gains expression outside the realms of the optional (and it is vital to
notice that he can do this—the denial of metaphysical realism does not usher
in a 1950s embrace of free choice and conventionalism).
If attributing modal value reflected free policies and choices, it would be
unclear why we should go in for it. The right attitude would seem to be that
which Wright attributes to his imagined 'Cautious Man'.12 This is the character
who agrees with us on all empirical truth. He agrees with us too in accepting
proofs; in arithmetic or logic, or in any more apparently metaphysical commit-
ments, such as those determining our basic ascriptions of temporal, spatial, or
causal categories, this character agrees with us. But he refuses to make modal
assignments. As far as he is concerned, it is enough that we accept, say, that 1
+ 1 = 2. It is unwise to go further and ascribe necessity to the proposition.
The challenge is reminiscent of Quine: would it not be better simply to
register our stronger attachment to some propositions than others, and then to
leave market forces to determine which ones maintain our loyalty? Even if we
abandon the self-image of decision makers, we confront essentially the same
problem. What would be lost if we simply did not modalize? Is it not foolish to
elevate mere imaginative limitations into iron necessities?
Quine thinks that even in the case of logic we would be better off doing no
such thing. Of course, in the context of positivism, Quine's strength lay not so
much in opposition to modal discrimination in itself, as in his insistence that
coming to the problem with notions of meaning or convention is coming with
dirty hands: there can be no modally innocent appeal to conventions, or
concepts or meanings or rules or languages, giving us an anterior understand-
ing from which to explain or justify those discriminations. In other words,
even if we can say things like 'analytic propositions are true in virtue of
meaning/concepts/constraints on the application of concepts . . .' this is no
help. It is no help because there is no identification of concepts, meanings,
etc., which does not itself involve knowing the modal liaisons of propositions
in which the concepts occur—what must be, may be, or cannot be true, if they
are so. Hence, any such appeal cannot explain or justify our modal commit-
ments: in a frequent metaphor, it keeps us within the same circle.
It may have been naive of the positivists to think that by retreating to
questions of meaning we obtained a cleanhanded empiricist aproach to modal-
ity. But overthrowing that is not the same as overthrowing the modal. Indeed,
the 'dirty-hands' argument is entirely two-edged: by showing how deeply the
12. Wright, op. cit., chapter 23.
62 Metaphysics

modal is entrenched in any 'conceptual scheme' it makes it less likely that

modalizing is left an unprotected optional extra in our thought. But so far as
the present essay goes, the point to insist upon is that there is clearly an
antecedent problem for any naturalistic sanitizing of the modal. This is to
explain the way in which we make modal judgements—the ease with which
we noncollusively agree upon them. Obviously, before we recommend that we
abandon modalizing, we want to know what it involves and why we do it. Our
capacity to make noncollusive modal discriminations requires explanation,
whether or not it is regrettable that we do so. But curiously enough (since the
task is one of naturalized epistemology), Quine's philosophy of the modal is
incapable of meeting this eminently naturalistic request, and when it is but-
tressed to do so, it loses its appeal, doing better by becoming quasi-realistic.
Or so I shall argue.

IV. Explaining Modalizing

Quine's consistent position has been that even when we think of the most
elementary trivialities of truth-functional logic, the best we can say is that they
are obvious. It is sometimes said that he changed his mind about this, and
that, discussing translation from allegedly pre-logical or alternative-logical
tongues, he conceded some very special status to truth-functional logic, in the
determination with which we would reinterpret others as conforming to it.
But in Quine's view this is no shift. It is just a consequence of the fact that we
always translate so as to save the obvious.13
Of course, not all truths naively called necessary are at all obvious, but
Quine can and does extend the explanation to those which can be proved by
obvious means from obvious starting points. Here we have the famous
Quinean picture in which the truths naively called necessary are those which
are obvious enough to lie far away from the theatres of war in which empirical
forces mould and break theories. It substitutes the one-dimensional web of
13. In 'Carnap and Logical Truth' (1954; reprinted in W. V. Quine, The Ways of Paradise and
Other Essays, New York: Random House, 1966, pp. 105-106) promoting the dirty-hands argu-
ment, he wrote "The considerations which were adduced in *1, to show the naturalness of the
linguistic doctrine, are likewise seen to be empty when scrutinized in the present spirit. One was
the circumstances that alternative logics are inseparable practically from mere change in usage of
logical words. Another was that illogical cultures are indistinguishable from ill-translated ones.
But both of these circumstances are adequately accounted for by mere obviousness of logical
principles, without help of a linguistic doctrine of logical truth. For, there can be no stronger
evidence of a change in usage than the repudiation of what had been obvious, and no stronger
evidence of bad translation than that it translates earnest affirmations into obvious falsehoods."
And in Philosophy of Logic (Englewood Cliffs, NJ: Prentice-Hall, 1970), in the same context,
again insisting upon the inevitability of our imputing classical logic to a translatee, he offers
almost identical terms: "Being thus built into translation is not an exclusive trait of logic. If the
natives are not prepared to assent to a certain sentence in the rain, then equally we have reason
not to translate the sentence as 'It is raining'. Naturally the native's unreadiness to assent to a
certain sentence gives us reason not to construe the sentence as saying something whose truth
should be obvious to the native at the time. Data of this sort are all we have to go on . . ."
 Morals and Modals 63

belief, with only a vague and pragmatic boundary between propositions that
face the test of experience routinely ('contingent') and those that at worst
would only face it in periods of exceptional theoretical turbulence ('neces-
sary'). And at first sight it gives Quine his answer to the problem of explaining
our noncollusive application of the notion. When we deem a proposition
necessary we express our apprehension of its obvious character.
But a little thought shows that this is quite inadequate. For a great many
truths are in Quine's central reservation, but would simply be classed as
contingent. These are truths that are central, certain, obvious to everyone—
that there exist trees and rocks, that houses keep off the rain, and so on.
There is no prospect of these being rocked by scientific change, nor of recalci-
trant experience casting doubt upon them. But we unhesitatingly class them as
contingent. How is Quine to explain this difference in the modal reaction, if
they are in the scientific archive, beyond the struggles of falsification and
modification?
Quine admits that logic is 'built into translation more fully than other
systematic departments of science. It is in the incidence of the obviousness
that the difference lies . . ,'.14 It looks as if this is to be developed when he
contrasts '1 + 1 = 2', which is 'obvious outright', with 'it is raining', which is
'obvious in particular circumstances'. But the point he apparently has in mind
is just that 'every logical truth is obvious, actually or potentially: each, that is
to say, is either obvious as it stands or can be reached from obvious truths by a
sequence of individually obvious steps'.15 This is the extension referred to
above. But it is not at all clear how it relates to the incidence of the obvious-
ness. And in any event, in a well-developed theoretical science, obviousness
can similarly transmit from obvious data through obvious principles of inter-
pretation and explanation, to bring hitherto unobvious conclusions into the
fold. There is no diagnosis of our different reactions to '1+1 = 2' and 'there
exist trees and rocks' here.
Quine's first thought about the contrast was the best: it is indeed in the
incidence of the obviousness that the difference lies: 'it is raining' is obvious
only in particular circumstances; '1 + 1 = 2' is 'obvious outright'. But 'obvious
in particular circumstances' versus 'obvious outright' is a dangerously sugges-
tive contrast: not far from 'assertible only in the light of particular experience'
versus 'assertible by conceptual means alone', or a posteriori versus a priori. If
the best theory of the incidence of the obviousness is that in the one case but
not the other it varies with particular contingencies, we are left with our
judgement that the truth of the one does so vary, and the truth of the other
does not. This once more is what common sense would say: 'there are trees' is
obvious in the light of something that, we know, could have been otherwise;
not so '1 + 1 = 2'. Another way of putting it is that common sense allows that
recalcitrant experience is possible in the one case but not the other: we could
tell a story in which it came to appear to us as if there were not trees, but not
14. Quine, Philosophy of Logic, p. 82.
15. Ibid., pp. 82-83.
 64 Metaphysics

one in which 1 + 1 is anything other than 2. But Quine cannot appeal to

entrenched modal intuitions to explain the division within the obvious.
The problem, remember, is that Quine is to explain our modal tendencies
before dismissing or sanitizing them—showing that nothing in the making of
them licenses epistemology to draw any grander distinction than his. He is
therefore quite within his rights to call upon his list of theoretical defects here.
Perhaps it is because we are in the grip of mythical theories of ideas, or
molecular theories of meaning, or use-mention confusion, that we distinguish
between equally certain or obvious judgements, identically remote from the
threat of overthrow: there being trees and 1 + 1 being 2. But is it clear from a
naturalist perspective that only a defect is involved—that there is no legiti-
mate point and purpose in the distinction, within the overall class of certain-
ties, between those that are necessary and those are are contingent? Surely
not, and a better explanation of our propensities is easy to produce. Let us
consider the matter from the opposite point of view. It is usually necessity that
is the bugbear, but if we suppose that it is the distinction between the neces-
sary and the contingent that requires understanding, we also can ask what we
miss if we lack the capacity to deem propositions contingent. This direction of
approach must be equally legitimate. In fact, I suspect there is some evidence
that contingency needs more explanation to children than necessity: the initial
tendency is to take everything that is so as having to be so. Suppose someone
who is modally blind in this way: he sees no point or purpose in accepting any
notion of contingency. Running the metaphysics and the epistemology in tan-
dem, we can suppose that epistemologically he can make nothing of the idea
that a particular judgement is a posteriori.16 So he can make nothing of the
idea that although there are trees there might not have been, nor that there
being trees is obvious only in the light of particular experience, so that if the
experience were different (or had been different), as it might be, the opposite
judgement would have seemed right.
What does he miss? The case is still aunderspecified. This person may,
perhaps like Leibniz or Spinoza, have a background theory that all apparent
contingency is disguised necessity. In that case, in the marketplace, or talking
with the vulgar, he could use a perfectly good surrogate for contingency—
perhaps one that he may regard as suited for finite beings. Or perhaps he is
like the Cautious Man, and claims to find some kind of hubris in expressing
verdicts in modal language, although he makes the same distinctions and the
same use of them (for instance, in distinguishing valid from invalid proofs, or
reflecting on alternative possibilities) that we do. This is theoretical or philo-

16. It is beyond the scope of this paper to explore the distinction between necessity and a
prioridty, where the one is thought as logical or metaphysical, and the other as epistemological. I
believe that an 'attitude'-based theory of necessity is able to explain Kripkean intuitions about the
distinction, although the story is not altogether straightforward. The difficult phenomenon to
explain would be the alleged conferring of necessity upon truths that were clearly arrived at a
posteriori. But it should be all right that, after we have discovered something ('water is H2O') we
should 'archive it' at least for some purposes: we do not of course regard such things as truths of
logic, nor can we 'make nothing of the thought processes of one who would deny them.
 Morals and Modals 65

sophical scepticism, and, like its counterparts elsewhere, is supposed to co-

exist with normal living. Such theorists draw the same distinctions as the rest
of us, except that when they think such things as that there might not have
been trees, they will (as it were) preface their assent with a universal qualifica-
tion: contingency becomes some species of apparent contingency, or not the
real thing. This is scepticism, or perhaps idealism, about modality, and not
what I intend. I want instead someone who does not even recognize the need
for a reinterpretation, for he cannot begin by recognizing even apparent con-
tingency as such.
It seems plain that blindness to the a posteriori status of propositions is
catastrophic. To such a person, failure to realize that it is raining here now is
like failure to realize that 1 + 1 = 2, an incomprehensible defect. He is unable
to make anything of a mode of thinking in which it is not realized that p, when
p is true, in just the way that we are unable to when p represents an elemen-
tary necessity. But what does he make of (for instance) sleep, of blindness, of
his own need for telephone directories or testimony, or of the difference that
different spatial and temporal position causes to his own information gather-
ing? How does he think of his own failures of omniscience or conceive of his
own changes of knowledge as he goes about the world? There seems no way
of answering these questions without stripping the subject of massive quanti-
ties of ordinary, nonmodal, empirical understanding—simple understanding
about the variation of belief with circumstance. It would be possible to fill out
the way in which the deficiency disqualifies him from interpreting others
reliably: he cannot rationalize them, seeing why various beliefs seem right to
them because he has no way of seeing how belief varies with point of view,
with use of the senses, with skill or luck. But ignorance of these things in this
context is just a species of ignorance of the way one thing varies with another.
The person who cannot understand how the cat's awareness of the bird varies
with whether it can see it seems little better than one who cannot understand
how the leaf's motion varies with the wind.
Conversely, if the subject has this understanding, he is in a position at least
to imitate modal discriminations. Crucially, he can do better than Quine sug-
gests in making distinctions within the class of the obvious. He can make
something of a way of thought in which it is not realized that there are trees,
just as he can make something of a way of thought in which it is not appreci-
ated that it is raining here now. Long-term confinement to treeless zones is a
kind of position he can understand, and whose impact upon a belief system he
could appreciate. He can say something better about 'it is raining here now'
and '1 + 1=2' than that they are equally obvious. He can say something of
what makes the former obvious, and describe people to whom it would not be
obvious; he can appreciate how there could be, or make something of, a way
of describing the world in which it is denied. Suppose he is set our task of
discriminating, among obvious truths, between those which are intuitively
necessary and those which are contingent; then he can at least approximate to
our division, by simply classing as contingent those which satisfy this condi-
tion: he can make something of ways of thought in which, for various reasons,
66 Metaphysics

they are either not accepted or are even denied. Here 'make something of
will include being able to explain how such a way of thought might arise,
knowing how it might be rectified, understanding the practices of those whose
thought it is, and so on. This will give the subject a sense of what would count
as recalcitrant experience, and what would have counted as such, even for
entrenched, obvious, but contingent, certainties. And, given that there is a
residual class of apparent beliefs where he cannot do this, he will have a
working substitute for the necessary and the impossible.
The upshot is that blindness to the a posteriori character of beliefs seems
impossible in subjects who have virtually any comprehension of the world.
Now naturalized epistemology is largely a study of the variation of belief with
circumstance. It can be done by us only when we can make something of the
variation of belief involved. In some cases we can; in residual cases such as
logic and mathematics we characteristically cannot. This difference can be
used naturalistically to explain our tendency to make modal divisions, and it
gives the explanation that Quine left himself without.
Is it an explanation that can be taken over by Quine? I believe so. Quine
has no reason to oppose our discrimination of contingency; it is the remainder
he dislikes. So his best path would be to accept the explanation of our propen-
sity to modalize, but to warn us against making too much of the imaginative
differences it cites. This would be to join forces with Wright's Cautious Man:
our imaginative limitations are facts about us; they may gain expression in our
modalizing and explain our discriminations, but they ought not to be taken as
any guide to what is necessarily the case.

V. Refining Imagination
The players, then, seem to align themselves into two teams. Both admit the
existence and centrality of imaginative blocks—of the fact that there are
propositions of whose falsity we can make nothing. The one side, encompass-
ing Craig, Wright, this new Quine, Forbes, and probably most others, finds
something distinctive about the Cautious Man, who goes this far, but refuses
to modalize. Quine recommends his modesty; Forbes thinks it would be a
mistake to project imaginative limitations. Craig does not go quite so far. He
indeed thinks there is a further step if we take our imaginative limitations as
guides to what must be the case, namely the step of supposing that the world is
transparent to our intelligence. He points out that in particular philosophical
climates the belief that the world is thus transparent, or the goal of making it
thus transparent, may be much more appealing than in others. In particular in
the twentieth-century pragmatic climate that Quine inhabits, this belief is less
prominent: it becomes enough that theory should enable us to 'anticipate and
control perceivable events', and genuine intelligibility is no longer a first
priority.17 In modalizing we are being Incautious, and even if Craig finds much
17. Craig, 'Arithmetic and Fact', p. 92.
 Morals and Modals 67

to admire in the old ideology that prompted us to be so, the sense remains that
sobriety requires the more Quinean attitude. This side then thinks that the
Cautious Man is distinctive in not modalizing. Either he does not possess a set
of concepts that we, somewhat unaccountably, do, or he exercises proper
caution in not making judgements with them.
The other side, where I feel rather isolated, queries the central doctrine of
these thinkers. When we understand what the Cautious Man lacks, we shall
be pleased that we have it. The central doctrine of the other team is, in Craig's
words, that 'we should not infer any absolute impossibility from the limita-
tions of our own imaginations.18 With modifications, I suggest that there is a
quite proper move or inference here; that what looks like intellectual hubris is
in fact not so. The shared doctrine of the other team is that there is a chasm
which the Cautious Man is admirable for not crossing. My claim is that it is
only in the shadows cast by illicit hankering after a realistic, truth-conditional
account of modalizing that the crossing seems so dangerous.
Craig thinks that there might be two sources for the idea that the crossing
can be made. One is that meanings are sufficiently transparent to our minds,
that we can know just by introspection that what we mean by some sentence
can never come out false. As he rightly says, nobody can succumb to that with
a clear conscience these days. The other is the assumption that our mental
powers are perfectly in tune with reality, and as he again rightly says, that can
only be credible within a specific philosophical climate. My source is different:
I am sceptical about the assumption that we know what we mean by 'absolute
necessity', or the real distribution of possibilities, in a way that allows us to
contrast them wholesale with the blocks that our only ways of thinking meet. I
am sceptical because I detect the influence of realism at just this point.
This scepticism will, I hope, appear less extravagant if we remember the
other, easier, fields on which projectivism + quasi realism fought. The equiva-
lent of the Craig-Wright-Quine team over morals would say: 'we should not
infer any ["absolute"] obligations from the direction of our own sentiments'
(for example). The equivalent of the Cautious Man would be someone who,
while conducting his practical reasoning in every respect as the rest of us do,
eschews the 'inference' to the proposition that we have, for instance, an
obligation to our children. He can make the same deprecatory remarks about
our right to think ourselves in tune with metaphysical moral reality. He can
even cite theological and philosophical climates in which this pride would
have seemed natural, but which no longer obtain. My reaction is that he has
mistaken the nature of the judgement: by thinking of it as 'made true' by some
possibly alien state of affairs he has made his scepticism inevitable; by seeing
the proper function of the proposition we avoid it. On a realist account, his
caution is correct, as is his refusal to moralize. But as it is he is actually missing
nothing (as I put it in essay 8, 'shmoralizing'—conducting practical reasoning
properly without a realistic backdrop—is just moralizing). Again, the colour
case provides an easier but slightly more distant analogy: we would be wrong
18. 'Arithmetic and Fact', p. 110.
68 Metaphysics

to be cautious over whether using our eyes tunes us to the real divisions and
distributions of colours, because our only concept of the reality of those
divisions comes from proper use of our eyes.
However, the other team has another weapon, again wielded powerfully
by Craig. Following the passage agreeing with Wright that it is a 'tendentious
step' to inflate our imaginative limitations into a metaphysical discovery,
Craig writes:
It certainly is a further step. In the first place, it is clear that there is a group
of possibilities which no argument from premisses about what we can and
can't imagine could ever rule out. We might, for instance, come to be able to
imagine what we can't now imagine, there may be other beings who can
imagine what we can't and never will be able to imagine, and so on. . . . [I]f
we close our minds to these possibilities then we make assumptions about
our present imaginative capacities for which we have no warrant.19
To address this, we need to make distinctions within the class of the 'un-
imaginable'. I wrote above of propositions whose truth we cannot imagine in
the sense that we could make nothing of ways of thought in which they are
asserted. Now this is to be taken fairly strictly, and so taken it does not quite
correspond to 'unimaginable' on an untutored reading. Suppose, for instance,
I announce that I am able to show you a new primary colour, quite distinct
from any mixture or shade of previous colours. You may doubt me, and you
would certainly be unable to imagine what I was going to show you, if my
claim is true. You might even express yourself by saying that it is impossible,
but you would be unwise to have much confidence in this claim, for in some
sense you can 'make something of the possibility that I am going to do what I
said. It is not as if I had said I would show you a circle with straight sides, or a
true contradiction.
Let us distinguish a proposition's being 'unimaginable', in the sense that
we cannot present to ourselves a sense of what it is like to experience it as
true, from its being 'inconceivable', where this involves the kind of block
just indicated, in which we can do nothing with the thought of its truth. It is
frequently pointed out that unimaginability is a poor symptom of inconceiv-
ability, and this is correct. The cases one would adduce include these: the
extra colour, the existence of infinite totalities, the bounded and shaped
nature of space or time, the existence of extra dimensions, perhaps the
operation of backward causation. Then there is the unimaginability of en-
tities like the self, or of the will, and in some frames of mind, we cannot
imagine the possibility even of rule-following, intentionality, and so on. The
lack of fit works the other way round as well—propositions might be prop-
erly classed as impossible, although the imagination freely allows them:
notoriously, the alleged possibility that I might have been Napoleon, or that
Fermat's theorem might be true (or false), one of which is imaginable,
although impossible.
19. 'Arithmetic and Fact', p. 90.
 Morals and Modals 69

Our imaginative powers change and develop. The child cannot imagine
the beliefs of the adult; those unacquainted with them cannot imagine the
taste of claret or the work of Rembrandt. These conditions can be altered,
which immediately gives us a sense of potential ways in which our own imagi-
nations are partial. Our experience is limited, and our imaginations not much
better. Just as people of limited experience have impoverished imaginations
compared with us, so we must accept that there are many things of various
kinds which we cannot now imagine—tastes, smells, insights, and presumably
truths. This, of course, accords well with Craig's caution: it is not just a modal
sceptic, but all of us, who will beware of inferring impossibility from just any
imaginative failure.
Using unimaginability as a good indication of impossibility is also a mis-
take because it depends upon too simple a notion of the relation between
experience and thought. It asks, as it were, that we should be able to see any
truth in a single picture. So, for instance, if we want to think of a theoretical
notion, such as that of force acting at a distance, we try to visualize the
process, and, failing, are apt to find the notion suspicious. We find it hard to
accept that full intelligibility can be earned by a proper place in a theory, even
if we cannot visualize the happenings of the processes. Consider, for another
example, the shape of space. Children find it incredible to think that space has
a shape, because they try to visualize it, or in other words imagine themselves
looking at it, which is what we normally do to observe the shape of things, and
the thought experiment collapses, for the observer cannot find a standpoint
from which the whole of space can be observed. But using that failure as a
reason for concluding that space must be infinite would be a mistake, for it
would ignore other ways in which a shape of space might be certified—ways
like those a man might use to find the shape of a container in which he is
confined. If these procedures certify that only certain routes in space are
possible, then the right conclusion may be that space is bounded and has a
shape, and we can explain why the enterprise of trying to visualize it fails.
Visualizing is a poor guide to states of affairs, because not all states of affairs
reveal themselves in a picture. Similarly, things may be impossible although
naive imagination allows them, because naive imagination does not tell us
how to describe the scenes it recreates; this is why it is so dangerous to use
imagination as a guide to the metaphysics of the self.
Here we have explanations of failures of imagination. And we can con-
ceive of superior positions from which some of our imaginative limitations
could analogously be explained. When we can do that, we will not take
imaginative limitations as a guide to impossibility. Now Craig in particular
notices all this. This is a difference, he writes, between the case of the extra
colour or difficult intermediate cases like that of extra spatial dimensions, and
full-blown cases like that of a deviant arithmetic: 'An explanation of our
inability to imagine the arithmetically deviant along the lines that served for
colour and spatial dimensions doesn't get started; so nothing checks our ten-
dency to project our incapacity and suppose that reality just couldn't be like
70 Metaphysics

that'.20 But Craig does not highlight the good use the projectivist can make of
this difference.
Consider again the parallel with moral projectivism. We do not find
it trivial to cross from a sentiment to a moral judgement. Only certain
sentiments—those of a certain strength, or with certain objects, or those
accompanied by sentiments about others who do not share them—form a
jumping-off point. We are also conscious that there are doubtless flaws and
failures in our sentiments, which are perhaps capable of explanation in the
same way that we explain the defects of those who are worse than ourselves.
But when the sentiments are strong and nothing on the cards explains them by
the presence of defects, we go ahead and moralize. We may be aware that our
opinion is fallible, but that is because we can do something with the thought of
an improved perspective, even when we are fairly certain that one will not be
found, and here as elsewhere commitment can coexist with knowledge that we
may be wrong. The 'step' from a fully integrated sentiment of sufficient
strength to the moral expression now becomes no step at all: the moral is just
the vocabulary in which to express that state. Avoiding it would not be an
exercise in modesty, but an impoverishing idiosyncracy of expression.
Why should it not be like this with logical necessity? We have arrived at
the residual class of propositions of whose truth we can make nothing. We
cannot see our failure to make anything of them as the result of a contingent
limitation in our own experience, nor of a misapprehension making us think
that their truth should be open to display in a way in which it need not be. We
express ourselves by saying that they cannot be true—that their negations are
necessary. There is the bare possibility of being shown wrong—perhaps our
search into the causes of our imaginative block was inadequate, or perhaps we
were under a misapprehension of what it might be for the proposition to be
true. We may be uncomfortably aware of even great philosophers who mistak-
enly projected what turned out to be rectifiable limitations of imagination—
the a priori has a bad history. But as Wright notices, we should have no wish
to make ourselves infallible when deeming things a priori. We make the
commitment in the light of the best we can do. There is no step, and no
illusion.

VI. Naturalism and Quasi-Realism

On this account, part of what it is for us to make nothing of the truths that we
deem impossible is that we cannot explain naturalistically our own failure to
see what it would be for them to be true. When we can see how, if a proposi-
tion were true, we might nevertheless be in bad circumstances to appreciate
how it might be, we release it from impossibility. It does not deserve ruling out
any more. But we cannot see how, if contradictions were true or if 1 + 1 = 3,
we might be in bad circumstances to appreciate how it might be. We could
20. 'Arithmetic and Fact', p. 106.
 Morals and Modals 71

have not even a sketch of a natural story of the block we face, because we can
make nothing of the starting point.
This provides a kind of Catch 22 in our attempts to theorize about the
modal. If we can see our tendency to rule out p as the outcome of a contingent
limitation, we are already making something of the thought that p might be
true, but that if it were, nevertheless we would not appreciate it because of
something or other. And this undermines any original commitment to its
impossibility. When someone starts: 'if there were an extra colour then . . .'
perhaps we can understand how it might be contingent limitations that make
the hypothesis hard to contemplate—but if that is all there is to it, we lose any
right to regard it as impossible. On the other hand, when someone says 'if 1 +
1 = 3 then . . .' and essays to show how, if this were true, we would be in a
bad position to appreciate it, the thought experiment breaks down, for we
cannot properly work through what is being supposed and how we might be in
a world of which it is true. But this means that there is bound to be a residual
'surd': our incapacity to make anything of the thought that some propositions
are true has to be resistant to natural explanation, if it remains a good candi-
date for modal commitment.
The fear of an inexplicable core motivates attempts, such as the positivists
gave, to remove any content from necessary truths. But we have accepted that
the dirty-hands argument shows that we will not explain this incapacity by
invoking uncontaminated knowledge of meaning, concepts, rules. We now
find that if any natural explanation of our imaginative block can be given, this
attacks our right to make the commitment. I think that here we get an alterna-
tive, or perhaps supplementary, explanation to that offered by Craig, of the
late twentieth-century opposition to the modal. It can arise not only from a
changed conception of what theories need to do, but also from a conviction
that nothing escapes naturalistic explanation.
When we have thoroughly tested the sense of a hypothesis and make
nothing of it, this is, in Wright's words, how things are with us. As Craig says,
if the quasi-experiment of working through how it would be if p is done on
ourselves, now, and if our attempts to work with p being true fail, then 'for
any logical guarantee we have, that may be as far as it goes'.21 But it goes a
little further, for in the light of what we have said, it will also be so that we
cannot see the incapacity as just one we happen to be subject to; we cannot
deem it a mere fact about ourselves, here, now. If we could see it in that light,
then that itself would destroy the modal commitment. This is why there is
something bogus in Kant's theory that it is the forms of inner and outer sense
that determine our a priori commitments. This looks illuminating because it
looks sufficiently parallel to the natural explanation we might give of the
imaginative limitations we can accept as no indication of impossibility—the
colour limitation, for example. But it is not really parallel, for if we can make
nothing of the possibility of other forms of sense, the 'fact' that ours is one
way or another is not intelligible as a genuine explanatory truth. Seeing it like
21. 'Arithmetic and Fact', p. 91.
72 Metaphysics

that would require thinking the other side of the boundary: understanding
how it might be, for instance, that although it is compulsory for us to use
classical arithmetic, with a different cast of mind it might have been compul-
sory to use another arithmetic. And this we cannot do.
The residual surd marks a large asymmetry between the moral and the
modal. In the case of moralizing, nothing stands in the way of a complete
naturalistic story of what it is, why we do it, and, quasi-realistically, why we
are right to do it. But the genesis of the way of thought is similar. The moralist
insists upon obligations. He rules out those who flout them, refusing ap-
proval, ignoring contrary temptations, bending his actions to conform. The
modalist insists upon necessities. He rules out ways of thought that flout
them, refuses theories that involve them, bends his thoughts to conform. The
moralist could just issue rules and penalties, but if he becomes self-conscious
he needs the moral proposition to stand as a focus for discussion and reflec-
tion, and he contemplates its truth as a way of doing so. The self-conscious
modalist needs the same. But the moralist can be quite completely aware of
the genesis and justification of his activity, whereas if what we have just said is
true, the modalist cannot be. In the case of the modal, the phenomenon is
antinaturalistic at its core.
Or is this unduly pessimistic? Some relief might be got by teasing out more
aspects of the core inability to 'make anything of a way of thought that
accepts a putative impossibility. Obviously, there are enterprises of thinking
through what modifications in logic are possible or what would be missing in a
way of thought that consistently tried to make 1 + 1 = 3. The business, for
instance, of thinking through how a science might be built around denial of
double negation, or of the distributive laws of logic (from P and Q v R, infer
(P & Q) v (P & R)) proceeds under the stimulus of constructivism, or of
quantum mechanics, respectively. So it ought to be possible to hold both that
these laws are necessarily true and that we can 'make something of ways of
thought that lead people to deny them. This is not a serious obstacle to the
direction of this essay. What we do is take a proposed deviation and follow it
until either the way of thought seems possible—and we no longer modalize
against it—or it breaks down. But 'breaks down' will mean: offends against
something that we suppose essential to any scheme of thought (such as some
distinction of truth and falsity, some stability of content, some embargo on
contradiction). Eventually we voice an inability to make anything of transgres-
sion against these norms: this is the surd that remains. If the thought processes
of the deviants are eventually seen to break down, then we can get a deeper
understanding of our own commitments: it is no longer so that we face an
entirely blank wall when we try to explain our own attachment to these laws.
This reveals the genuine scope for explanatory work, and it may do a little to
moderate the antinaturalistic pessimism. We can certainly hope to show why a
way of thought that is committed (say) to noncontradiction, or to supposing
that not all propositions are true, or to other elementary necessities, is also
committed (say) to '1 + 1 = 2 ' , since we can hope to prove (relying, inevita-
bly, on moves that we find inescapable) that if they are necessary, then so is
 Morals and Modals 73

this. This would give a complete bill of health to the modal if, as the positivists
hoped, the propositions finally bearing the burden were free of genuine con-
tent, or owed their truth to some naturalistically explicable fact about us—a
decision or convention, for instance. But these escapes no longer appear, and
in default of a leap outside the system of necessities, the final surd seems set to
remain.

Addendum
In this essay I do not press an argument against Lewis's modal realism that I
did express in Spreading the Word; this argument nevertheless hovers in the
background. This argument is that, as well as problems of saying how we get
as far as possible worlds, the realist has a problem of getting us back from
them: when we use a counterfactual, for instance, in pursuit of a concern with
the actual world, why should we be interested if things are thus and so in a
neighbouring world, or in all neighbouring worlds? It sounds like a change of
subject. This argument was assailed by Bob Hale in his review (The Com-
pleat Projectivist', Philosophical Quarterly 1986). Hale in effect plays the
equation between 'this wire might have been live' and 'there is a possible
world in which this wire is live' backwards, pointing out that since we have
excellent reason to be interested in the former, and since according to the
modal realist the latter means the same, we have excellent reason to be
interested in the latter.
This mistakes the nature of the problem. My concern, as usual, was expla-
nation, and the point is that a realist construction of the neighbouring-
possible-world proposition plays absolutely no role in explaining why we
should be interested in the 'might have been' proposition with which it is
identified. If anything, it seems to make such an interest strange or even
inexplicable. It is no good replying that we are after all interested in the
'might' proposition, so we can expect the possible-world proposition to inherit
that interest: the point is that the interest is not explained, and becomes
harder to explain, if we give each of the claims other-worldly truth conditions.
There is an immaculate treatment of this by the late Ian McFetridge in the
collection of his papers, Logical Necessity (London: Aristotelian Society
Monographs, 1990), pp. 144-46. McFetridge also correctly breaks the alleged
parallel with Kripke's notorious argument against Lewis's counterpart theory.
Another puzzle with modal realism that I do not develop is that the realism
seems to take the modality out. 'Necessarily 2 + 2 = 4' and '2 + 2 = 4
everywhere' do not mean the same. But, says the realist, what if 'everywhere'
means 'in all possible worlds'? The question is ambiguous. If the collection of
all possible worlds were given extensionally (w1, w2 . . .), then again the iden-
tity would be lost: someone might think that 2 + 2 = 4 in all those worlds,
without thinking of 'all those worlds' as exhausting the possible worlds. If the
totality were given under some other heading than modality, the modal con-
tent would be lost. It is only if the collection is given under the heading of
74 Metaphysics

modality that the two mean the same, but we are not any further in under-
standing what it is to think of a set of worlds under that heading. This is no
objection to using possible-worlds talk, but it shows that the idea that when
we do so we refer to real things just like the actual world provides no explana-
tion of the nature of modal commitment.
It is natural to worry whether the use of the idea of an imaginative block is
a fig leaf, disguising what must ultimately be thought of in more conven-
tionalist terms, as for example adherence to a rule of language. In a way, and
for the purposes of this essay, I do not mind very much whether this is so (it
would matter much more to Craig, whose campaign has been directly con-
cerned with refuting conventionalism). In the last few lines of the essay I do
indeed express pessimism for the prospects of any theory of why we face the
blocks we do when we set about thinking in terms of impossibilities. But for
my purpose it is more important that this block is identified and properly
located as the source of our propensity to modalize.
 4
Opinions and Chances

I
Ramsey was one of the few philosophers who have fully appreciated the
fundamental picture of metaphysics that was originally sketched by Hume. In
this picture the world—that which makes proper judgement true or false—
impinges on the human mind. This, in turn, has various reactions: we form
habits of judgement and attitudes, and modify our theories, and perhaps do
other things. But then—and this is the crucial mechanism—the mind can
express such a reaction by 'spreading itself on the world'. That is, we regard
the world as richer or fuller through possessing properties and things that are
in fact mere projections of the mind's own reactions: there is no reason for the
world to contain a fact corresponding to any given projection. So the world,
on such a metaphysic, might be much thinner than common sense supposes it.
Evidently the picture invites us to frame a debate: how are we going to tell
where Hume's mechanism operates? Perhaps everywhere: drawing us to ideal-
ism, leaving the world entirely noumenal; or perhaps just somewhere; or
nowhere. Hume's most famous applications of his mechanism, to values and
causes, are extended by Ramsey to general propositions, which to him repre-
sented not judgements but projections of our habits of singular belief, and
also to judgements of probability, which are projections of our degrees of
confidence in singular beliefs.
If we are to assess his views we must be sure of what counts as an argument
for or against this projectivist picture. The main burden of my essay is that
most ways of framing the debate underestimate the resources available to the
projectivist. I think it is also clear, particularly from his 1929 paper 'General
Propositions and Causality', that Ramsey himself was optimistic about those
resources, in a way that has not been widely recognised.
The usual way of attacking the projectivist is this: He is saddled with a
particular view of the meaning of remarks made in the area in question. This
view is then shown not to correspond with some feature of the meaning that
we actually give to those remarks. It is triumphantly concluded that proj-
ectivism is inadequate, and that we must adopt a realistic theory, seeing the
remarks as straightforward descriptions of a part of the world that we are
(somehow) able to cognize. This kind of attack is clearly worthless unless it is

75
76 Metaphysics

clear that the projectivist is indeed committed to the theory of meaning attrib-
uted to him. Yet, I shall argue, the theory of meaning to be linked with
Hume's picture is variable, subtle, and obscure: if, as I suspect, it is as yet
unclear what resources the projectivist has in this matter, it follows that all
such attacks so far made are unsuccessful.
A couple of examples may help. It used to be thought that a subjective
theory of value entailed identifying the assertion that X is good with the
assertion that the speaker himself liked X. This is properly refuted by point-
ing out that the two have entirely different truth-conditions (or assent-
conditions), and the subjectivist is discomfited. But it is now widely recog-
nized that only a very naive subjective theory of value commits this error: a
theory of value as a projection of our attitudes can adopt a much better
account of what then is said by attributions of value—primarily in terms of
expression of such attitudes. Again, it used to be thought (perhaps it some-
times still is) that someone who, like Hume, thinks of the world as a succes-
sion of distinct events, and who accords no real distinct existence to necessary
connections between those events, must think that we mean no more than
regular succession when we talk of cause. But there is no reason for saying
that this is what we must mean when we project a certain habit of reliance on a
regularity, or some other attitude toward it, onto the world. Perhaps, for
example, we express some special attitude to the regularity or dignify it in a
certain way, and then many stock objections to regularity theories (factory
whistles blowing at the same time and so on) are entirely irrelevant. They
simply draw attention to regularities that, for some reason yet to be explored,
we do not dignify.1
But expressive theories of meaning are themselves attacked. It is probably
necessary at present to distinguish two kinds of rejection. One, which I shall be
coming to, joins issue over some particular aspect of meaning, such as the
occurrence of the disputed remarks in subordinate clauses, to which the theory
is supposed to be inadequate. The other is hostile to the whole idea of there
being a debate. I have in mind the conservative, pessimistic, and perhaps
Wittgensteinian view that we cannot do much with our language except speak
it, or at best put down rules for buildng up meanings by establishing rules that
govern the components of sentences with those meanings. But if that enter-
prise leaves us with such things as the placid truths that 'good' is satisfied by
good things, 'chance' refers to chance, and that A, B satisfies 'x causes y' when
A causes B, then we should rest and be thankful. Sometimes, indeed, it is felt
that the very endeavour to find semantic structure in a rule-governed way rules
out Hume's kind of theory by committing us to a correspondence theory of
truth, as if coherence theorists, or pragmatists, or the projectivists I am inter-
ested in, must half-wittedly deny that 'London' refers to London, and so on. In
fact, in constructing such a theory of semantic contribution we use our lan-
guage simply to describe itself, and leave perfectly untouched the question of
which metaphysics is appropriate to that use. It is puzzling to think why people
1. This is explored further in essay 5.
 Opinions and Chances 77

still associate the creation of a formal truth-theory for a language with particu-
lar views of truth. 2 To exorcise this temptation, imagine a formal theory of
Arabic numerals enabling us to deduce which number a sequence of digits
refers to, given axioms saying to what individual digits refer; the one essential
rule merely captures the way the position of a digit indicates the power of ten
by which it is multiplied. Does such a theory, or the interest of creating it, tell
us what it is to refer to a number, or commit us to a correspondence theory of
such a thing? Of course not. It is entirely silent on the issue, and merely uses
the notion, while telling us nothing about it.3
I think the argument to the contrary that confuses people goes: a corre-
spondence theory of truth needs to identify some fundamental word-to-world
relations; Tarski's style of theory can be taken to offer the relations of refer-
ence and satisfaction for this job; hence Tarski's style of theory is, or at least
helps, a correspondence theory. This would be fine if there were independent
argument that what it is for words to refer to things, or things to satisfy
predicates, is well thought of in terms of correspondence. But it might just as
well be thought of in terms of the predicates taking part in a coherent system;
or being used in promoting certain ends; or in terms of the things having the
reactions of the mind projected onto them. In other words, we might just as
well argue that a coherence, pragmatist, or subjective theory needs to identify
some fundamental word-to-world relations; hence Tarski's style of theory
helps them too. This shows its irrelevance to this issue. Yet, although it gains
nothing from truth-theories, we cannot dismiss the view that Hume's theory is
not debatable. The difficulty that perplexes me is that if, as I shall suggest, the
projectivist can make perfect sense of apparently realistic practice, it is not
clear what intellectual quirks mark him off, nor what is left to fight over
except harmless images and metaphors. The interest, at any rate, comes in
seeing what he can do by way of incorporating apparently realistic practice:
this is a programme that can be called 'quasi-realism', and I see Ramsey as
one if its patrons.
Specific charges against projectivist theories will concentrate upon ways in
which our thinking about the area in question appears to accord an objective
or independent standing to the things allegedly projected. Primarily, chances,
laws, and causes (not to mention values and goods) are all things about which,
we say, we can be ignorant. Our opinions about them can be wrong, defective,
in various ways. We allow the possibility that we think of them as existing
when they really don't and that we are unaware that they exist when they
really do. We acknowledge experts, so that some person's views of, say,
probabilities, become authoritative enough to count as knowledge and to
enter into books as physical constants like values of masses and densities (and
other people's opinions are often not worth a straw). Yet even the experts
might be wrong: it is not their opinion that defines laws and chances; the laws
2. M. Platts, Ways of Meaning (London: Routledge & Kegan Paul, 1979).
3. This issue was also joined in Spreading the Word, chapters 7 and 8. The idea that truth
theories dominate metaphysics is not nearly so popular now as when this was written.
78 Metaphysics

and chances would have been what they are regardless of whether people had
known about them.
Ramsey is usually thought of as one of the fathers of a 'subjective' theory
of probability that denies or at best struggles with such facts. On that theory, a
distribution of confidence across any totality of propositions is coherent if it
satisfies some very weak constraints. But those constraints allow for the most
bizarre confidences and agnosticisms. Yet coherence is all that there is. As
Kyburg and Smokier put it in the introduction to their 1964 collection, for
subjectivism any degree of belief in any statement is permissible, but there are
restrictions placed on the distribution of degrees of belief among sets of
related statements. 4 Since there is nothing to be wrong about, the view has
been summarized as claiming that 'sincerity is enough'. On the more modern
version, which I discuss in part III, it is also mandatory to stick by opinions
through time and as various kinds of observation are made. However, even
this gives us no title to say that a man who announces a quite outrageous set of
confidences is 'wrong'; the only vice he could display would be a kind of
fickleness as time goes by. It is easy to see why this implausible theory is
fostered onto a projectivist. In probability, as in the theory of value, if projec-
tion is all there is, there is surely nothing to be wrong about. But all this flies in
the face of the objectivity of our usage and renders the theory an easy prey to
criticism.
But Ramsey was well aware of the shortcomings of a purely subjective
theory of laws and chances. He explicitly denies, for instance, that chances
correspond to anyone's actual degrees of belief;5 he knows that we believe in
unknown laws (and he would have said the same about chances);6 he knows
that some opinions about chances are much better than others. 7 His effort is to
show that these phenomena do not refute an antirealist, projectivist, theory of
chances and laws, but actually are explicable given such a theory. It is the fact
that he made this quasi-realist attempt that seems to me to show that Ramsey
was much better aware of the resources of projectivism than many of his
apparent followers (Carnap8 being an honourable exception). How far can his
programme succeed?

II

Hume forged the essential tool for the projectivist to use as he attempts to
reconcile his theory with the objectivity of usage. In his great essay 'On the
4. H. E. Kyburg and H. E. Smokier, eds., Studies in Subjective Probability (New York:
Routledge & Kegan Paul, 1964), p. 7.
5. F. P. Famsey, Foundations of Mathematics, ed. R. B. Braithwaite (London: Routledge &
Kegan Paul, 1931), p. 206.
6. F. P. Famsey, Foundations, ed. D. H. Mellor (London: Routledge & Kegan Paul, 1978),
pp. 139, 150.
7. Ibid., p. 95.
8. R. Carnap, Logical Foundations of Probability, 2nd. ed. (Chicago: Chicago University
Press, 1962), p. 16 ff.
 Opinions and Chances 79

Standard of Taste', he points out, in effect, that it is no part of a projectivist

metaphysic to claim that one projection is as good as another. Some may be
inferior, some superior, and even the best may, in principle, be capable of
improvement. Thus, let us take the difficult case of moral evaluations. If
values are projections of a habit of forming some kind of attitude to some
kinds of thing, how can I be aware that my own attitudes might be defective,
and capable of improvement? (If we prefer societies to individuals, the ques-
tion becomes: 'how can we be aware that ours are'?) The answer is that I know
that people are capable of habits of projection which from my own standpoint
are deplorable: they judge things of which they are ignorant, and their views
are the function of fears and fantasies, blind traditions, prejudice, and so on.
But then who am I to be sure that I am free of these defects? This thought is
quite sufficient to enable me to understand the possibility of my attitudes
improving. They ought to be formed from qualities I admire—the proper use
of knowledge, real capacity for sympathy, and so on. If they are not, and if the
use of those capacities and the avoidance of the inferior determinants of
opinion would lead me to change, then the resulting attitudes would be not
only different, but better. It is true that in saying this I am presupposing one
kind of evaluation in giving sense to the possible deficiencies of the other. An
attitude to the processes of attitude formation is used to give sense to the
possibility not merely of change but of improvements in moral judgement. But
this gives nothing an axiomatic status: at the end of a process of reevaluation,
everything may have changed. The right analogy is with the rebuilding of
Neurath's boat, and we know that in principle the result of that might be an
improved boat. Equally we can understand and fear the possibility of deteriora-
tion. It follows that a projectivist picture of values need have little to do with
the frivolities of traditional moral subjectivism ('one opinion is as good as
another' and so forth). By pursuing the point we might begin to see how a
projectivist can incorporate notions of truth and knowledge.
In the case of empirical judgements of chance the matter is much easier in
two respects. The first is not my main concern in this paper, but it is worth
noticing. Projectivism in moral philosophy is open to attack on the grounds
that the reaction of the mind that is supposedly projected is itself only identifi-
able as a reaction to a cognized moral feature of the world. The specific
attitudes and emotions (approval, indignation, guilt, and so on) can, it is
argued, be understood only in terms of perception of right and wrong, obliga-
tions, rights, etc., which therefore cannot be reflections of them. Myself, I do
not think that this is true, nor do I think, if it were true, that it would refute
projectivism. For it is not surprising that our best vocabulary for identifying
the reaction should be the familiar one using the predicates we apply to the
world we have spread. Thus, to take a parallel, many people would favour a
projectivist view of the comic, and they may well be right even if our best way
of describing the reaction which we are projecting onto a situation we describe
as comic is 'that reaction we have when we find something funny'. I don't
think a behaviourist analysis is either required or helpful, for obviously the
behaviour, to someone with no sense of humour, would be incomprehensible.
80 Metaphysics

In any case, a projectivist theory of probability meets no such objection.

For it is easy to identify the main component projected when we attribute a
good chance to an event or a high probability to a judgement: it is, of course,
simply a degree of confidence.
Degrees of confidence in propositions are 'intervening variables' in psycho-
logical theory. We can know about them through interpreting the behaviour
that, supposedly, they explain. The measurement of degrees of confidence is
not necessarily straightforward, but this is no problem. It may even be indeter-
minate, very often, what a person's degree of confidence in a proposition
actually is, or whether indeed he has one. But in the same way it is often not
straightforward to know what a person's belief is; it may be indeterminate
what his belief about some matter is, and even indeterminate whether he has
one. Yet the notion of a belief is a proper theoretical concept in psychological
explanation. Here we should notice that a projectivist needs no more degrees
of confidence than a person has beliefs about chances; it is no part of his view
that, for instance, a real number should be in principle assigned to every
proposition an agent has ever thought of, representing his degree of belief in
it. We need no such extravagance: often we express ourselves by saying that
we have no idea what the chance is—and this attitude need not co-exist with a
particular confidence in a proposition. We shall see how to interpret it on a
projectivist picture later. If the whole notion of a degree of confidence were
suspect, as some authors claim, then we would need at least to indicate
another projected psychological state. I am inclined to suggest that it would
not be fatal if, as in the moral or even the causal example, our success in doing
this without using the vocabulary of chance were only partial. But, in fact, I
doubt whether it would be too difficult.
The second respect in which things are easier for probability is the backing
we can give to our standards for evaluating projections. We need standards for
assessing projections of degrees of confidence, enabling us to say that some
are better than others, that even the best may be capable of improvement,
that some are worthless. But there are obvious sources for such standards: an
opinion might be formed in the light of experience of observed frequencies or
fit into an otherwise successful scheme of projections, and most fundamen-
tally it might give its possessor the habit of belief in what happens and dis-
belief in what does not. And it is this which is the lynchpin of Ramsey's
theory. We could say that it imports a pragmatic standard for evaluating
projections, but this might be misleading. For it is not as though the standard
is in any way optional or avoidable if we adopt different goals or purposes. It
is necessary that truth count as success in judgement, and that the proportion
of successes achieved by a habit of making judgements be a measure of the
confidence that ought to be felt in the beliefs to which that habit leads one.
The standard is mandatory.
For suppose we have a thin, Humean view of the world. What is our
purpose in projecting onto it chances and probabilities? Ramsey writes that
we 'judge mental habits by whether they work, i.e. whether the opinions they
lead to are for the most part true, or more often true than those which
 Opinions and Chances 81

alternative habits would lead to'. The opinions he is talking about are of
course not opinions about probabilities, for that would get nowhere, but the
opinions of particular matters of fact that judgements of probability will lead
us to form. Fortunately the world displays patterns allowing us to have success-
ful habits of particular belief: faced with partial or complete regularities we
can form partial or complete confidence in new cases, and the world grants us
success if we are careful. As Ramsey writes, the best habit of belief formation
will have us forming confidence of a strength proportionate to the ratio of
particular truths to falsities that the habit leads us to believe in. (For why, see
section IV.) So it seems that Ramsey is going beyond mere coherence of sets
of belief, in a thoroughly sensible and necessary way. That standard is too
permissive, since, on the face of it, a set of beliefs may possess the virtue of
coherence while having the disadvantage of enjoining confidence that things
happen which never do, or that things don't happen which often do. Pragma-
tism must supplement coherence.
But this charge ignores the work that has recently been done on the
relation between subjectivism and 'learning from experience'. This work
makes it plausible to believe that a subjective theory of probability, relying
only on the constraint of coherence, can show that the process called condi-
tionalization is obligatory. It may then seem as though conditionalizing is
itself a process that forces opinions to converge, and that what they converge
upon is a value for probabilities in accord with observed frequencies. If all
this is true, then the constraint of coherence would be sufficient to give the
standard of evaluation that Ramsey wants. In the next section I assess this
argument, and in the last I go on to develop a projective theory in more
detail.

III

In the succeeding sections I shall use the standard terminology, in which a

chance setup exists, and there is an actual or hypothetical series of trials that
yield a stable frequency of various outcomes. The set of trials we can call A,
and the outcome in which we are interested, B. There is a slight strain in
adapting this terminology to, say, the chance of a person being a gin drinker,
but for the moment we are not interested in any problems caused by extend-
ing the notions.
We now suppose that we have conducted a reasonably extended investiga-
tion, and in a large number of trials the proportion of 5s is tending to stabilize
around some figure, p. Our problem is to give a projectivist account of the
natural judgement that such evidence (in the absence of other evidence)
would lead to 'the chance of an A being B is p'. But this way of putting it blurs
a vital distinction. There are two sorts of chance judgement that could be
made. One is local, and concerns only the chance of one of the trials being B
in the set that makes our evidence. The other is not: it is a judgement that
concerns other trials and has implications for our confidence in future cases. It
82 Metaphysics

is one thing to say 'the chance was p of one of the examples of A we have
considered having been B'. It is quite another to say 'the chance is p of any A,
including ones yet to be realized, being B\ The first judgement is local or
restricted to the class of trials in which we already know the frequency. The
second involves a prediction, or more accurately, a commitment to a particu-
lar kind of confidence in situations not yet brought about. It involves an
apparently inductive step. Yet, although it will evidently be more complex to
identify the thoughts that license it, I shall consider it first.
It is not surprising that a man observing and recording results from a
process that generates a certain frequency of outcomes B among events A,
with no discernible pattern, should come to have a degree of confidence,
proportionate to the frequency of Bs among .As, that an arbitrary A, such as
the next one, will be B. But why should he? The simplest answer has two
stages. First, the man has the inductive habit: he expects the process to go on
generating roughly the frequency it has done so far. Second, given that this is
so, he will be right to have the degree of confidence in a particular outcome
identical with its relative frequency, because that is the standard for rational-
ity. There is nothing mysterious about this second point. If the inductive
expectation is right, then the relative frequency remains stable. If that is so,
then a habit of adopting and acting upon any other degree of confidence in
particular expectations would lay you open to certain loss if you are required
to act out your confidence by buying or selling bets at the corresponding rate,
given that your partner is someone more straightforward. It would be like
having confidence other than 0.25 that a card from a shuffled pack is a heart: a
hopeless position if you are required to buy or sell bets at a corresponding
rate. Nor should we worry that there is anything unrealistic (undemocratic, as
it were) in criticizing someone for having a set of confidences that would lead
to loss if he were required to post odds on which he could be required to buy or
sell bets: it is not a satisfactory defence to reply that we are not often required
to gamble. The defective degrees of confidence are like bad dispositions that
may nevertheless remain unrealized. They can still be criticized by pointing
out what would happen if they were to be acted upon.
The two-stage answer relies on induction, and the rationality of that is left
dangling. It is tempting, therefore, to hope that work on conditionalizing
achieves an answer without relying on specifically inductive habits. It is not,
indeed, likely that this could be so, since induction appears to be a necessary
component of any answer, in that if there were no reason for expecting the
process to generate the same frequency as hitherto there would indeed be no
reason for expecting the next A to be a B with any particular confidence. So
unless work on conditionalization provided some justification of induction, it
could not provide the requisite standard.
The central argument in this area is credited variously to David Lewis and
Patrick Suppes. It is, in effect, an extension of the standard Dutch book
argument for coherence to an agent's probability distributions through time.
The standard argument makes coherence at any one time a necessary condi-
tion of a rational distribution of confidence. The new argument, which we can
 Opinions and Chances 83

call the dynamic Dutch book argument, or DDB, extends this to prevent a
rational agent from wiping his slate clean at any time and forming whatever
new confidences he fancies (although until Ian Hacking first made it clear, it
was not widely recognized that one might do this, so that subjectivists had
happily helped themselves to conditionalization anyhow).9 The new argument
is designed to prove the connection between rationality and conditionaliza-
tion. Following Paul Teller we can see the DDB like this.10 We imagine an
agent with a set of beliefs at time 0, described by a function P0 giving the
confidence with which every proposition in the domain is believed. A change
in belief in A is described as conditionalized upon evidence E if, at time n,
after E becomes known (so that Pn(E) = 1), Pn(A) is equal to P0(A & E)/
P0(E). That is, at time n the new confidence in A is equal to the old condi-
tional probability of A upon E. As Teller shows, we can generalize everything
to the case where E merely changes probability, but this does not matter.
Now it is quite clear that, sometimes, changes of belief that are not condi-
tionalizations are legitimate. One may rethink a problem afresh, and come to
regret one's old confidences. One may think up new alternatives. But we can
avoid objections based on this by restricting ourselves to cases where nothing
of this sort occurs, but where someone has in advance, at time 0, a settled
policy or habit of not conditionalizing. In other words, one has a policy or
habit that should E come about, will lead him to some confidence in A
greater or less than his present confidence of A upon E. The DDB shows that
a man known to have such a plan, and required to buy and sell bets according
to his confidences, can be made to buy and sell bets on which he has a net
loss, whatever happens, by an opponent who knows no more than him (ex-
cept, perhaps, that he has the habit or policy). The general proof is complex,
but its principle is quite easy to grasp. Suppose I am following a plan, or have
a habit, which means that I now have a large confidence that A will occur if E
does, but which enjoins that if E does occur I will only have small confidence
in A. Suppose I think there is a 60 percent chance that John will play a spade
and a 90 percent chance that if he does so he will be left with a court card.
But, flouting conditionalization, I am settled that I will have only 30 percent
confidence that John will have a court card after he has played a spade. The
nub of your strategy for profit is this. You sell to me a bet to yield (say) 1 if he
plays a spade and has a court card, and 0 otherwise. I will pay a relatively
large amount for that (0.54, in fact). You plan that if he plays a spade you will
cover that bet by buying from me a bet to yield you 1 if he has a court card, 0
otherwise, and you know you will be able to do that cheaply, for since I will
then have little confidence in the court card, I will want little for such a bet
(0.3, in fact). You then only need to arrange side bets to give you a modest
profit if he does not play a spade, and you will profit in any event. If you can
sell me bets when I am confident, and buy when I am less so, you profit.
Conversely, if I had announced that although now I regard it as only 30
9. I. Hacking, 'Slightly More Realistic Personal Probability', Philosophy of Science 1967,
pp. 311-25.
10. P. Teller, 'Conditionalization and Observation', Synthe.se 1973, pp. 218-58.
84 Metaphysics

percent probable that if he plays a spade he will have a court card, yet, I
agree, if he does play a spade I will be very confident that he has a court card
(I know I get excited). You buy from me a bet on both things now, and sell to
me a bet on the court card later, if he plays a spade. Again, arranging side
bets to cover him not playing a spade, you profit whatever happens. I ask
little now for the first bet, and am prepared to pay a lot later for the bet which
covers it.
It is not quite right to say that this gives an effective method of profiting
from a nonconditionalizing agent. The direction of his departure from the
present value of conditional probabilities (the probability of A upon E, now)
must be known. If we know that he will inflate his confidence in A we can
profit, and if we know that he will deflate it we can profit, but it does not
follow, and it is not true, that if we know that he will do one or the other we
can profit. We have to know the direction of his aberration before we know
whether to buy or sell bets, but of course a general tendency in an agent one
way or another could also be exploited over time.
What does this argument show? It shows that an agent known to plan a
definite confidence in some proposition receives certain evidence that is
either higher or lower than the value he now attaches to the conditional
probability of the proposition on the evidence, can be made to lose, what-
ever happens, if he is required to act out those confidences. Let us agree
that such a plan is irrational. Does it follow that we should expect rational
confidence to converge upon frequencies, thereby bypassing the apparently
inductive step? The feeling that it may have something to do with it comes
like this. Suppose we antecedently hand people a number of hypotheses
about the chance of an A being a B, and invite them to form a distribution
of confidence among them avoiding the pathological values 0 or 1. We then
amass frequencies, and since the agents must conditionalize, the posterior
probabilities gradually increase for hypotheses giving the chance a value
near the observed frequency, and fall away for the others. Eventually opin-
ion converges upon a high probability for the chance being as near as possi-
ble to the observed frequency. And the value given the chance dictates our
confidence in the next A being a B.
As an attempt to either bypass or cast light upon the inductive step, this
argument clearly fails. It falls to a dilemma: either the original hypotheses are
consistent with changes in chances over time, or they are not but relate solely
to the trials already conducted. They are local in my sense of the term. If they
are, then whatever our confidence that the chance of an A being a B took a
certain value in generating the frequencies we have so far observed, we need
inductive confidence to transpose that confidence to the future. It is no easier
to argue that since the chances of an A being a B have always been good, they
will continue to be, than it is to argue that since nature has always been
ordered, it will continue to be so. On the other hand, if the original hypothe-
ses describe eternal chances, so that in accepting a hypothesis concerning
chance I would indeed be committing myself to a uniformity—to the probabil-
ity that a stable observed frequency of As among 6s can be extended indefi-
 Opinions and Chances 85

nitely with a similar value—then the induction is presupposed in setting up the

request; it is only if we have an inductive faith in such uniformities that we
should be inclined to distribute confidence over the initial finite selection of
hypotheses. Otherwise we should simply point out the many other things that
nature might do instead of giving a uniform chance for an A being a B.
This is easily seen if we imagine a concourse of souls in an antecedent
heaven, each of whom is handed a ticket describing a different course that
nature might take in the world they are about to enter. Some give eternal
constancy to the relation between As and Bs, some see it as altering over time,
so that although up until a certain time a certain degree of confidence in an A
being a B accords with frequencies, after that time a different one does. These
souls may dream up conditional probabilities for themselves, giving the world
an x percent chance of continuing to conform to their ticket if it does so until a
certain time. But unless they have an a priori reason for expecting the world
they are to enter to favour uniformities, there is no reason for x to be different
for those with straight tickets and for those with bent ones. And if all possible
hypotheses are ticketed, then x must be vanishingly small, corresponding to
the fact that at any time there will be an indefinite number of tickets conform-
ing to the world up until that time, but subsequently divergent. Of course, I
am not here denying that we may be able to think of a reason why x should be
placed higher for those with straight tickets: that is solving the problem of
induction. But it is quite clear that if this can be done, it is not by simply
proving the virtue of conditionalizing. For that is something which the bent
can do with the straight. In a nutshell, conditionalizing appears interesting
only if we pose the problem in a way that presupposes inductive good sense.
(Philosophers of a sociological and Popperean bent are liable to point out that
only hypotheses about stable chances would be of interest to scientists, who
would regard the others as crazy. This is true. It is true because scientists, like
the rest of us, possess inductive good sense.)
It appears, then, that hypotheses about chances that carry implications for
future distributions of confidence are not automatically the outcome of condi-
tionalizing changes of opinion. Induction is needed, as indeed we might have
expected. Furthermore, we can escape the unrealistic element of seeing learn-
ing from experience in terms of conditionalizing: namely, the nebulous nature
of the prior distributions of confidence needed. It is usually much more natu-
ral to see our experience as putting us in mind of some hypothesis about
chance, rather than merely modifying the degree of confidence with which we
used to hold one. The rational man does not have to spring from the womb
fully armed with an infantile probability distribution across all the hypotheses
that experience teaches him to believe. But if we now turn to purely local
assessments of chance, things may appear easier. Here we merely want to say
that the chance of an A being a B on the trials we have conducted was p.
Whether it remains so can be, so far as that judgement goes, entirely up to the
gods, and depending on our opinion of them we can expect what we like about
the next A. What is the rationale for such local judgements?
The connection between judgements of chance and confidence that we
86 Metaphysics

have relied upon has been quite simple. We have imagined judgements of
chance dictating a corresponding degree of confidence: the reason why that
must conform to relative frequency is evident if we demand that the confi-
dence could be acted upon in willingness to buy and sell bets. If we imagine a
closed set, like a pack of cards, any confidence in an outcome (e.g. a card
being a heart) with a known frequency must conform to that frequency, for
otherwise, if the confidence were acted out, loss would be certain. So it might
seem as though everything ought to be very simple for local judgements of
chance. We know the frequency of As in our set being Bs; the right confidence
to have that an arbitrary member of the set is a B must conform to that
frequency, so the right chance judgement must be one that expresses that
degree of confidence, that is, the judgement that the chance is identical with
the frequency. But the trouble is that although this may be correct in the sense
that, if we knew nothing else, it would be the proper estimate of the chance, it
does not follow that we regard it as true.
For there is actually no compulsion on us to identify local chances with
local frequencies: we all know that there is a chance that in any finite set of
trials the frequency with which an outcome actually occurred differs markedly
from the chance of that outcome occurring. The chance of getting a B might
have been q, even if the obtained relative frequency of Bs was p. Here realism
seems to triumph: what kind of account can a projectivist give of this modal
claim? Equally we may persist in actually believing that the As in our set had a
chance q of being B, although we know that the relative frequency was p, so if
we were told to bet on whether an arbitrary A had been a B, p would be the
right figure to act upon. Again, what account can a projectivist give of the
belief about chance, when, as in this case, it appears to diverge from the right
degree of confidence to have about arbitrary members of the set?
The answer can come only from seeing the parts that induction and science
play. If we think the chance of an A being a B was actually q, we think that q
would have been the right betting rate antecedently to the set of trials. We can
believe this because we can believe that there was or could have been a longer
set of trials in which the proportion of Bs tended to stabilize on the correct
figure, q, and we believe that none of the things that affect such frequencies
was different on our actual trials. This last is a scientific belief, in the sense
that our causal theories of the world are what tell us whether particular factors
that do influence the frequency of 5s were present. Of course, if the figure p
arose from a sufficiently long series, this in itself will be evidence that such a
factor was present, whatever it may have been. But the point is that we may
not be forced to think that, and it may be easier to believe the reverse. The
judgement of local chance, when it diverges from actual frequency, is then an
expression of the confidence that should be felt in a hypothetical situation: a
situation where none of the things that, we believe, affect frequencies of Bs
among As would be different from those which obtained on our trials. Simi-
larly the modal claim expresses our fear that the hypothetical series might
exhibit the different frequency q, and our actual set of trials may be a very
poor indication of it. However, there is no reason to be depressed by these
 Opinions and Chances 87

possibilities, for induction works, and what produces a frequency of outcomes

one day is very likely to do substantially the same the next.
It seems to me that this is an account of our apparently realistic talk (in this
case, modal claims and claims apparently divorcing local judgements from
expressions of confidence) that yet concedes nothing to realism. Chances have
not entered as real facts, capable of explaining or causing events. They remain
projections, even if we are interested in spreading them not only over the
actual events that have confronted us or that will confront us, but over events
that could have happened as well.
In my last section I want to take up another problem that may tempt us to
realism: the problem of our willingness to talk about knowledge of chance,
and our subtlety in so doing.

IV
Let us suppose that the best evidence is that the frequency of 5s among As
indeed approximates to p, and will keep doing so. Clearly, then, a man with
the degree of confidence p has the habit of singular belief that meets Ramsey's
standard: he has a degree of confidence proportionate to the number of times
he is right. But there is another standard needed. For it does not follow that
we should endorse this confidence, nor this judgement of chance. Consider
that we may be able effectively to divide the As into two classes and rightly
predict a high ratio of Bs among the CAs and a low ratio among the others. I
shall call this 'effecting a partition of the As'. If this were done we would be in
a better position than him in this very straightforward sense. Using our knowl-
edge we can gamble with him and consistently win, by buying bets at a rate
corresponding to p on CAs being B, and selling on ~ CAs. And in a variety of
less mercenary ways we can see that our habit is more useful than his. It is
more accurate, more efficient. It is not necessarily incorrect to say that the
chance of an A being a B is p, even when effective partitions exist. We can,
after all, talk of the chance of an animal being a carnivore, or of a human baby
being Chinese. But if an acceptable practical application is to be made of such
a remark, then the context must be one in which the participants themselves
cannot easily put the subject matter into one or the other partition. If they
could, then there would be something deficient about the distribution of
confidence: it could effectively be improved. We might express this by saying
that there is no case of an A, neither those which are C nor those which are
not, on which we should accept such an estimate of the chance of it being a B.
But if an acceptable single-case judgement is to follow from such a remark,
then the conversational context will be one in which the participants cannot
easily put the subject into one or other of the partitions. It remains true that if
this can be done, the judgement of chance is deficient. (Of course, there is no
implication here about negligence or otherwise in being ignorant about C.)
We here have the beginnings of a reasonably clear view of two topics that
sometimes perplex analyses of probability: the rationality of seeking the nar-
88 Metaphysics

rowest reference class when we want single-case judgements, and the propri-
ety of restricting the terminology of chance to phenomena which satisfy von
Mises's second condition for an empirical collective, namely that there should
be no effective selection procedure for singling out a subset of the members
with a different overall frequency of the relevant property. (This is felicitously
called 'the requirement of excluded gambling systems'.) To take the second
issue first, if we believe that a selection procedure (corresponding to the
property C in the last paragraphs) can easily be found, then it follows that we
believe that our distribution of confidence can easily be improved, by the
standards we have seen. While we believe such a thing we are obviously in a
deficient position, and, depending on the consequences of a judgement and
the degree of ease of improving our knowledge, we may wish to attach no
weight to our judgement—to suspend it, in fact. The projectivist thus has
excellent pragmatic reasons for confining our judgements of chance of an A
being a B to reference classes for which we do not expect an easy method of
partition. There is no metaphysics of randomness required. It is just that if we
cannot effect a partition, a judgement of chance leads to our best possible
habit. If we can, it does not.
The single-case problem is in effect the same. If you want to have a degree
of confidence in the judgement of a particular philosopher that he drinks gin,
the statistic of the proportion of men who drink gin helps. But that statistic
can easily be refined: middle-class people, academics, academics in arts sub-
jects, of a certain age . . . The pragmatic motive for seeking the narrowest
reference class is just the same as before: a degree of confidence based solely
on the wider statistic leaves its possessor in an inferior position vis-a-vis some-
one who can partition the class of men and attach different degrees of confi-
dence to singular judgements depending upon the subclass in which the sub-
ject is found. Again, of course, the extent to which it is worth seeking statistics
for narrower reference classes depends upon the expected benefit of the more
discriminating judgement and the expected homogeneity of the reference
class. If nothing much hangs on it and if our prior judgement is that gin
drinking is not likely to be much different in any subsets we can think of, it
may not be worth a research programme to find out. Nevertheless the point
remains that unless we make our evidence as weighty, in Keynes' sense, as we
can, by considering possible partitions of our class, we cannot be sure that we
are properly serving the purposes of judgement, and the right think to do may
be to form no opinion. A good example of such reticence comes from legal
suspicion of a 'mere' statistic putting a defendant into a class with a high
frequency of guilt. Unless the class is as weighty as it can be made, it would be
impermissible to be confident that the defendant is guilty.11
It seems to me that this pragmatic perspective on these issues has a clear
advantage over a realist metaphysics. Notoriously, in trying to give sense to
the single-case judgement, a realist metaphysic of chance becomes tangled in
11. L. J. Cohen, The Probable and the Provable (Oxford: Clarendon Press, 1977), esp.
chapter 7.
 Opinions and Chances 89

the issue of determinism. If it is determined now that I will, or will not, catch a
cold next winter, it is hard to see what sense to make of talk of my propensity
to do so or not, just as, if I did not die before I was thirty, it is very unclear
what could be meant by saying that I had a propensity to do so. Yet I had a
chance of doing so. Equally, only if a chance setup is indeterministic will it be,
for a realist, true that trials on it form a collective. Otherwise, it is in principle
possible to select trials with a different long-run frequency of outcomes. For
the projectivist this becomes simply irrelevant. Our purpose in making judge-
ments of chance, and the standards to use in following out those purposes, are
perfectly indifferent to whatever secret springs and principles lie behind the
empirical collectives that form the subjects of those judgements. When our
best judgement is that we should treat the trials on some setup as forming a
collective, we are entitled to project a chance and form confidence accord-
ingly; when we suspect that we can find a partition, we should not. But there
need have been nothing wrong with us if we have treated something as a
collective, but at some improved state of knowledge a partition is found. Not
all ignorance is culpable.
The realist is apt to be impatient with such laissez-aller attitudes. True, he
will say, you can tell us when we talk of empirical chance, and perhaps you can
give some pragmatic understanding of why we do. But for all that, we may be
wrong to do so in some cases even when we are not culpable. Warranted
assertibility is not truth. A later discovery that a trial on a chance setup A fell
into a subclass with a different ratio of 5s from that shown overall would show
that we had spoken falsely; but what made our remark false was not the
discovery but the fact about the trial. Unless indeterminism is true, there are
always such facts, in principle, forcing falsity on all attributions of chance
other than 0 or 1. This argument is all the more persuasive because it cannot
be avoided by mentioning warranted assertibility in the long run, the usual
pragmatist substitute for truth. Once we admit that there are facts determin-
ing whether a particular outcome will occur or not, we cannot very well claim
that a long run of improving investigation would not find them; and in that
case determinism will entail that all chances of particular events are 0 or 1,
even on this definition.
But it is the definition that is at fault. A projectivist need have no use for
truth, about the chance of a single case, as the limit of the degree of confi-
dence to which progressive omniscience would tend. He does need a proper
account of fallibility, enabling him to admit that a particular estimate of
chance might in principle be improved. This we have provided him, without
any involvement in the metaphysics of determinism. This means that along
with a judgement he has the concept of a standpoint yielding a possibly
improved judgement. It does not mean that he needs the concept of a stand-
point from which all other judgement is seen to be wrong, and this is what the
unnecessary notion of the limit is attempting to import.12
In practice this means that although we have plenty of use for warranted,
12. See also essays 1 and 2.
90 Metaphysics
defensible, careful, estimates of chances of particular events, we have less use
for claims that we know such chances. The claim to knowledge entails, I
think, the claim that no improved standpoint, yielding a revised estimate, is
possible. To know something is to know that no judgement contradicting one's
own could be really preferable. To know that the chance of an outcome on a
single trial is p we would need theoretical knowledge that no partition exists,
thus we are entitled to say that we know the chance of an individual outcome
on an individual trial only if we are entitled to say that we know that the
chance setup admits of no partition. Now we can adopt different standards for
saying this. The clearest case (where, as it were, even God cannot partition
the trials) is one where we know that the system is indeterministic. Yet our
standards need not be so absolute. We may know that whatever God could
do, there could be no practicable project of partitioning the trials. The sys-
tems on which people gamble are designed so that this is so. In such a system
we have effectively ruled out the possibility of an improved judgement about
any single trial, and we can properly express belief that we have succeeded in
doing this by claiming knowledge of the chance. In other cases, responsible
judgement is all that we want or need.
How far can our quasi-realism, our attempt to found apparently realist
practice on a subjective basis, succeed? A possible stumbling block would be
talk of chances as explaining events, or as themselves being things that need
explanation. This would seem to invest them with some ontological standing,
with a real influence on the world, which fits ill with the projectivist picture.
But perhaps the appearance is deceptive, for everything will depend on the
interpretation we make of such explanatory claims. Clearly, if we have a
generalized antirealist attitude toward science, then chances could be as hon-
est inhabitants of theory as any other posit can be. But even without this, once
we have incorporated the notion of truth, or right opinion, there will be
natural things to ask and say about why chances are what they are. We can ask
what it is about the world that makes it the case that one particular distribu-
tion of confidence over propositions about some subject matter is right. And
we can, in asking why something is the case, cite that some such distribution is
right as part of our answer, even if this looks dangerously like giving some
chances a real, causal, place. It need not be doing so, for instance, because if
we endorse one distribution (say, give a 0.5 chance to heads in a coin toss) it
will standardly follow that we ought to endorse others (a 0.25 chance of two
heads on two trials) and it is no surprise that if asked why this is the chance on
two trials, we reply by citing the first 'fact'. This is not the place to enter into
all the moves a projectivist might make in tackling explanatory contexts, but
their mere existence is unlikely to be much of a problem for him. The empiri-
cal part of science connects frequencies with whatever factors influence them,
and our reaction to this knowledge is our talk of chance. If, in expanding such
reactions, we find ourselves talking of chances explaining things and needing
explanation, the proper response is to ask what we are projecting onto the
world by making these remarks.
A more serious threat is that with its very success quasi-realism takes much
 Opinions and Chances 91

of the impetus out of subjectivism. Responsible subjectivism is less fun. If,

one by one, a quasi-realist programme takes over the things that realists used
to consider their special private property, then the view that we have real
metaphysical options becomes more doubtful. We all thought we knew what
we meant by subjectivism, as opposed to propensity theories, frequency theo-
ries, and so on. But, if I am right, the intellectual practices supposedly defini-
tive of these different positions can be available to all, so that the old defini-
tions and divisions appear quite artificial.
In any case, if there is a metaphysical issue, then the subjectivists, so far as
we have yet discovered, may have been right about it. But the subjectivism is
not the irresponsible and therefore inefficient brand with which Ramsey is
wrongly associated. It maintains standards for proper projection, and those
standards go beyond coherence and beyond the dynamic coherence that I
discussed. They involve a proper respect for frequencies, arising from a
proper respect for induction. The main consequence of this responsibility is
that subjectivism, as a metaphysic, becomes immune to a large variety of
abuses. It also finds itself able to give sense to most of the thoughts that tempt
us to realism. However, it maintains what can be seen as clear gains: it avoids
the metaphysical problems of indeterminism, since proper single-case projec-
tions can be made when we are perfectly indifferent to that issue, and by
trying to purchase realist practice from a more austere metaphysic we may
come to feel more secure in that practice. There would, of course, be much
more to be said if we were to expand the quasi-realist programme, particularly
to cover the hypothetical that are involved in the modal claims I discussed.
But my instinct is that if there are obstacles on this route, then they afford
opportunities for delightful scrambles, rather than excuses for retreat. And I
think that was Ramsey's view as well.

Addendum

It may be instructive to add some notes comparing the resolutely antirealist

approach sketched in this paper (and of course much more technically elabo-
rated in works such as Jeffrey's The Logic of Decision (Chicago: University of
Chicago Press, 1983) and Skyrms's Choice and Chance (Encino, Calif.: Dick-
enson, 1975) with the eclectic theory of David Lewis's 'A Subjectivist's Guide
to Objective Chance'. Lewis believes that the practice of science requires, in
addition to measures of reasonable partial belief, a notion of 'objective
chance'. However, our best grip on this notion is given via our conception of
what reasonable credence concerning it amounts to: the Principal Principle.
This identifies the chance of an event at a world and a time as the unique
degree of belief in the event that would arise from any reasonable initial
degree of belief, by conditionalizing on the complete history up to the time
and 'theory of chance' at the world.
This is the upbeat side (I owe the description to James Logue). It seems,
initially, grist to the antirealist mill, since direct evaluations of objective
92 Metaphysics

chance now turn out to be the kinds of thing a subjectivist can make—
provided, of course, he can access the history and the 'theory of chance'.
Objective chance might stand to confidence like values stand to attitudes. In
the philosophy of value we might say that the objective value of a state is or
reflects the attitude you would have to it given the complete facts ('history')
and theory of value. Provided that having a theory of value is something like
having a sensibility, that is, a fact-in/attitude-out grinder (or having such a
grinder plus proper concern for coherence and attitudes of fallibility, describ-
able by the antirealist), all is well. It would not be well if having such a theory
were having a description of a particular part of the way of the world, because
the antirealist will resist the metaphysics that this seems to be bringing with it.
Lewis does not see it like this. He wants to leave objective chance as a real
aspect of things, and it is instructive to see the difficulties this causes (the
downbeat side). These centre around finding a coherent status for the theory
of chance, or the 'history-to-chance' conditionals. Lewis himself appears to
vacillate over the crucial feature of this theory already mentioned, namely
whether we are to think of it as true or not. When it first comes in 'it may or
may not be a complete theory, a consistent theory, a systematic theory, or a
credible theory. It might be a miscellany of unrelated propositions about what
the chances would be after various fully specified particular courses of events.
Or it might be systematic . . .' (p. 96). But it rapidly transforms itself into the
theory that holds at w, or in other words, the truth about how history deter-
mines chance at w (p. 97). The transformation is critical for the subjectivist
versus objectivist opposition, of course. In the former guise the theory is
naturally thought of as the subject's own set of procedures or norms for
distributing confidence in the face of empirical knowledge; in the latter guise
it seems more like an objective, potentially unknown and even unknowable
specification of how real chances relate to real events in the particular world.
Only in the latter guise does it sustain Lewis's realism about chance. It is also
worth remarking that only in the latter guise does it seem the kind of thing
chances might supervene upon. We do not normally think of values, for
instance, as supervening on our own dispositions to value this or that, and it
can falsify practical reasoning to think of them doing so (see chapter 9 for
more on this. The theme is usefully taken up in Philip Pettit and Michael
Smith, 'Backgrounding Desire', Philosophical Review, 1989).
The difficulty that I interpret Lewis's paper to be exploring is not this, but
rather the problematic modal status for history-to-chance conditionals. Either
they are necessary, or they are contingent. If necessary, then any two worlds
sharing the same Humean mosaic (or even the same historical Humean mo-
saic, or set of admissible events) share the same chances. If contingent, what
are they contingent upon? Lewis dislikes the first option because he cannot
see belief in such conditionals as rationally compulsory. But we are fairly
familiar with the idea of necessary truths that are not rationally compulsory.
My own reason for disliking it would be that it appears to fly in the face of the
methodology implicit in judgements of chance. We make such judgements in
the light of finite evidence in accordance with various methodologies from the
 Opinions and Chances 93

family of laws of large numbers. When a finite pattern is interpreted as a

Bernoulli sequence in which each event has a chance p, we understand that
the same pattern could have been the outcome of a different chance set up: a
non-Bernoullian sequence in which chances change, or a sequence in which
every event had a somewhat different chance than p. The history-to-chance
conditionals are not treated as necessarily true. This is not the 'assertibility
condition' for chances.
But if contingent, contingent upon what? The question will not trouble an
antireductionist realist, content to see chance as another self-standing domain
of fact, which may or may not vary with other things. It may not trouble a
realist roughly of Armstrong's type, content to think in terms of universals and
relations between them forming self-standing 'gridlocks' or instructions for the
ways things fall out. But it rightly troubles Lewis (who doubts whether these
theories can give any account of the Principal Principle, which sums up all we
take ourselves to know about chance), and it troubles anyone who thinks that
the Humean mosaic is all that there is. Perhaps one might think this but also
think that a chance exists (how?) of there being other things that would in turn
have altered the chances at the Humean world. But let us call a resolutely
Humean world one where the mosaic makes up all that there is and all that
there is any chance of being. What can history-to-chance conditionals be contin-
gent upon in a resolutely Humean world? How could they be true in one such
mosaic and not true in another identical mosaic? And if they could not so vary,
we seem to be back with the necessity of history-to-chance conditionals.
The dilemma is sharp. But it should trouble Lewis only because of his
realist leanings. If we jettison those we can give a perfectly good description
of the way in which norms governing belief in chances should be governed by
patterns in the Humean mosaic, the way in which those norms are neither
rationally compulsory, nor, in the light of the large-numbers methodology,
thought of as necessarily yielding the truth (the last word about chance at a
world). There is no final puzzle to dispel over how chances ('the things them-
selves') can be contingently related to the mosaic: finding that intractable is a
symptom of the truth-conditional hankerings I tried to diagnose for modality
in essay 3. (My thanks to David Lewis for extensive reaction to an earlier
commentary on his paper, and especially to James Logue.)
 5
Hume and Thick Connexions

I. Two Approaches
Recently there has been a pronounced shift in the interpretation of Hume on
causation. The previous weight of opinion took him to be a positivist, but the
new view is that he is a sceptical realist.1 I hold no brief for the positivist view,
but I believe it needs replacing by something slightly different, and that at best
it shows an error of taste to make sceptical realism a fundamental factor in the
interpretation of Hume.
Let us call any concept of one event producing another, or being necessar-
ily a cause or consequence of another, and that involves something in the
events beyond their merely being kinds of events that regularly occur to-
gether, a thick concept of the dependence of one event on another. Then on
the positivist account, Hume believes that no thick notion is intelligible. On
the positivist view there is very little that we can ever understand and mean by
a causal connexion between events. All we can understand and properly mean
by talk of causation is that events fall into certain regular patterns, and the
positivist interpretation is that Hume offered this as a reductive definition of
causation. This is the famous regularity theory, summed up in the 'philosophi-
cal' definition: 'an object, followed by another, and where all the objects
similar to the first are followed by objects similar to the second'.2 The
sceptical realist view denies that Hume offered any such reduction or analysis
of the notion of causation. It takes seriously the many passages in which
Hume appears to allow that we are talking of some thick notion of depen-
dence of one event on another, going beyond regular succession. It takes it
that Hume acknowledges that there is some such thick relation, even if it will
be one about whose nature and extent we are doomed to ignorance. Hence, in
John Wright's phrase, sceptical realism.
At first sight the difference between positivism and sceptical realism is
reasonably clear, and it is plausible that if these are the two options then
Hume is better seen as tending toward the second. But, as proponents of the
1. I have in mind Edward Craig, The Mind of God and the Works of Man (Oxford: Claren-
don Press, 1987); Galen Strawson, The Secret Connexion (Oxford: Clarendon Press, 1989); and
John Wright, The Sceptical Realism of David Hume (Manchester: Manchester University Press,
1983).
2. Enquiry Concerning Human Understanding, ed. Selby Bigge, section VII, part II, p. 76.

94
 Hume and Thick Connexions 95

sceptical realist interpretation realize, there is one big problem, arising from
Hume's theory of meaning. Sceptical realism seems to demand that we under-
stand what it would be for one event to depend thickly upon another, even if
we are ignorant of the nature of this relation; Hume seems to insist that we
have no impression and hence no idea of any such dependence.
The problem here is a problem for any interpretation and can be focussed
on a contradiction, to which Hume seems to be committed:
(1) We have no ideas except those that are preceded by suitably related
impressions.
(2) There are no impressions that are suitably related to the idea of a thick
necessary connexion between distinct events.
(3) We have an idea of a thick necessary connexion between distinct
events.
The 'suitable relation' spoken of includes direct copying, in the case of
simple ideas, and whatever is covered by 'compounding' in the case of com-
plex ideas that are compounded out of simple ones.
The positivist interpretation takes Humes to be claiming that when we talk
of causation we mean only something that strips out the thick element of
necessity and substitutes regular contiguous succession. So (3) is false. The
difficulty is that Hume apparently denies this:
Shall we then rest contented with these two relations of contiguity and
succession, as affording a compleat idea of causation? By no means. An
object may be continguous and prior to another, without being consider'd as
its cause. There is a NECESSARY CONNEXION to be taken into consideration;
and that relation is of much greater importance, than any of the other two
above-mentioned.3
The central problem in interpreting Hume is coping with the contradiction.
The sceptical realist strategy is to downplay the importance of the theory of
understanding, so that even if Hume officially said (2), it played a negligible
part in his view of causation.

II. A Doubtful Distinction

How then does the sceptical realist deal with the problem of meaning? Ed-
ward Craig and Galen Strawson draw attention to a distinction that occurs in
Hume's writings.4 When the theory of ideas threatens our idea of external
existence or 'body', it is said that Hume invokes a distinction between what
we can 'suppose' and what we can 'conceive', the idea being that we can
coherently suppose that there are things of some sort (external objects) even
when strictly we have no idea of what it is that we are supposing. Another way
of putting it is that we have a 'relative' idea of things whose 'specific' differ-
3. Treatise of Human Nature, ed. Selby Bigge (Oxford: Oxford University Press, 1888), p. 77.
4. Edward Craig, op. cit., p. 124, and Galen Strawson, op. cit., chapter 12.
96 Metaphysics

ence from other things we cannot comprehend. We could say that we have no
representative idea of what we talk about, but a relative or relational idea,
locating it by its role. We would talk of a 'something-we-know-not-what' that
does something or bears some relation to an aspect of the world of which we
do have an idea. This distinction solves the contradiction by distinguishing
between the terminology of (2) and (3). Hume thinks we have no representa-
tive idea of causation: we have no impression of it, and in some important
sense it remains incomprehensible, and we cannot represent to ourselves what
it is. What we do have, however, is a relational idea of it: it is whatever it is
that issues in regular successions of events or upon which such patterns de-
pend, or whatever forces such regularities. The negative side is given in (2)
but the positive side in (3).
The texts, however, give no direct support to this interpretation of Hume.
While he does indeed use both a 'relative' versus 'specific' distinction and the
possibility of 'supposing' what we cannot 'conceive', he uses them very spar-
ingly indeed. In fact he never uses either, nor mentions either, in connexion
with causation. He never uses or mentions either in the Enquiry or in the
Dialogues in any context at all. This alone makes them unlikely candidates for
a central role in understanding his mature philosophy.5 But worse, there are
warning signs to be noticed when they occur in the Treatise. There are four
occurrences: on pp. 67-68, p. 188 referring back to them, p. 218, and p. 241.
In none of these cases is Hume actually contrasting a specific versus a relative
idea of any one property or relation, enjoining us that we can understand a
property or object by its relations even if we cannot understand it by some
stricter standard derived from the theory of ideas. On the contrary, in each
context it is the impossibility of conceiving a 'specific difference' between
external objects and perceptions that is the focus of attention. 'Specific' quali-
fies the properties supposedly differentiating external objects from ideas, and
of these specific qualities we know and understand nothing by any standard at
all. Here are the two major passages with enough surrounding context to
matter:
Now since nothing is ever present to the mind but perceptions, and since all
ideas are deriv'd from something antecedently present to the mind; it fol-
lows, that 'tis impossible for us so much as to conceive or form an idea of any
thing specifically different from ideas and impressions. Let us fix our atten-
tion out of ourselves as much as possible: Let us chase our imagination to the
heavens, or to the utmost limits of the universe; we never really advance a
step beyond ourselves, nor can conceive any kind of existence, but those
perceptions, which have appear'd in that narrow compass. This is the uni-
verse of the imagination, nor have we any idea but what is there produc'd.
The farthest we can go towards a conception of external objects, when
suppos'd specifically different from our perceptions, is to form a relative
idea of them, without pretending to comprehend the related objects. Gener-
ally speaking we do not suppose them specifically different; but only attri-

5. It is particularly odd that Strawson relies upon them, since he conceives of the Enquiry as
embodying Hume's official theory of causation.
 Hume and Thick Connexions 97
bute to them different relations, connexions and durations. But of this more
fully hereafter, (pp. 67-68)
Philosophers deny our resembling perceptions to be identically the same,
and uninterrupted; and yet have so great a propensity to believe them such,
that they arbitrarily invent a new set of perceptions, to which they attribute
these qualities. I say, a new set of perceptions: For we may well suppose in
general, but 'tis impossible for us distinctly to conceive, objects to be in their
nature any thing but exactly the same with perceptions. What then can we
look for from this confusion of groundless and extraordinary opinions but
error and falsehood? And how can we justify to ourselves any belief we
repose in them? (p. 218)
It requires some daring to take these passages as a model for sceptical
realism. Hume is far—about as far as can be—from saying that we actually
possess a going idea of the external world that allows us to understand, by
some weak standard, what the externality is that we do not know about. Each
of the two passages gives the strongest contrary impression. The first affirms
idealism ('. . . we never really advance a step beyond ourselves'). The second
introduces the 'supposes' versus 'conceives' distinction only while he simulta-
neously dismisses its effect out of hand. Its dismissal justifies Hume in describ-
ing his Philosophers (the culture whose spokesman is Locke) as actually
inventing new perceptions, rather than inventing new things different from
perceptions. This is the very opposite of the view a sceptical realist Hume
should take. He should admit that a Lockean succeeds in introducing a (rela-
tive) notion of an external object as something that has various relations to
our perceptions, and then go on to worry how much we know about such
objects. Hume does not do this: he simply dismisses the idea that we have a
set of determinate, intelligible, propositions about which, unfortunately, we
shall never know the truth. It is not that we understand something, but
cannot know whether it is true; it is that we give ourselves explanations that
seemed to introduce an intelligible concept, but in fact fail (the demands put
upon an external world independent of perception are simply inconsistent).
We are in the domain of a 'confusion of groundless and extraordinary opin-
ion' where our only hope is to abandon reason altogether. So, even when it is
used, the 'specific' versus 'relative' distinction is not used as Craig and
Strawson would have it used in the different area of causation, to which, as I
have said, Hume never applies it.

III. Another Distinction

Before proceeding it is necessary to have in mind two things that might be
asked of 'thick' causation. When we think of a causally connected pair of
events, such as the impact of the first billiard ball causing the motion of the
second, we want there to be a further fact than (mere) succession, or even
mere regular succession of these kinds of event. We want there to be a depen-
dency or connexion, a fact making it so that when the first happens the second
98 Metaphysics

must happen. Call this the desire for a causal nexus. But now suppose we shift
our gaze to the whole ongoing course of nature. Again, we may want there to
be a further fact than mere regular succession. We feel that the ongoing
pattern would be too much of a coincidence unless there is something in virtue
of which the world has had and is going to go on having the order that it does.
We want there to be some secret spring or principle, some ultimate cause, 'on
which the regular course and succession of objects totally depends'.6 This is
whatever it is that ensures the continuation of the natural order, that dispels
the inductive vertigo that arises when we think how natural it might be, how
probable even, that the constrained and delicate pattern of events might fall
apart. Call the desire for this further fact the desire for a straitjacket on the
possible course of nature: something whose existence at one time guarantees
constancies at any later time. 7
A fact alleviating this vertigo has to be a very peculiar fact, for the follow-
ing reason. It has to be something whose own continued efficacy through time
is subject to no possibility of change or chance of failure. For otherwise the
fact that it keeps on as it does would itself be a case of coincidence or fluke,
another contingency crying out for explanation and engendering inductive
vertigo. Some think they can point us toward a fact with this potency. Some
draw comfort from God's sustaining will (as if anything understood on the
analogy of our own mental states could be timeproof!). David Armstrong
believes that a kind of necessary, timeless, gridlock of universals will do.8
Galen Strawson takes comfort in fundamental forces constitutive of the na-
ture of matter.9
It is easy to conflate the desire for a nexus, case by case, with the desire for
a straitjacket. But Hume (sometimes—but see below) is clear that they are
different. They are different because whatever the nexus between two events
is at one time, it is the kind of thing that can in principle change, so that at a
different time events of the same kind may bear a different connexion. Thus
suppose we grant ourselves the right to think in terms of a thick connexion
between one event and another: a power or force whereby an event of the first
kind brings about an event of the second. Nevertheless, there is no contradic-
tion in supposing that the powers and forces with which events are endowed at
one time cease at another, nor in supposing that any secret nature of bodies
upon which those powers and forces depend itself changes, bringing their
change in its wake. Hume emphasizes this point in both the Enquiry and the
Treatise.10 It is his reason for denying that the problem of induction can be
solved by appeal to the powers and forces of bodies. But it is equally a reason
for separating the question of a nexus from that of a straitjacket. Nexuses by
6. Enquiry, p. 55.
7. At least. It may be that its existence at one time should entail its existence at any
previous time as well. But one way of gesturing at what is wanted is to imagine God creating it by
some kind of fiat or act of law-giving, whose writ would run only into the future.
8. What is a Law of Nature (Cambridge: Cambridge University Press, 1983), p. 88 ff.
9. E.g., p. 91, pp. 254-55.
10. Treatise, pp. 90-91; Enquiry, p. 37.
 Hume and Thick Connexions 99

themselves do not provide a straitjacket. The ongoing regularity and con-

stancy even of a thick nexus between one kind of event and another is just as
much a brute contingent regularity as the bare regular concatenation of
events.11 In each case we have something that can engender the inductive
vertigo, or whose continuation through time might be thought to demand
some kind of ground or ultimate cause or straitjacket.
The difference between a nexus holding on some particular occasions and
a straitjacket guaranteeing the continuation of a pattern of connexions is easy
to overlook. This is because of a lurking epistemological difficulty. Suppose
one thinks that a particular nexus can be known for what it is, for instance by
some observation whose content is more than the mere succession of events.
One might report this by claiming to have seen that the one event had to
happen, given the other. But if you see a 'must' in one pair of events, would
you not thereby see that it will hold for every pair of some kind that the
original pair enables you to identify? How could you see it without seeing
something with general implications, and ones that are immune to temporal
change? In other words, you will take yourself to have seen a timeproof
connexion: one that rigidly governs how things could ever fall out. To put it
the other way round, if things were not to fall out as expected, the original
claim to have seen that the one event had to follow the other is refuted. This in
turn makes it hard to see how a particular nexus could be an object of observa-
tion. Observation extends only to limited periods of space and time: how
could we have within our view something that essentially casts its net over the
whole of space and time?
This problem probably explains one puzzling feature of Hume's proce-
dure. He repeatedly affirms that someone who has a full apprehension of a
thick causal connexion would be in a position to make an a priori claim about
the way events will fall out and what kind of event will be caused by another.
He argues that because we cannot have this time-proof knowledge we do not
apprehend the causal connexion, for instance in the exercise of our own will.12
The argument seems initially to be, as Craig describes it, a muddle, since
there is no evident reason why someone apprehending a nexus on one occa-
sion should thereby know that the same nexus will obtain on another—the
very point Hume himself emphasizes when arguing that powers and forces will
not solve the problem of induction.13 I suspect that Hume sees that nothing
would really count as apprehension of a particular 'must' unless it carried with
it implications of uniformity for the general case. It is to be (per impossible) a
particular apprehension, but one with the consequences of apprehending a
straitjacket. Someone apprehending a straitjacket for what it is will as a
11. One might seek to avoid this by the verbal manoeuvre of identifying kinds of events by
their causal powers, in which case it will follow that events of the same kind will bear the same
causal connexions. But as Hume in effect points out, inductive vertigo then transfers itself to the
contingent question of whether future events with the same sensory appearance will turn out to be
of the same kind.
12. Enquiry, section 7, passim.
13. Treatise, p. 91, and Craig, p. 97.
100 Metaphysics

consequence know its immunity to time and chance: he will know the timeless
'must' that it guarantees. He will be apprehending the impossibility that
events should ever transpire otherwise. He has therefore a piece of knowledge
that, although it took an empirical starting point in the apprehension of an
individual thick necessary connexion, can be seen a priori to have implications
for all other places and times. And it is this that Hume treats as his target,
even when the issue ought to be the apparently lesser one of the particular
nexus.14
There may be some room for manoeuvre over the lesser claim to have
apprehended a particular, but not necessarily timeproof, thick connexion.
One might try allowing the particular apprehension not to carry any implica-
tions for what might be present on other occasions.15 The difficulty will be that
an apprehension of a mutable thick connexion does not gives us quite what we
want from knowledge of causation. That knowledge has to have a conse-
quence: the subject possessing it must be prepared to foretell the one kind of
event on the appearance of the other. It is not at all clear how apprehension of
a particular relation obtaining at a particular place and time could automati-
cally carry any such consequence: one might, as it were, say that this is how
events are connected today, and form no expectation, and not know what to
expect to happen tomorrow.
Sceptical realism might characterize Hume's position on either the nexus
or the straitjacket. But unless we understand the extraordinary demands on a
straitjacket we shall fail to see that realism concerning it is hardly important
compared to his scepticism. Thus when Strawson opposes the Regularity
Theory, with its ongoing flukes, by citing 'fundamental forces' essentially
constitutive of 'the nature of matter', and invokes these to soothe away induc-
tive vertigo, he is surely forgetting Hume's point.16 Even if forces are taken 'to
latch on to real, mind independent, observable-regularity-transcendent facts
about reality',17 they need something further in order to serve as a strait-
jacket. They need necessary immunity to change; they need to be things for
which the inductive vertigo does not arise. Equally, if the 'nature of matter' is
to help, then the continuation of matter must not be just one more contin-
gency, whose falling out the same way instant after instant, time after time, is
a cosmic fluke. The force that through the green fuse drives the flower might
falter, and so might the fuse and the flower, but a straitjacket must not. Its
immunity to change must be necessary, for if it is contingent then either it is a
fluke—of any changes that might occur, none ever does—or else this regular-
ity is itself not brute but demands some further straitjacket in the background,
of which we have even less inkling. The point is that we will not locate it by
ordinary talk of 'force' and its cognates. For even if Hume can countenance

14. On these issues, see also Peter Millican, 'Natural Necessity and Induction', Philosophy
61, 1986.
15. G. E. M. Anscombe, 'Causality and Determination', in Metaphysics and the Philosophy
of Mind (Oxford: Blackwell, 1981).
16. Strawson, p. 91.
17. Ibid.
 Hume and Thick Connexions 101

understanding of a thick nexus, the theoretical pressures on a straitjacket are

a great deal more demanding.18
Hume's main interest in causation is to destroy the idea that we could ever
apprehend a straitjacketing fact: we have no conception of it, nor any concep-
tion of what it would be to have such a conception, nor any conception of how
we might approach such a conception. In particular we must not think of the
advance of science as targeted on finding such a thing. The lesson drawn from
Newton is that just as Principia gives us the operation of gravitational force
but does not tell us what it is, so any conceivable advance in science can do
only more of the same. It can put events into wider and more interesting and
exception-free patterns, and that is all. 'The most perfect philosophy of the
natural kind only staves off our ignorance a little longer."19
Would it be easy for Hume to allow us a 'relative' idea of a straitjacketing
fact—a something-we-know-not-what that governs/brings about/explains the
continuing order of nature? We understand this only insofar as we understand
the relation of governing or bringing about. But can we understand the rela-
tion? Can Hume say the relation part of the relational idea is intelligible? The
question is whether we know what governing or bringing about would be
when we have no example, and indeed no conception, of the kind of fact
alleged to be doing it. Hume, given his endorsement of Berkeley's theory of
ideas, must say that we cannot take relational ideas (governing, forcing,
grounding, issuing in, bringing about) out of the context within which they
have intelligible application, and apply them without blush in contexts in
which they do not.20 We can generate the general idea only if we have particu-
lar examples. Otherwise, comprehension fails.
Nevertheless, it will be said, even if this shows that we have no idea at all of
what would count either as straitjacket, or as knowledge that some kind of fact
provides one, it seems plain that Hume allows that there is one, even while
insisting on scepticism about its nature. Sceptical realists might be right that he
allows us a 'relative' idea of such a fact, silently betraying the Berkeleyan

18. Strawson is probably betrayed into this conflation by using the one term 'Causation'
(with a capital 'C') equally for a thick nexus and a thick straitjacket.
19. Enquiry, p. 31. This is the famous point where Newton said 'hypotheses nonfingo', and
the point that left contemporary scientists such as Huygens and Leibniz, who had wanted to know
what gravity was and not merely how bodies moved under its influence, feeling badly let down.
Newton was quite within his rights to want more scientific understanding of gravitational attrac-
tion, and Hume does not oppose the goal. But if Newton and his contemporaries wanted a
different thing—an understanding of the impossibility that events should ever fall out otherwise—
then Hume stands in his way. Hume does not magnify the difference between himself and New-
ton, but if Newton was aiming at this superlative piece of understanding and thought that the
methods of natural science might give it, then Hume is clearly opposed. He was the first to see that
what Newton did was the only kind of thing that could ever be done.
20. Berkeley's rigour on this is apparent in his constant polemic against 'abstraction' and in
such matters as his embargo on taking causal relations away from the domain of the will, given
that it is this that is the basis of our understanding of them. More directly relevant is his insistence
that if you try to introduce a 'relative notion' of matter as whatever it is that supports various
properties, you mean nothing. As well as the passage quoted in the text, see Principles of Human
Knowledge, part 1, section 80.
102 Metaphysics

background. Even if this were technically correct—and we have seen how far
it stretches the texts—it would still misplace the stress; this is why I originally
described it as an error of taste rather than an outright mistake. The point is
that Hume is utterly contemptuous of any kind of theorizing conducted in
terms of such a thing. We are at the point where anything we say 'will be of
little consequence to the world',21 or in the world of 'notion(s) so imperfect
that no sceptic will think it worthwhile to contend against (them)'. 22 His atti-
tude must be the same as that he holds to an equally noumenal substratum,
supporting the qualities of matter:
But these philosophers carry their fictions still farther in their senti-
ments concerning occult qualities, and both suppose a substance support-
ing, which they do not understand, and an accident supported, of which
they have as imperfect an idea. The whole system, therefore, is entirely
incomprehensible. . . ,23
He is here directly echoing Berkeley:
Lastly, where there is not so much as the most inadequate or faint idea
pretended to: I will not indeed thence conclude against the reality of any
notion or existence of any thing: but my inference shall be, that you mean
nothing at all: that you imply words to no manner of purpose, without any
design or signification whatsoever. And I leave it to you to consider how
mere jargon should be treated. 24
Craig especially makes the case that there is importance in the positive claim
that something-we-know-not-what exists, and the importance is sceptical: it
enables Hume to destroy any pretension to finding what we might anteced-
ently have hoped to understand about nature. I agree entirely that this critical
aim is essential to Hume, and at least as important as the theory of understand-
ing itself. But Hume enjoys this realignment without himself making any
positive claim about the existence of any mysterious, straitjacketing fact or
facts. The realignment of our self-image, our philosophy of what real discov-
ery and understanding might be, is independent of any such assertion. We do
not ourselves have to think the other side of the line to learn how tightly the
line defining the limit of all possible empirical enquiry is drawn. The point is
that our real engagement with the world, in our understanding and our sci-
ence, and our self-image or philosophical understanding of the notions we
actually use must sail on in complete indifference to any facts transcending
our ideas. 'Relative' ideas of such facts play no role any more than relative
ideas of many things: Cartesian egos (simple, indivisible entities whose perma-
nence ensures the identity of the self); the substratum in which properties
inhere; objective goods commanding the will of all those who apprehend
21. Enquiry, p. 155.
22. Treatise, p. 168. 'I am indeed, ready to allow, that there may be several qualities both in
material and immaterial objects, with which we are utterly unacquainted; and if we please to call
these power or efficacy, 'twill be of little consequence to the world'.
23. Treatise, p. 222.
24. Berkeley, Three Dialogues between Hylas and Philonous, dialogue 2, paragraph 121.
 Hume and Thick Connexions 103

them, and so on. Since the actual business of making judgements about the
identity of the self, or the possession of properties by things, or what is good
or bad, goes on in complete indifference to these things, they play no role in
our real understanding.25 They have no use at all: nothing will do just as well
as something about which nothing can be said.

IV. The Nexus

Perhaps the same is not true of individual thick connexions, that is, the particu-
lar causal nexus obtaining between specific events at a time. Don't we give
every employment to such a notion? And if sceptical realists are right that
Hume is not giving us a positivist reduction, do they remain in possession of
the field here at least? I do not think so, for there is a third option: a truer
description of Hume on ordinary empirical causation would be that he is
neither a positivist nor a sceptical realist, but rather a not-so-sceptical anti-
realist.26 That is, he gives us a story explaining and even justifying our use of
the vocabulary of causation, while denying that we represent a real aspect of
the world to ourselves as we use it.
The outline of Hume's positive theory of causation is well known. The
mind's perceptions, which form the material with which it must work, reveal
only a regular succession of events. However, upon experience of such a
regular succession the mind changes. It does not change by forming an impres-
sion or idea of any external property invisible in one instance alone. It
changes functionally: it becomes organized so that the impression of the ante-
cedent event gives rise to the idea of the subsequent event. No new aspect of
the world is revealed by this change: it is strictly nonrepresentative, just like
the onset of a passion, which which Hume frequently compares it.27 But once
it takes place we think of the events as thickly connected; we become confi-
dent of the association, we talk of causation, and of course we act and plan in
the light of that confidence.
There are two separate components in this story: the contribution of the
world to our apprehension, and the functional change in the mind itself.28
These are the two aspects separated in the famous two 'definitions' of cause:

An object, followed by another, and where all the objects similar to the
first are followed by objects similar to the second.

25. Strawson is at pains to show that not all Hume's reference to straitjacketing facts are
ironic, but I do not think he shows that they are not contemptuous.
26. Or, quasi-realist, in the sense of the essays.
27. Norman Kemp Smith, The Philosophy of David Hume (London: MacMillan, 1942),
chapters I and II. These present convincing evidence that this comparison was the prime mover of
Hume's theory of causation. It opened up the 'New Scene of Thought' of which he speaks in the
1734 letter to (probably) George Cheyne.
28. As clear a statement as any is Hume's recapitulation, Enquiry, pp. 78-79.
104 Metaphysics

An object followed by another, and whose appearance always conveys

the thought to the other.29

The first 'philosophical definition' describes the contribution of the world,

insofar as we can apprehend it, and the second 'natural' definition describes
the nonrepresentative, functional difference in the mind that apprehends the
regularity. The parallel with Hume's philosophy of ethics is so far complete:
again, there is a neutral starting point in the mind's apprehension of some
nonethical facts, and then the onset of nonrepresentative passions ready to be
voiced in our moralizing.
It is only now that complexities start, but unfortunately Hume gives less
help with them than one would wish. The theory so far tells us of a
nonrepresentative change, a change in the structure of our expectations, that
gets expression when we deem two events to be causally connected. But it has
not yet conjured up a full theory of the content of propositions about cause. It
does not tell us, for example, what we are bothered about when we wonder if
A caused B, what we are saying when we say that every event has a cause, or
whether we can sensibly talk of unknown causes. We need more detail about
the way in which cause becomes objectified so as to be spoken of as a feature
of the real world, if its origin is in a feature of our own minds. Hume shows
little interest in such questions, and indeed against the background of the
theory of ideas, he can point only in misleading directions. He says, for
example, that by a necessary connexion we 'mean' a connexion in the mind,
leaving himself open to interpretation as a kind of Berkeleyan, taking the
idea of necessity to be a representation of some thick connexion we are aware
of in our own minds. He then has to spend Part 1 of Section VII of the
Enquiry averting this misunderstanding. 30 In his theory of morality he simi-
larly seems unclear whether he is saying that virtue and vice are 'nothing in
the objects', but only sentiments in us, or that they are the qualities of objects
that tend to arouse those sentiments.31 What he lacks is a link between the
real functional difference and the thick content to give causal judgements: the
way we talk and think in terms of a projected property of things.32 A telling
point here is that in both the Treatise and the Enquiry he produces the 'two
definitions' only at the end of the discussion, and in each place he does so
apologetically, in effect telling us that they are not to be regarded as strict
29. Enquiry, pp. 76-77.
30. An interesting scholarly question, to which I do not know the answer, is why he took
such elaborate care in the Enquiry, section VII, to distinguish his theory from Berkeley's, when
the Treatise contains no corresponding passages. It is one of the very few cases where the Enquiry
is fuller than the Treatise. Did some review or correspondence make the need evident to him?
31. His lists of virtues (e.g., Enquiry Concerning the Principles of Morals, appendix I, p.
258) specify the properties of people, such as benevolence, serenity, and so on, that make us love
them; his official position (e.g., Treatise, pp. 471, 614) identifies the virtue with the sentiment
itself.
32. Hume is quite prepared to allow that our common notion of cause contains defective
elements—see the footnote on p. 77 of the Enquiry. But overall he is perfectly friendly to the way
we think.
 Hume and Thick Connexions 105

definitions. On the view I am recommending this is right: they separate the

two different aspects of the matter—the contribution of the world, and the
change in us. But they do not give us a lexicographer's analysis, and we
should not expect one. There is no way of moralizing without using a moral
vocabulary, and no way of causalizing without using the vocabulary of cause,
efficacy, or power.
Notice, however, how many cards Hume holds in his hands. The basic
theory is flexible enough to accommodate many points that are usually raised
against him. Our reactions to nature are subtle: not all regularities betoken
cause, and sometimes we attribute cause after miniscule experience of regular-
ity.33 Well and good: the basic theory need put no limits on the input to our
causalizing, any more than his theory that in moralizing we voice a passion
puts a limit on the input to our moralizing. On the output side, the change in
the structure of our thought after we have deemed a sequence to be causal
may also be complex. Its heart is that we 'make no longer any scruple of
foretelling' one event upon the appearance of another. But there may be
other changes. We may become willing, for example, to hold the sequence
constant as we think about what would have happened if something else had
happened, or what would happen if something else were to happen. Once we
view a sequence as causal, it is held fixed as we conduct counterfactual and
conditional deliberations. Well and good: the basic theory puts no limits on
such consequences either. The theory also happily predicts the 'intuitions' that
lead people to detest the Positivistic 'regularity' theory of the content of our
causal sayings. Someone talking of cause is voicing a distinct mental set: he is
by no means in the same state as someone merely describing regular se-
quences, any more than someone who appreciates some natural feature as
good is in the same state of mind as someone who merely appreciates the
feature. The difference in this case is in the sentiment or passion that the
feature arouses, and, in the causal case, in the fixity that the sequence of
events takes in our thinking. Finally, the contradiction I identified at the
beginning of this essay is sidestepped by distinguishing a representative idea
of a connexion, which we do not have, from a capacity to make legitimate use
of a term whose function is given nonrepresentatively, which we can have.
There are, I believe, only two ways in which this kind of theory could be
opposed. One is to deny that a Humean could forge the missing links between
the functional difference we are expressing and the surface content of our
causal judgements. The other is to deny that we have here a distinctive posi-
tion, by assailing the limits on 'representation' under which Hume operates.
The first attack presses the point that Hume needs to tell us what happens not
just when we think that A causes B, but also when we think that there exist
unknown causal connexions, that regardless of whether we had ever existed
there would still have been causal connexions, and so on. We think in terms of
causation as an element of the external world, and there remains a real
33. Hume discusses these complexities in Treatise I, pp. iii, xv: 'Rules by Which to Judge of
Causes and Effects'.
106 Metaphysics

question of how much of this thought Hume can explain and how much he has
to regret. However, his prospects for deflecting this first criticism must be
quite bright. For as we have seen, he is working with exactly the same ingredi-
ents in the case of ethics. Here too there is the task of explaining the appar-
ently objective content of moral judgements given their source in the pas-
sions, but here it is much harder to believe that the problem is insoluble, and
Hume certainly did not believe it to be so.
The second attack need not deny Hume his ingredients. It simply claims
that we can cook with them in a different way, awarding ourselves the right to
a genuinely representative concept of causation. For when should we say that
we have a representative idea of a property or relation? One answer would
be: when we can picture it holding, or exhibit to ourselves in imagination a
scene in which the property or relation is visibly instanced. This is a natural
empiricist answer and the one that leaves Hume poised to argue that we have
no representative idea of thick causal connexions. For a view (or succession of
other experiences: sounds, felt pressures, and so on) in which there is given a
certain succession of events, and in which one event causes another, need be
no different from a view in which the one event does not cause another, but in
which the same succession happens anyhow. This is why we have to interpret
sequences as causal, and, however automatic this act is, it is still one that
needs to be performed. But empiricism nowadays sounds like prejudice: why
should we not have a theoretical concept of a thick causal connexion, allowing
that there is a step from the raw appearance of a scene to the belief that it
instances such and such connexions, but also insisting that we know what it is
for such connexions to exist? We have a theoretical idea of them, and the idea
represents the way the world is when they are present.
The real problem with this is that it only works if we also understand the
relation between the thick connexions and the ongoing pattern of events.
Thick connexions make events happen; they guarantee outcomes, they issue
in patterns of events, and so on. But these are terms of dependency or causa-
tion, so we understand the theory only if we understand them. And this
understanding in turn is queried by the problems described above: any realist
theory needs to tell us how the 'musts' present on one occasion throw their
writ over others. Otherwise it fails to give us what we want from a causal
understanding of the world. For all the story goes, someone might be a
virtuoso at detecting particular thick connexions, yet have no idea what to
expect or how to conduct counterfactual and conditional reasoning.
The net result is that any such realist theory looks extravagant. It asks
from us more than we need. To see this, imagine a character we might call the
Bare Humean. The Bare Humean misses out this capacity for apprehension
or theory, so does indeed lack the representative idea of thick connexions that
these are supposed to give us. But she goes through the functional change that
Hume describes, and conducts her expectations and actions accordingly. She
can be an enthusiastic natural scientist, finding concealed features and con-
cealed patterns in nature to aid prediction. She can understand that rinding
ever more simplicity and ever more general patterns may be 'set us as a task',
 Hume and Thick Connexions 107

so that there will always be more to know about nature. She will need a
vocabulary to express her confidences and her doubts, and to communicate
them to others; she will be a virtuoso at the salient features that are usable day
by day to control her world. What else does she need? Are we sure she is
missing anything at all—isn't she a bit like you and me?
This page intentionally left blank
 II
ETHICS
This page intentionally left blank
 6
Moral Realism

I
Granted that it is correct to reply to a moral utterance by saying That's true'
or 'That's not true', the question remains of what sort of assessment is indi-
cated by these responses. Dummett, with whose paper this emphasis is associ-
ated, also says that a statement is false if a state of affairs obtains such that a
man asserting the statement and envisaging that state of affairs as a possibility
would be held to have spoken misleadingly.1 But there are ways and ways of
speaking misleadingly, of which speaking falsely is only one; and indeed not
always one, as when you say something false knowing that the hearer will
misinterpret it as something true. Even apart from the possibility of mislead-
ing someone only as to my beliefs or my authority, still I can mislead someone
about the truth of the subject under discussion when a state of affairs exists,
and I envisage it as a possibility (and indeed more than a possibility), and I say
nothing but the truth. I would do this for example if I told an interested child
that some elephants are at least twice as large as some fleas. Worse than that,
in what way does a man speak misleadingly if he sincerely judges a thing to be
good, envisaging the possibility that he may be wrong, according to standards
of which we are aware and with which we disagree? We hold his judgement
false, but I don't think we would hold him to have spoken misleadingly, unless
we just meant: what he said would mislead anyone who believed it as to the
truth about whether the thing is good. But then the relevant assessment of
misleadingness is itself to be understood in terms of truth.
Nevertheless, perhaps some candidate for the correspondence relation
other than one constructed from misleadingness exists. It is the purpose of this
essay to argue that there are features of the claim that a moral judgement is
true that prevent this being taken as the claim that it corresponds in any way
with a state of affairs. It follows that an attribution of truth to such a judge-
ment must involve a different sort of assessment, and it is a secondary purpose
of the paper to ask what that might be, and whether it should be thought
surprising that we have a form of utterance with this form of assessment. I
shall call the view that the truth of moral utterances is to consist in their
correspondence with some fact or state of affairs 'moral realism'; it is this view
1. M. Dummett, 'Truth', Proceedings of the Aristotelian Society, 1958, p. 150.

1ll
112 Ethics

that I shall argue is false. Another way of characterizing realism, adopted by

Dummett, 2 would be as the belief that for any moral proposition there must
be something in virtue of which either it or its negation is true. This certainly
points to the same view: the advantage in stressing that it is the existence or
nonexistence of a state of affairs that determines truth or falsity, according to
the realist, is that it is precisely the disparity between the logical conditions
upon the existence of states of affairs, and those upon moral truth, that is to
be urged in what follows.
There exist already two well-known arguments for the falsity of moral
realism. Both take one feature of assent to a moral proposition and claim that
this is not a feature of belief that some state of affairs obtains; whence it
follows that assent to a moral proposition—which may be expressed as belief
that it is true—is not belief that some state of affairs obtains, and thence that
the truth of a moral proposition does not consist in the existence of any state
of affairs. The features that these arguments seize upon concern respectively
our attitudes and our actions.
The first argument, then, emphasizes the connection between real assent
to a moral proposition and the possession of a certain attitude to its subject.
At its strongest, it can be stated as follows. Belief that a thing is good entails
possession of a certain attitude towards it. No belief that a thing enters into a
state of affairs entails the possession of any attitude towards it. Therefore,
belief that a thing is good is not belief that it enters into a state of affairs, and
moral propositions must be distinguished from propositions with realistic
truth conditions.
The trouble with the argument lies in the second premise. For it is not at all
evident that no belief that a thing enters into a state of affairs—no belief, that
is, with realistic truth conditions—can entail3 the possession of an attitude
toward its subject. Consider, for example, the belief that a person X is in
possession of the truth about some matter. This is clearly a factual belief. But if
I come to hold this belief about X, doesn't it follow that my attitude toward X
alters? I become prepared to defer to A"s opinion on the matter, or commend
his views to other people whom I wish to inform, and each of these is quite
naturally construed as possession of an attitude toward X. Again, consider the
way in which one's attitude to a statement must change according to whether
one believes it to be true. Finally, there are beliefs—for example, that some-
thing is alive—that may also entail the possession of an attitude, even if all that
can be said about the attitude is that it is one that one feels toward things that
are alive, but not toward other things. Now these examples may produce a
certain impatience. All that they could prove, it may be thought, is that the
notion of an attitude can be used to cover such things as a belief that something
is alive, or true. Seen this way, Wittgenstein's remark, 'My attitude towards
him is an attitude towards a soul. I am not of the opinion that he has a soul',4
2. Truth', p. 157.
3. I am using 'entails' throughout in the logical sense, so that 'P entails Q strictly implies
that P strictly implies Q.
4. L. Wittgenstein, Philosophical Investigations (Oxford: Blackwcll, 1958), part II, section IV.
 Moral Realism 113

tells us more about attitudes than about the judgement that people have souls.
Now this can only be part of the truth, for it fails to explain why Wittgenstein's
remark is a good deal more intelligible than, say, 'My attitude toward the room
is an attitude toward a room with a bottle in it. I am not of the opinion that
there is a bottle in it', but even if this was all that my examples indicated, it is
enough to topple the anti-realist argument that we are considering. For if it is
correct to say that real assent to the proposition that a thing is alive, or true, or
a true proposition, then even if real assent to the proposition that a thing is
good entails that my attitude toward it is an attitude toward a good thing—
namely one of approval—still goodness may be as much a property of a thing
as life or truth.
To this it will be replied that there are features of moral approval that are
not shared by my other examples of 'attitudes', that enable it to be identified
independently of statement of a moral belief. These features concern the
necessary consequences for our choices and actions of holding a moral posi-
tion. So the second argument we are to consider hopes to find, in the connec-
tion between moral belief and the will, a proof that moral realism is false. In
so doing it would hope to find a practical aspect of moral approval that will
distinguish it from other examples of attitudes, and so reinstate the first
argument. I shall briefly describe two difficulties into which this line of
thought runs, not in order to show that it cannot be correct, but in order to
show that the obstacles in the way of its proper statement are sufficiently great
that a new line of attack on moral realism may be not entirely otiose.
The position, then, is that the anti-realist is to describe a connection be-
tween moral belief and the will in such a way that it is clear that moral belief
necessarily has connections that no realistic belief need have. The first trap is
that of simply giving a false description of the connection, making it, for
example, logically impossible to do something that one believes to be wrong,
or logically impossible to will something that one believes to be bad. When it is
realized that these things are possible, various things may be said about the
situations in which they happen. This is the second trap: the failure to identify
a connection that will perform what the argument requires. For example, if the
claim is made that necessarily (other things being equal) an action done con-
trary to a moral belief produces remorse or guilt, then unless remorse or guilt
can be identified as something further than attitudes felt toward actions that
the person believes himself to have performed and that he believes to have
been wrong, nothing to the purpose has been achieved. It would be like
proving that the belief that a proposition is true cannot be a realistic belief,
because it has this connection with the will: that upon considering a past action
of stating the negation of a proposition believed true, one must feel the atti-
tude where is that attitude felt toward the past production of falsehood.
A better approach might be to try to identify this connection by consider-
ing the place that a moral belief holds in practical reasoning. The best state-
ment of this for the purpose of the argument would, I think, go something like
this. Consider the situation in which a person X is wondering whether to do A
or to do B. Suppose a person Y tells him that A is the right thing to do.
114 Ethics

Suppose that X believes Y. Then it is logically necessary that this belief is

relevant to his decision. On the other hand, there can be no realistic belief of
which this is necessary, for, as Hume saw, the relevance of belief that some
state of affairs obtains to a decision is always contingent upon the existence of
a desire whose fulfilment that state of affairs affects. Now this argument, as its
antiquity suggests, is not easy to assess, and I shall not say by any means
everthing that ought to be said about it. But the central difficulty is clear. To
say that a belief is relevant to a decision is, analytically, to say that it ought to
be taken into account if the decision is to be the right one. So the first premise
of the argument says that it is logically necessary that the belief (that A is the
right thing to do) ought to be taken into account if X's decision is to be the
right one. Now this may be logically necessary, but only as it is logically
necessary that the belief that A is the cheapest thing to do ought to be taken
into account if the decision is to be the cheapest one. In short, the production
of this argument serves only as a prelude to an explanation of the analyticity
above, for otherwise it merely underlines connections between moral con-
cepts. So until such an explanation is produced, this approach cannot be said
to have produced the required connection between morality and the will.

II

There are, then, obstacles in the way of disproving moral realism by consider-
ing assent to a moral proposition. What I now want to do is to present an
argument against moral realism that starts directly from the notion of the
truth of a moral proposition. The form of the argument is this. I shall first
describe two properties of moral truth, and then try to show how, jointly, they
provide an insuperable difficulty for a realistic theory. The argument is not
original,5 but so far as I know it has never received a clear or detailed presenta-
tion in print.
The first property is that of supervenience. It is widely held that moral
properties are supervenient or consequential upon naturalistic ones. The gen-
eral notion of supervenience is capable of various slightly different expres-
sions. We can take first:

(S1) A property M is supervenient upon properties N1 . . . Nn if M is not

identical with any of N1, . . . Nn, nor with any truth function of
them, and it is logically impossible that a thing that is M should
cease to be M without changing in respect of some member of
N,... Nn.

This is equivalent to saying that a thing ceasing to be M strictly implies that it

has changed in respect of some member of Nl . . . Nn. So a contraposed
version says that a thing remaining the same in respect of all members of

5. It was given to me in conversation by Dr. C. Lewy.

 Moral Realism 115

N1 . . . Nn strictly implies that it does not cease to be M. However, the notion

of supervenience that we want in ethics should be slightly stronger than this,
for moral properties are capable of degree, and we should, if we believe in the
supervenience of moral properties, also hold that a thing cannot come to
possess them, or come to possess them in a greater or lesser degree than
hitherto, without changing some of its naturalistic properties. So a better
version is:

(S) A property M is supervenient upon properties N1 . . . Nn if M is not

identical with any of N1 . . . Nn nor with any truth function of them,
and it is logically impossible that a thing should become M, or cease
to be M, or become more or less M than before, without changing in
respect of some member of Nl . . . Nn.

Again, a contraposed version says that a thing remaining the same in respect
of all members of N1 . . . Nn strictly implies that it remains the same in point of
possession, lack of possession, or degree of M-ness.
(S) defines a notion of a supervenience of a property applying to one thing
over a period of time. But we also believe that if two things are the same in
their naturalistic properties, then it follows that they are identical in their
moral properties, that is, have the same moral worth. This notion of super-
venience can be defined:

(S2) A property M is supervenient2 upon properties Nt . . . Nn if M is

not identical with any of N1 . . . Nn or with any truth function of
them, and it is logically impossible that two things should each
possess the same properties from the set N1 . . . Nn to the same
degree, without both failing to possess M, or both possessing M, to
the same degree.

For our purposes we need not distinguish between supervenience and

supervenience2. It is difficult to think of a property that might be supervenient
upon some others in the sense of (S) but not supervenient2 upon those others
in the sense of (S2), and I can think of no argument that might be used to show
that moral properties are supervenient upon naturalistic ones in one sense but
not the other. So from this point, when I talk of supervenience, I shall take it
to include both notions, and the claim that moral properties are supervenient
upon naturalistic ones embraces the claims that they are supervenient both in
the sense of (S) and in the sense of (S2). It should be noted that (S) and (S2)
define supervenience in terms of logical impossibility, and it is a logical claim
about the interrelations of sets of properties that is being made. It may, for
example, be a physical impossibility that two materials be identical in their
crystalline properties but one strong and the other not, or a biological impossi-
bility that two twins come from the same ovum but one be a zebra and the
other not. But neither of these impossibilities indicates supervenience in the
sense that I have defined, for neither is a logical claim. Similarly, if it were
116 Ethics
only a moral impossibility that two things should be identical in naturalistic
properties but different in moral worth, then moral worth would not be super-
venient upon naturalistic properties in my sense. One thing, then, that must
be established in defending this part of the argument is that if somebody
claimed, say, that an action was absolutely identical in every respect with
another, except that it was much worse; or that a feature of character like
courage had changed in no way in its nature, relations, consequences, yet was
of much less value than formerly; it would be a logical and not merely a moral
mistake that had been made.
The other feature of moral truth that is needed to support the argument is
even more commonly believed in than its supervenience. It is that the posses-
sion of moral worth is not entailed by the possession of any set of naturalistic
properties whatsoever, in any degree whatsoever. This lack of entailment has
often been claimed, and often disputed, and often confused with other claims,
such as that moral properties are not identical with naturalistic ones, or that
moral properties have no criteria in a Wittgensteinian sense. So the precise
claim that is being made is this:

(E) There is no moral proposition whose truth is entailed by any propo-

sition ascribing naturalistic properties to its subject.

This is not the claim that moral properties are not identical with any naturalis-
tic ones. For although such an identity would ensure an entailment both from
naturalistic to moral propositions and vice versa, to suppose that the existence
of the entailment would ensure the identity is simply to indulge in wishful
thinking. There are counter-examples to the thesis that Fa only entails Ga if F
is identical with G or some conjunct of which G is a member, which have
never been satisfactorily explained away, and it cannot be confidently as-
sumed at the outset that this would not be one of them. Of course Moore's
concern in Principia Ethica was to disprove the thesis that there is an identity,
not the thesis that there is an entailment.
Second, (E) is not the claim that there are no naturalistic properties that
are necessarily reasons for an ascription of a moral property. To say that P is
necessarily a reason for Q is6 to say that, necessarily, coming to know P ought
to increase one's confidence in Q. Now suppose that there exist naturalistic
properties such that necessarily coming to know that a thing possesses such a
property ought to increase one's confidence that it is good. It would by no
means follow that (E) is false. For it does not follow that any statement that a
thing possesses one of these properties can be given that is a conclusive reason
for the thing being good, in the sense that having come to know that state-
ment, whatever else one learns about the thing in question, one is right to be
certain that it is good. Whereas if an entailment exists, that is precisely what
can be done, for if P entails Q, then the conjunction of P with any proposition
whatsoever entails Q. This sort of consideration should be familiar from other
6. With minor complications.
 Moral Realism 117

areas of philosophy. For example, it is very plausible to suppose that some

statements about what seems to be the case to some observers are necessarily
reasons for supposing certain things to be true of the external world. That is, it
is plausibly supposed to be necessary, and not contingent (for what could it be
contingent upon?), that something's seeming to be yellow under appropriate
conditions to an apparently normal person is a reason for supposing it to be
yellow. But few people are prepared to believe that such statements about
what seems to be the case can entail that something is the case. A better
example comes from consideration of induction. It is very plausible to sup-
pose that, necessarily, the knowledge of the existence of certain past regulari-
ties ought to increase one's confidence in certain appropriately related predic-
tions. Yet there is no entailment between the evidence and the prediction.
Now it is no part of my purpose to claim that these positions about perception
and induction are correct in affirming that some things are necessarily reasons
for others while no entailment can be produced. But the evident plausibility of
supposing this is sufficient to show that claims about lack of entailment must
be very sharply separated from claims about lack of a necessary reasoning
relationship, and that it cannot be safely assumed that argument against the
latter claim will transfer to the former. A fortiori, one must keep the claim
that, necessarily, reasons for moral judgements must come from certain areas
(facts concerning human interests, for example) entirely distinct from the
claim that there is an entailment from some propositions describing things in
those areas, to a moral proposition.
Third, (E) is not the denial that there exist criteria, in a Wittgensteinian
sense, for the ascription of moral properties. What is meant by saying that P is
a criterion, in a Wittgensteinian sense, for Q is not at all clear. It certainly
doesn't mean just that P is necessarily a reason for Q, for there are cases
where it is plausible to suppose that one proposition is necessarily a reason for
another, yet extremely unnatural to call that proposition a criterion, in any
sense, for the other. Nor would anyone who believes in the necessary reason-
ing relationship want to express it in this way. I have already given one
example where this is true, in pointing out that perhaps certain facts about the
past necessarily provide a reason for certain predictions; for surely nobody
would want to express this by saying that perhaps some facts about the past
are criteria, in a Wittgensteinian sense, for the truth of certain predictions.
Again, for at least most values of P and Q, P is necessarily a reason for the
conjunction P&Q. Yet surely P is not therefore a criterion, in any sense, for
the truth of P&Q. Or, to take a familiar case, just as some philosophers claim
that someone's exhibiting certain behaviour is necessarily a reason for suppos-
ing that he is in pain, so the knowledge that some unobserved person was in
pain at a certain time may be necessarily a reason for supposing that he
exhibited certain behaviour at or around that time, but it is obviously not a
Wittgensteinian criterion for us exhibiting that behaviour. 7
7. Strictly speaking, the propositions that might be thought necessary are not of the form
'coming to know P ought to increase one's confidence in Q' but rather 'If the situation is (in some
118 Ethics

Perhaps some condition limiting the type of fact with which P and Q are
respectively concerned can be added to the condition that P is necessarily a
reason for Q to give an acceptable sense of 'criterion'. The result would be a
conjunctive definition: 'P is a criterion of Q if and only if P is necessarily a
reason for Q and knowing the sort of fact that P describes is the only way of
knowing propositions like Q' This would be a plausible attempt, and one that
would give the result that if P described a naturalistic fact, and Q was a moral
proposition, and P was necessarily a reason for Q, then P would be a criterion
of Q. But then the question of whether there are naturalistic criteria for moral
propositions would in effect be the question of whether there are necessary
reasoning relationships, and, as we saw, this question must not be confused
with that of the truth of (E).
These preliminary remarks serve only to distinguish (E) from other claims
that a naturalist might wish to dispute; the important point is that (E) denies
the existence of an entailment, not of a 'logical connection' of a weaker sort,
nor of logical constraint on the areas from which morally relevant consider-
ations must be adduced. This much said, I shall postpone further discussion of
(E) and (S) until I have described the difficulty that they provide for moral
realism. For it will turn out that a principle in some respects weaker than (E)
can serve as a basis for the argument, so that although (E), being a familiar
claim to moral philosophers, may provide the clearest starting point, the anti-
realist might rest his position on something less controversial.
What, then, is the difficulty for realism that the lack of entailment de-
scribed in (E) and the supervenience described in (S) and (S2) jointly prom-
ise? Suppose that we ask a moral realist to describe his position, showing it to
be compatible with the lack of entailment and supervenience. He has to say
that the truth of a moral proposition consists in the existence of a state of
affairs, which it reports; that the existence of this state of affairs is not entailed
by the existence of other, naturalistic facts; yet that the continuation of these
facts entails that the moral state of affairs continue as it is. Now this may at
first sight seem harmless enough, and perhaps it is not actually inconsistent,
but it is very mysterious. To make the preculiarity of the view evident we can
put it like this. Imagine a thing A, which has a certain set of naturalistic
properties and relations. A also has a certain degree of moral worth: say, it is
very good. This, according to the realist, reports the existence of a state of
affairs: A's goodness. Now the existence of this state of affairs is not entailed
by A being as it is in all naturalistic respects. This means, since all the proposi-
tions involved are entirely contingent, that the existence of this state of affairs
is not strictly implied by A being as it is in all naturalistic respects. That is, it is
logically possible that A should be as it is in all naturalistic respects, yet this
further state of affairs not exist. But if that is a logical possibility, why isn't it a

respects) normal, coming to know P ought to increase one's confidence in Q'. For example, it
could at best be 'If there is no reason to suppose that he is pretending or abnormal, then coming
to know that he exhibited certain behaviour ought to increase one's confidence that he was in
pain', which could be supposed necessary.
 Moral Realism 119

logical possibility that A should stay as it is in all naturalistic respects, and this
further state of affairs cease to exist? If it is a logical possibility that A be as it
is in all naturalistic respects, and not be good, why isn't it a logical possibility
that A remain in all naturalistic respects as it was when it was once good, and
yet cease to be good? The existence of the naturalistic facts doesn't guarantee,
logically, the moral state of affairs, so why should their continuation give a
logical guarantee of the continued existence of the moral state of affairs?
Again, these questions can be put in terms of (S2), giving the puzzle for the
realist: if A has some naturalistic properties, and is also good, but its goodness
is a distinct further fact not following from its naturalistic features, and if B
has those features as well, then it follows that B also is good. And this is a
puzzle for the realist, because there is no reason at all, on his theory, why this
should follow. If the goodness is, as it were, an ex gratia payment to A, one to
which A is not as a matter of logic entitled in virtue of being as it is in all
naturalistic respects, then it should be consistent to suppose that although
goodness was given to A, it was not given to 5, which merely shares the
naturalistic features that do not entail the goodness.
It may not, at first sight, be clear why these problems are particularly acute
for the realist. If joint supervenience and lack of entailment are problematic
for any theory, then it is no objection to realism that it finds them particularly
hard to explain. But the situation is not like that. For although the correspon-
dence theory of truth, of which moral realism is an instance, is often thought
to be entirely vacuous, it does in fact offer a distinct picture of the truth of
propositions: each proposition may be construed as asserting that a certain
state of affairs exists (namely, that state of affairs in whose existence or
nonexistence the proposition's truth or falsity consists). And, however grave
for some purposes the circularity introduced in the parenthesis, the definite
consequence is that the truth of propositions should be subject to just the
logical constraints that govern the existence of states of affairs. But this is
exactly what does not happen with moral truth, where we are asked to make
intelligible the notion of a state of affairs subject to the constraint that its
existence does not follow from the naturalistic facts being as they are but its
continued existence follows from the natural facts staying as they are. Now,
while I cannot see an inconsistency in holding this belief, it is not philosophi-
cally very inviting. Supervenience becomes, for the realist, an opaque, iso-
lated, logical fact for which no explanation can be proffered.
We can now see why a claim weaker than (E) will support the argument
equally well. (E) asserts that no proposition ascribing a degree of worth to a
thing is entailed by any naturalistic proposition, however complicated. But all
we need for the argument is that there are some propositions ascribing worth
to things which are not entailed by any naturalistic proposition, however
complicated. For if there are some such moral propositions they nevertheless
still ascribe properties that are supervenient upon the naturalistic facts, and
their supervenience is again an objectionable peculiarity for the realist. Of
course, if we weaken the premise in this way, it is theoretically open to a
realist to say that some moral truths are entailed by naturalistic facts, and his
120 Ethics

realism is a theory about these, and others are not, and his realism does not
apply to these. But in that case the position has the disadvantage of simply
postulating a large but hitherto unsuspected dichotomy in moral truths, and
then leaving one half (the interesting half) of this dichotomy completely unex-
plained. Since this is not very attractive, the argument is just as strong if we
use as one premise, instead of (E):

(E') There are some moral propositions that are true, but whose truth
is not entailed by any naturalistic facts about their subject.

This is quite a significant strengthening of the argument, for using this premise
we need not defend ourselves against the view that in some cases 'is' does
entail 'ought'. All we hold in adopting (E') is that in some cases it does not,
yet that the 'ought' proposition can be a moral truth all the same. Thus J. R.
Searle's well-known argument8 that in some cases there are purely naturalistic
criteria for undertaking an obligation, and that sometimes it can follow from
an obligation having been undertaken that, other things being equal, it ought
to be fulfilled: even if it convinces us that there is an entailment and that the
'ought' in the conclusion is a moral 'ought' (so that someone who considers
the fact that he promised to do A irrelevant to the decision whether to do A or
not A is illogical rather than immoral) this argument is beside the point. For
there is no prospect at all of extending these considerations to cover all moral
truths. Searle's argument, as he himself notices, can cover only moral proposi-
tions that are true because of certain institutional facts. When an action is
described truly in the terms appropriate to an institutionalized activity ('He
was clean bowled'), it would seem to follow that consequences which the rules
of that activity prescribe ought to be brought about ('He ought to stop bat-
ting'). But of course no morality is going to suppose that all moral truths are,
in this way, institutional. For a start, that would make it impossible to raise a
moral question about the existence of the institution, and in any case there are
features of human life whose lack of desirability obviously is not a matter of
the rules of institutions.
Does the position sometimes called 'neonaturalism', associated with the
work of Philippa Foot, advance reasons for denying (E')? I do not think so. A
central claim of this view is that necessarily certain qualities—courage, jus-
tice, temperance—are qualities by which men come to act well: because they
are needed in any enterprise whatsoever, other than an accidental fulfilment
of any desire or need, anyone, whatever his moral nature, has to prize them.
It seems to me that there can be no need to deny this sort of claim in defending
(E'). For how can this view, even when allied with the view that human harm
and injury are necessarily bad and human benefit necessarily good, be used to
demonstrate every moral truth? Consider, for example, the view that a state
of satisfaction is a good thing only if its object is somehow appropriate, so that
it adds nothing to the worth of a state of affairs if, say, somebody achieves
8. J. R. Searle, 'How to Derive "Ought" from "Is" ', Philosophical Review, 1964.
 Moral Realism 121

content or satisfaction by contemplating or creating something totally unfit-

ting. This is an important moral view, underlying many people's discontent
with utilitarianism. And of course many people would reply that it is illiberal
and false, that although it is a pity that people should contemplate or create
unfitting things, it is nevertheless better, given that they do, that they should
be satisfied with them than not. Now, is it likely that, even allowing Foot's
claim that certain things are necessarily good, we shall find naturalistic facts
entailing that one of these views is true and the other false? I cannot see that it
is. Of course, it might be argued as follows: necessarily human benefit is good,
necessarily satisfaction is a benefit, so it follows from the fact that someone
has achieved some satisfaction that, whatever its object, something good has
come about. But this is a very weak way of trying to show the anti-utilitarian
to be illogical rather than immoral,9 for it denies that it is even true, let alone
necessary, that all cases of satisfaction are in any morally relevant sense a
benefit to their possessors. Satisfaction at inappropriate objects is, according
to this argument, something that should not be wished upon anyone, even if
the only alternative is dissatisfaction at those same objects.
Another line would deny that the notions of truth or falsity are applicable
in these cases, where no entailment exists. It would thus be possible to deny
(E') by denying that there exist any further moral truths beyond those that are
entailed by naturalistic facts. So neither the proposition that satisfaction is
good only if its object is appropriate nor its negation is true or false, and the
dispute between the utilitarian and his opponent is not to be construed as a
dispute about moral truth. But this is not only ad hoc, it is itself a serious
departure from realism. For one consequence of realism is that the law of the
excluded middle applies: either the state of affairs in which the goodness of
unfitting satisfaction consists exists or it doesn't. In the former case the propo-
sition that such satisfaction is good is true, and in the latter case, false. The
point is that the correspondence theory seems to allow no third alternative,
and quite obviously the picture of moral truth that the realist presents would
make the description of such an alternative a most unenviable task.
It seems then that (E') is reasonably secure, at least so far as recent
argument, which might certainly seem to be in opposition to it, is concerned.
The question now remains whether the other premise, that moral worth is
supervenient upon naturalistic properties, is indisputable. I doubt very much
whether anybody would feel entirely happy about disputing it, so even if there
is no very good argument for it that a realist need accept, it may be quite
proper to accept it as an axiom of meta-ethics. However, what can be done is
to show how certain natural ways of explaining the feature of supervenience
are not open to the realist, so that it is an objectionable factor for him but not
for others. We can also show that if, faced with this, he does try to deny the
necessity of supervenience, this can only be at the cost of making moral truths
unimportant. It has often been realized that moral realism runs this danger;
9. G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1903), chapter
VI.
122 Ethics

what I wish to do is to show how this escape route from the argument leads
directly to the danger. For an anti-realist, a natural way of explaining the
supervenience of moral properties would be something like this. There can be
no question that we often choose, admire, commend, or desire, objects be-
cause of their naturalistic properties. Now it is not possible to hold an attitude
to a thing because of its possessing certain properties and, at the same time,
not hold that attitude to another thing that is believed to have the same
properties. The nonexistence of the attitude in the second case shows that it is
not because of the shared properties that I hold it in the first case. Now, moral
attitudes are to be held towards things because of their naturalistic properties.
Therefore it is not possible to hold a moral attitude to one thing, believe a
second to be exactly alike, yet at the same time not hold the same attitude to
the second thing. Anybody who appears to do this is convicted of mis-
identifying a caprice as a moral opinion.
This line of thought is insufficient for the realist. We saw earlier that he
need not refrain from talking about moral attitudes: it is just that he thinks
this a less clear alternative to talking about moral beliefs. But if we rephrase
the preceding paragraph in terms of moral belief, it is obviously insufficient to
explain supervenience. For we obtain: we hold moral beliefs about things
because of their naturalistic properties; it is not possible to hold a belief about
one thing because it satisfies some condition, and, at the same time, not hold
that belief about another thing that is believed to satisfy the same condition;
therefore, it is not possible to hold a moral belief about a thing, believe a
second to be exactly alike in all naturalistic respects, yet at the same time not
hold the belief about the second thing. But this doesn't explain supervenience
at all: it merely shows the realist putting conditions upon what can be believed
to be the truth, not upon what is the truth. Our belief, he is saying, has to be
consistent across naturalistic similarities—but this is no explanation of why,
on his theory, the truth has to be. Furthermore, once we have grasped the
inadequacy of his picture to account for supervenience, we may doubt his
right to help himself to conditions upon belief that mean that beliefs are to be
constrained as though supervenience were true. Why, that is, should a realistic
theory accept that things believed to be naturalistically alike cannot be be-
lieved to be of distinct worth? Surely the condition on belief is that it should
be true, and the trouble with inconsistency is that it is false: but the realist, as
we have seen, offers no explanation of why this sort of inconsistency in moral
belief should yield falsehood.
Finally, suppose the realist takes the apparent consequence of his position
seriously, and denies supervenience. Then he holds that it is possible that the
worth of, say, a feature of human life, such as courage, should alter, although
its intrinsic nature, consequences, relation to our desires, and so forth remain
the same. The only conclusion is that it is possible that the worth of courage
should be irrelevant to our interest in it, attitudes toward it, or preparedness
to urge it on our children or to criticize or commend someone for possessing
it. For when moral truth is that pure there can be no reason for being inter-
ested in it—nothing hangs upon the worth of courage changing if its relation
 Moral Realism 123

to everything perceptible remains the same, and no reason could possibly be

given for being interested in this fact.
Moral realism, then, is false. The purpose of the next section is to intro-
duce the consequences of this for the logic of moral discourse. For it is natural
to suppose that the falsity of realism entails the falsity of any prepositional
account of moral discourse. For this reason many people might suppose that
the results of section II could better be expressed as a disproof of the thesis
that moral utterances express propositions, or of the thesis that goodness is an
attribute. But this is too hasty. If propositions are what can be true, false,
believed, hypothesized, or premises of arguments, or if they are what can
enter into entailment relations, then it is not at all attractive to suppose that
the statement 'Courage is an intrinsically good thing' expresses no proposi-
tion, even if the only natural theory about the truth of that proposition is
certainly false.

Ill
The argument of section I showed that it is easy to overestimate the amount
that we gain from talking in terms of moral attitudes rather than moral beliefs.
However, in section II, I claimed that an anti-realistic theory does have a
distinct advantage in being alone consistent with supervenience and lack of
entailment, and I sketched a way in which a theory based on attitudes might
account for the former phenomenon. However, this was just a sketch, and we
must now see whether it could be successful by describing the connection that
an anti-realist must draw between moral attitudes and moral predicates. The
problem is this. It is easy to think that the meaning of moral predicates has
been given if it is said that they are used in sentences that when uttered
assertively, express the speaker's attitude to the subject of the discourse. At
any rate, if this is true, and if a suitable account of the attitude is given,10 it is
easy to think that all that is needed to give an analysis of moral discourse has
been done. However, it has been pointed out by J. R. Searle11 and others that
this is not so. For such a description of the use of moral sentences leaves us
completely in the dark as to what happens when such sentences are used not
assertively, but in less direct contexts. For example, suppose we are told that
the sentence 'Courage is an intrinsically good thing' is standardly used so that
in asserting it a speaker is expressing his attitude, namely approval, to cour-
age. This does not, it is argued, enable us to construe: 'If courage is an
intrinsically good thing, then organized games should be a part of school
curricula', or 'No statement of the naturalistic properties of courage entails
that courage is an intrinsically good thing', since in neither of these is any
more value attributed to courage, nor is any attitude to it expressed. So the
10. This is a problem to which this paper makes no contribution. For an investigation of it,
see Roger Scruton, 'Attitudes, Beliefs and Reasons', in Morality and Moral Reasoning, ed. J.
Casey (London: Methuen, 1971).
11. 'Meaning and Speech-acts', Philosophical Review, 1962, pp. 423 ff.
124 Ethics
problem, for an anti-realist, is that of showing how the original insight as to
what is done when a moral proposition is asserted also gives him an explana-
tion of what is done when the sentence expressing it occurs in such contexts.
There are, I think, two natural ways for the anti-realist to attempt to meet
this challenge, and it can be shown that neither of them will do. The first is to
suppose that in an indirect context12 some proposition about an attitude is
being hypothesized, said to be entailed, or otherwise involved. Now this is
clearly unnatural, because the anti-realist is not claiming that in a direct
context, when asserted straightforwardly, moral sentences express proposi-
tions about attitudes. This would be to hold a realistic theory, if the proposi-
tion expressed was supposed to be one simply claiming the existence of certain
attitudes, or to fail to give a theory of moral discourse at all. If, for example,
the view was that in asserting that courage is intrinsically good one is asserting
that an attitude of approval to courage is appropriate, then this, although true,
is unhelpful. For the judgement that approval is appropriate is a moral judge-
ment, and giving it as a synonym for the original removes no problems of
analysis or epistemology that the original causes. So if an anti-realist tried to
answer Searle's point by saying that what is hypothesized is 'If an attitude of
approval to courage is appropriate, then . . . ', no ground is gained at all, for
still nothing has been done to explain how moral judgements can occur in
indirect contexts. The same point applies to more complicated versions.
Urmson, discussing only judgements of something as good of a kind or from a
point of view, suggests that 'good' means 'satisfies a description such that no
one can correctly dissent from a favourable evaluation (as of some kind or
from some point of view) of an object that satisfies that description'.13 But if
this were extended to try to give a theory of moral judgement, precisely the
same objection would apply: to say that nobody can correctly dissent from a
favourable evaluation of a good thing is true but completely unhelpful.
The other thing that an anti-realist might try to do is to suppose that in an
indirect context a naturalistic proposition is being hypothesized, said to be
entailed, or otherwise involved. Which naturalistic proposition would, I take
it, be determined by the moral views of the person who utters the sentence
expressing the hypothetical, or other compound proposition. Thus take

(H) If courage is an intrinsically good thing, then organized games

should be a part of school curricula.

This would be a sentence that is interpreted on a particular occasion of utter-

ance by our knowing the moral standards of the man uttering it and supposing
that he is hypothesizing that courage has the naturalistic properties that he
would give as reasons for its being intrinsically good. Again, there is difficulty
in reconciling this view of indirect contexts with the view that it is not a
12. I do not, fortunately, need to embark on the analysis of the notion of an indirect context.
For the purpose of the argument it is just taken to embrace those contexts in which the moral
proposition is expressed, but not asserted.
13. J. O. Urmson, The Emotive Theory of Ethics (London: Hutchinson & Co., 1968), p. 142.
 Moral Realism 125

naturalistic proposition that is asserted when it is said that courage is a good

thing. For if you and I, with different standards, each assert (H), we have
hypothesized different things, but then, unless in asserting that courage is
intrinsically good we assert different things, at least one of us is construed as
hypothesizing something other than what he asserts, which makes it difficult
to see how modus ponendo ponens is valid when used on moral propositions.
Also some hypothetical will become tautologous when they have a moral
antecedent and naturalistic consequent, namely those in which the naturalistic
consequent gives the facts that the speaker would use as his standards for the
truth of the moral antecedent. Thus, 'If courage is intrinsically good, then it is
a quality by which men must act to achieve happiness' might be tautologous as
uttered by me. But it doesn't look much like a tautology, for disagreement
with it seems to be disagreement with a substantive moral claim, a claim about
the necessary conditions of intrinsic worth of a quality, rather than the sort of
disagreement about the use of words that disagreement with a tautology
usually involves. In short, if an anti-realist takes this view of indirect context,
he is opening himself to the classical, Moorean objection to naturalism, and
the apparent sophistication of the position renders it no less vulnerable.
The trouble with both these attempts to escape from Searle's argument is
that they try to answer the question of what is being hypothesized, said to be
entailed, and so on, and we have a definite view of how to set about questions
of this sort. We try to give an analysis of the moral proposition, and this turns
out to be either incorrect or unhelpful. Instead, I propose to consider the total
sentence expressing the hypothetical, or entailment, and to give a theory of
what these say and why we have these locutions to say it. The nature of the
theory can best be presented via an example. Consider:

(I) No statement of the naturalistic properties of courage entails that it

is an intrinsically good thing.

Here is a sentence in which the proposition that courage is an intrinsically

good thing is expressed but not asserted, so what can an anti-realist theory,
based on the way in which an attitude is expressed when a moral proposition is
asserted, say about it? The answer is given if we extend Frege's theory of such
contexts to sentences that express attitudes. Frege thought that sentences
occur in contexts in which they refer to propositions that they normally ex-
press. I suggest that in (I) we refer to an attitude: an attitude is the subject of
the proposition expressed by (I), and it is quite easy to see what is said about
that attitude. For, according to an anti-realist, (I) is not asserting the logical
independence of the existence of two states of affairs, the moral and the
naturalistic. It is an expression of a fact about the moral attitude of approval
toward courage, and says that nobody who fails to approve of courage while
knowing all that there is to be known about it can be convicted of a logical
mistake. This is a claim about the nature of the moral attitude, and the
statement (I) is what I shall call a 'prepositional reflection' of this claim. By a
'prepositional reflection' I mean roughly any statement that, while appearing
126 Ethics

to make a factual claim about states of affairs, their interrelations, and their
logic, is actually making claims about attitudes, although none of the proposi-
tions involved in the statement is to be analyzed into one whose subject is an
attitude. I am afraid that this is probably quite obscure, and I shall not, in this
essay, make the notion nearly as clear as I think it could be made, but further
examples of the use of the notion should help.
If we turn again to the anti-realistic explanation of (S), we can see that it
shares with the explanation of (I) the feature that an attitude—the attitude of
moral approval—is said to have certain properties, and this by itself is the
truth of which (S) is, in the above sense, a propositional reflection. Thus, the
moral attitude is said to be necessarily held because of the naturalistic proper-
ties of its objects, and the statement of supervenience, made in terms of which
differences entail which others, is a realistic-appearing way of putting the view
that difference in moral attitude to two things must, logically, be justified by
difference in beliefs about them. So the theory, if it works for other examples
of indirect contexts, does give an account of the relation between moral
predicates and moral attitudes that allows the anti-realist explanation of super-
venience to be successful.
The idea fares well when we consider hypotheticals again. A hypothetical,
such as (H), although it appears to be making a claim that one state of affairs
exists if another does, must be taken as a propositional reflection of a claim
about attitudes. This claim is that an attitude of approval to courage in itself
involves an attitude of approval to organized games as part of the curriculum
in every school. It does not, of course, involve this as a matter of logic, but
then neither is (H) true as a matter of logic. To show that one attitude does
involve the other it is necessary to show that organized games are intimately
connected with the production of the quality of courage and lack other dis-
advantages. This is precisely what would have to be shown to verify the
original hypothetical (H). If we attack the claim that one attitude involves the
other ('Surely not—organized games at school promote cruelty and coward-
ice'), we produce just the propositions that are reasons against the original
hypothetical. Clearly with moral as with natural discourse there is a variety of
grounds on which an 'if ... then . . .' proposition might be held, but this
elasticity is preserved by varieties of involvement. Consider, for example,
logically necessary propositions using moral antecedents: 'If you ought to
divide the pound evenly among ten people, then you ought to give them two
shillings each'.14 This is necessarily true. The analysandum that refers to
attitudes must therefore be necessarily true also. And 'Approval of dividing
the dollar evenly among ten people involves approval of giving them twenty
cents each' can easily be given a sense in which it expresses a necessary
truth—namely, that one attitude cannot be consistently held without the
other. This logical constraint upon consistency in attitude must not, of course,
be thought to derive in any way from the logical interrelations of moral

14. An example suggested by Tom Baldwin, who most helpfully critieized an earlier version
on this point.
 Moral Realism 127

propositions that it analyses. It is rather explained by pointing out that it is

logically impossible to divide a dollar evenly among ten people without giving
each of them twenty cents, so for this type of example the analysis can show in
a genuinely explanatory way how the hypothetical refers to attitudes and
expresses their relations.
The validity of modus ponendo ponens as a rule of inference is preserved
by this treatment of moral hypothetical. The question, for the anti-realist, is
that of how his account of the hypothetical is to cohere with his account of
straightforward assertion in such a way that 'P, and if P, then Q' entails Q,
where P is a moral proposition. The theory tells us that anybody asserting 'P,
and if P, then Q' where P attributes worth to a thing expresses his attitude to
that thing, and asserts that that attitude involves a further attitude or belief.
There is, when that has been done, a logical inconsistency in not holding the
further attitude or belief. It is this logical inconsistency that is expressed by
saying that modus ponendo ponens is valid: its validity is a reflection of
possible logical inconsistency in attitudes and beliefs.
A moral proposition can be the antecedent of a hypothetical of which the
consequent is not a moral proposition. Here the claim of which this form is a
prepositional reflection is that the attitude involves a belief, not another
attitude. 'If courage is an intrinsically good thing, then it is a quality by which
a man must act to achieve happiness' is a reflection of the view that an attitude
of approval to courage in itself involves belief that courage is a quality by
which a man must act to achieve happiness. This is a statement of one's moral
standards: only someone with a specific moral position will hold that the
attitude involves that belief, and this reflects perfectly well the point that I
made earlier, that such a hypothetical is a statement of a standard, and if
attacked is attacked as such and not, for example, as a sort of contradiction or
empirical falsehood.
It may be thought at this point that too much weight is put upon the word
'involves'. For how can an attitude involve a belief (or a belief an attitude) in
such a way? We recognize that people can hold the moral attitude without the
belief and vice versa: what other involvement is there? To say that the attitude
of approval to courage involves the belief that courage is a quality by which a
man must act to achieve happiness is, as I said, to express a moral standard, to
make a moral claim. The subtle thing is that the subject of the claim is now the
attitude of approval of courage, and it is said that this ought to involve the
belief that courage involves happiness. It is important to see that there is
nothing in the least troublesome about this being a moral proposition: the
object was not to show that such standard-giving hypotheticals are not moral
propositions, but to show how they could be so. And they are quite straight-
forward moral propositions expressing the way in which attitudes ought to
vary with beliefs.
This analysis has the further property that it enables us to deal with expres-
sions of moral fallibility, in my opinion the hardest context of all for an anti-
realist to understand. Consider I strongly believe that X is a good man, but I
may be wrong'. Here we express an attitude to X, but then what happens? For
128 Ethics

a realist, we simply state the possibility that our belief does not accord with
the facts, but what can an anti-realist say about it? It is easy to say that here I
express an attitude to my moral attitude to X, and of course I do, but the
question is whether this second-order attitude can be described consistently
with anti-realism. For if we have to describe the second-order attitude just as
'belief in the possibility that the attitude of approval to X is unjustified', then
we are in no better position than those who rest their analysis on attitudes
being 'appropriate'. But let us consider what we envisage when we admit that
we may be wrong in a moral belief. We envisage that we could, at some future
date, want to change our attitude to the subject of the belief, not because the
subject has changed, but because we realize that we have made a mistake.
The mistake could be one of two kinds. Taking the example of 'X is a good
man', we can envisage that we may have been mistaken about X: he isn't
really kind, he only likes to appear so, and so forth. Or I can envisage a
mistake in my standards. I think that a man's being kind is a reason for
thinking that he is a good man. That is, I think that the belief that a man is
kind ought to tend to create an attitude of approval towards him. This, as I
explained above, is another moral belief, whose subject is the variation of an
attitude with a belief. But I may be mistaken here, too: I can envisage, at least
as a bare possibility, that some argument, or someone, or some event, should
come along and show me that I am mistaken about kindness: some sorts of
kindness kill the soul of the recipient, and so forth. Then I no longer think
that the belief ought to tend to create the attitude. The prepositional reflec-
tion of all these possibilities of errors is that the belief that something is good,
although I may hold it, may be false. And thus, if I am right, a theory based
upon attitudes can encompass those elements in moral thought that make us
reflect on how difficult it is to know what is right and wrong, good or bad.
Finally we may consider tenses. Tenses can certainly provide extreme
difficulty for analysis that concentrate upon what is done when something is
asserted to be the case now: thus the Toulmin account of 'It is probable that
F, according to which this is used to make a guarded assertion that F pro-
vides no account of 'It was probable that P would happen'—or, at least, no
account that does not confuse it with the very different 'It is probable that P
happened.' And if we consider 'X was a morally fine man', it might appear
that we have no option but to construe this as The attitude of approval
toward X was appropriate'—an analysis of the form that I claimed to be
totally unhelpful. However, a device is available that, although it may look
unnatural, appears to me to take account of tensed assertions without having
to view them as applying tensed predicates to attitudes. The question we must
ask is whether anybody who asserts 'X was a morally fine man' is expressing a
moral attitude. It is from the supposition that he is not that difficulty arises,
just as difficulty arises when it is realized that by saying 'It was probable that P
would happen' I am not guardedly asserting anything. But anybody asserting
'X was a morally fine man' is expressing a moral attitude. Not, it is true, to X
as he is, for X may in the interim have degenerated. But he is expressing the
attitude to X as he was. Quite clearly, anyone saying 'X was a morally fine
 Moral Realism 129

man' is sincere only if he approves of X as he was, morally, and in making his

assertion it is that approval which he expresses. The difficulty is thus averted
by realizing that the subject of approval when a tensed assertion is made is not
the object as it is now, but the object as it was then.
I have now given several examples of the device of prepositional reflec-
tion, the ways in which expressions of attitude and propositions concerning
the interrelations of attitudes with each other and with beliefs are given a
syntax that makes them appear to relate to facts in a peculiar, unobservable,
moral realm. It must in no way be considered surprising that this device
should exist. For disagreement in moral attitude is one of the most important
disagreements there is, and working out the consequences of moral attitudes,
one of the most important subjects there is. The device of propositional
reflection enables us to bring the concepts of propositional logic to this task. It
enables us to use notions like truth, knowledge, belief, inconsistency, entail-
ment, and implication to give moral argument all the structure and elegance
of argument about facts.
But isn't the theory saying that there is really no such thing as moral truth,
and nothing to be known, believed, entailed—only the appearance of such
things? Not at all. It is a complete mistake to think that the notion of moral
truth and the associated notions of moral attributes and propositions dis-
appear when the realistic theory is refuted. To think that a moral proposition
is true is to concur in an attitude to its subject: this is the answer to the
question with which I began the essay. To identify this attitude further is a task
beyond the scope of this essay, but it is the central remaining task for the
metaphysic of ethics. To think, however, that the anti-realist results show that
there is no such thing as moral truth is quite wrong. To think there are no
moral truths is to think that nothing should be morally endorsed, that is, to
endorse the endorsement of nothing, and this attitude of indifference is one
that it would be wrong to recommend, and silly to practise.15

15. I comment further on the complex issues of this essay in the next, and from a recent
perspective, in its Addendum.
 7
Supervenience Revisited

I
A decade ago, in an article entitled 'Moral Realism', I presented an argument
intended to show that two properties, which I called supervenience and lack of
entailment, provided together an unpleasant mystery for moral realism.1 This
argument originally was suggested to me in a discussion with Casimir Lewy,
which in turn was directed at a paper of G. E. Moore entitled 'The Concep-
tion of Intrinsic Value'.2 The intervening decade has provided a number of
reasons for revisiting my argument. First of all, it was couched in an idiom that
subsequent work on modal logic—particularly the distinctions of various
kinds of necessity and the general use of possible worlds as models—has made
a little quaint. It would be desirable to see if the new notions allow the
argument to stand. Second, we have seen a great deal of interest in super-
venience as a notion of importance beyond moral philosophy. Thus, in conver-
sation and correspondence I have heard it suggested that my argument must
be flawed, because exactly the same combination of properties that I found
mysterious occurs all over the place: for example, in the philosophy of mind,
in the relationship between natural kind terms and others, in the relation
between colours and primary properties, and so on. Since anti-realism in
these other areas is not attractive, this casts doubt upon my diagnosis of the
moral case. Last, moral realism is again an attractive option to some philoso-
phers; although when I wrote I might have seemed to be shadow boxing, the
argument is becoming relevant again. In any case, enough puzzles seem to me
to surround a proper analysis of supervenience to warrant a fresh look at it.
Suppose we have an area of judgements, such as those involving moral
commitments or attributions of mental states. I shall call these F judgments,
and I shall also talk of F truths and F facts: this is not intended to imply any
view at all about whether the commitments we express in the vocabulary are
beyond question genuine judgements, nor that there is a real domain of truths
or facts in the area. Indeed part of the purpose of my argument was to find a
1. S. Blackburn, 'Moral Realism', in Morality and Moral Reasoning, ed. J. Casey (London:
Methuen, 1971); essay 6, this volume. More than a decade ago now, alas.
2. G. E. Moore, 'The Conception of Intrinsic Value', in Philosophical Studies (London:
Routledge & Kegan Paul, 1922).

130
 Supervenience Revisited 131

way of querying just these ideas. At this stage, all this terminology is entirely
neutral. Now suppose that we hold that the truths expressible in this way
supervene upon the truths expressed in an underlying G vocabulary. For exam-
ple, moral judgements supervene upon natural judgements, or mental descrip-
tions of people upon physical ones (either of the people themselves or of some
larger reality that includes them). This supervenience claim means that in
some sense of 'necessary' it is necessarily true that if an F truth changes, then
some G truth changes, or, necessarily, if two situations are identical in point of
G facts, then they are identical in terms of F facts as well. To analyze this more
closely, I shall make free use of the possible-worlds idiom. But it must be
emphasized that this is merely a heuristic device and implies no theory about
the status of the possible worlds. Let us symbolize the kind of necessity in
question by N and possibility by 'P': for the present it does not matter
whether these are thought of as logical, metaphysical, physical, or other kinds
of modalities. We are now to suppose that some truth about a thing or event or
state, that it is F, supervenes upon some definite total set of G truths, which we
can sum up by saying that it is G*. Of course, G* can contain all kinds of
relational truths about the subject, truths about other things, and so on. In
fact, one of the difficulties of thinking about all this properly is that it rapidly
becomes unclear just what can be allowed in our conception of a totality of G
states. But intuitively it is whatever natural or physical states bring it about
that the subject is F. I shall express this by talking of the set of G states that
'underlies' an F state. Belief in supervenience is then at least the belief that
whenever a thing is in some F state, this is because it is in some underlying G
state, or is by virtue of its being in some underlying G state. This is the minimal
sense of the doctrine. But I am interested in something stronger that ties the
particular truth that a thing is F to the fact that it is in some particular G state.
We can present the general form of this doctrine as characterizing the relation
'U' that holds when one 'underlies' the other:

The formula says that as a matter of necessity, if something x is F, and G*

underlies this, then anything else in the physical or natural (or whatever) state
G* is F as well. There is no claim that G* provides the only way in which
things can become F: intuitively, we know that something might be, say, evil in
a number of different ways, and that something in one given physical state
might possess some mental property that it could equally have possessed by
being in any of a family of related physical states. The supervenience claim (S)
is thus in no opposition to doctrines that now go under the heading of 'vari-
able realization'. To get the claim that these doctrines deny, we would need to
convert the final conditional: . . But the resulting doctrine
is not one in which we shall be interested.
I now want to contrast (S) with a much stronger necessity:
132 Ethics

Of course, (N) does not follow from (S). Formally they are merely related like
this: (S) necessitates an overall conditional, and (N) necessitates the conse-
quence of that conditional. So it would appear there is no more reason to infer
(N) from (S) than there would be to infer Nq from N(p q). Hence also there
is no inconsistency in a position that affirms (S) but also affirms:

At least, this is the immediate appearance. In my original essay it was the

nature of theories that hold both (S) and (P) (which I shall call the (S)/(P)
combination) that occupied me. Such a theory would think it possible (in
some sense commensurate with that of the original claim) that any given G
state that happens to underlie a certain F state, nevertheless might not have
done so. In other words, even if some G setup in our world is the very state
upon which some F state supervenes, nevertheless, it might not have been that
F state that supervened upon it. There was the possibility (again, in whatever
modal dimension we are working) that the actually arising or supervening F
state might not have been the one that supervened upon that particular G
setup. My instinct was that this combination provided a mystery for a realist
about judgements made with the F vocabulary, and that the mystery would
best be solved by embracing an anti-realist (or as I now prefer to call it, a
projectivist) theory about the F judgements.
To pursue this further, we might question whether there could be any
motivation for holding the (S)/(P) combination. Consider the following possi-
ble doctrine about 'underlying', and about the notion of the complete specifi-
cation of an underlying state, G*:

The rationale for (?) would be this: Suppose there were a thing that was G*
and F, so that we were inclined to say that its being in the G state underlies its
F-ness. But suppose there were also a thing that was G* and ~ F. Then would
we not want to deny that it was x's being G* that underlies its being F, but
rather that it was its being G* and its being different from this other thing in
some further respect—a property that explains why the other thing fails to be
F? We can call that a releasing property, R, and then F will supervene only on
G* and ~ R. More accurately, G* would denote a set of properties that do not
really deserve the star. We would be wrong to locate in them a complete
underlying basis for F.
This raises quite complicated questions about the form of these various
doctrines. Let me put aside one problem right at the beginning. Since (S) is a
conditional and contains an existential clause as part of the antecedent, it will
be vacuously true if nothing is G* and F; the necessitation will likewise be
vacuously true if nothing could be G* and F. So if (S) captured all that was
meant by supervenience, we could say, for instance, that being virtuous super-
vened upon being homogeneously made of granite. Necessarily, if one thing
 Supervenience Revisited 133

homogeneously made of granite were virtuous, and this quality underlay the
virtuousness, then anything so made would be. But this is just because it is
impossible that anything of this constitution should be virtuous. I am going to
sidestep this problem simply by confining the scope of F and G henceforth to
cases where it is possible that something with a set of G properties, denoted
by G*, should be F. In fact, we are soon to deal with different strengths of
necessity and possibility, and I shall suppose that this thesis is always strong
enough to prevent the conditional being satisfied in this vacuous way.
The next problem of logical form is how we construe the denotation of a
set of properties made by the term 'G*'. First, we do not want the super-
venience thesis to be made vacuously true through its being impossible that
any two distinct things should be G*—it then following that if one G* thing is
F, they all are. And the threat here is quite real. If, for instance, G* were held
to include all the physical properties and relations of a thing—if it were that
and nothing less which some property F supervened upon—then assuming the
identity of indiscernibles, we would have (S) satisfied vacuously again. To get
around this I am going to assume a limitation thesis. This will say that when-
ever a property F supervenes upon some basis, there is necessarily a boundary
to the kind of G properties that it can depend upon. For example, the mental
may supervene upon the physical, in which case the thesis asserts that neces-
sarily there are physical properties of a thing that are not relevant to its mental
ones. A plausible example might be its relations to things with which it is in no
kind of causal connection (such as future things). Again, the moral super-
venes upon the natural, and the thesis will tell us that there are some natural
properties that necessarily have no relevance to moral ones—pure spatial
position, perhaps, or date of beginning in time. Given the limitation thesis,
(S) will not be trivialized by the identity of indiscernibles. The last problem of
form that arises is whether "G*' is thought of as a name for some particular set
of properties (which form a complete basis for F), or whether it is built into
the sense of 'G*' that any set of properties it denotes is complete. The differ-
ence is easily seen if we consider a very strong kind of necessity—say, concep-
tual (logical or analytic) necessity. It is unlikely to be thought analytic that
being made of H2O underlies being water. One reason for this is that it is not
analytic that being made of H2O exhausts the kind of physical basis that may
affect the kind to which a substance belongs. That is a substantive scientific
truth, not one guaranteed in any more a priori way. I am going to build into
the sense of 'G*' that in at least one possible world it denotes a set of proper-
ties sufficient to underlie F. I do not want it to follow that this is true in all
worlds, although that is a very delicate matter. Fortunately, so far as I can see,
it does not occupy the centre of the stage I am about to set.
If we accept (?) as a condition on what it is for a set of properties to
underlie another, and hence on what it is for a property to supervene upon
such a set, the relationship between supervenience and (E) changes. Suppose
that there is something whose G*-ness underlies its f-ness:
134 Ethics

then we can now derive (N). In other words, (?) and (E) together entail (N).
And as I have already said, (?) is an attractive doctrine. But it does mean that
supervenience becomes in effect nothing but a roundabout way of committing
ourselves to (N): the prima facie simpler doctrine that some set of underlying
truths necessitates the F truth. This is in fact the way that supervenience was
once taken by Kim; it enabled Kim to suppose that where we have super-
venience, we also have reductionism.3 Another way of getting at the attrac-
tions of (?) would be to cease altogether from mentioning the requirement
that there is something that is G* and F. After all, surely some moral property
might supervene upon a particular configuration of natural properties, regard-
less of whether there actually is anything with that set. Or some mental
property might supervene upon a particular physical makeup that nobody
actually has. If we took this course, we would replace (S) by a doctrine:

and then the doctrine that would give us (N) immediately would be

Yet supervenience claims are popular at least partly because they offer some
of the metaphysical relief of reductions, without incurring the costs; I want
therefore to preserve any gap that there may be for as long as possible. This is
particularly important in the moral case, where supervenience is one thing,
but reductionism is markedly less attractive. So I am going to stick with the
original formulation, subject to the caveats already entered, and while we
should remain well aware of (?), I do not want to presuppose a verdict on it.
If we put (?) into abeyance, we should be left with a possible form of
doctrine that accepts both (S) and (P): the (S)/(P) combination. It is this which
I originally claim to make a mystery for realism. If there is to be a mystery, it is
not a formal one, and I actually think that with suitable interpretations, there
are relations between F and G vocabularies that are properly characterized by
(S) and (P). It is just that when this combination is to be affirmed, I believe it
needs explanation. In the moral case I think that it is best explained by a
projective theory of the F predicates. But in other cases, with different inter-
pretations of the modalities, other explanations are also possible. I shall argue
this later.
Here, then, is a way of modelling (S) and (P) together. In any possible
world, once there is a thing that is F, and whose F-ness is underlain by G*,
then anything else that is G* is F as well. However, there are possible worlds
in which things are G* but not F. Call the former worlds G*/F worlds, and the
latter, G*/O worlds. The one thing we do not have is any mixed world, where
some things are G* and F, and some arc G* but not F. We can call mixed

3. J. Kim, 'Supervenience and Nomological Incommensurables', in American Philosophical

Quarterly, 1978, pp. 149-56.
 Supervenience Revisited 135

worlds G*/FvO worlds. These are ruled out by the supervenience claim (S):
they are precisely the kind of possible world that would falsify that claim. My
form of problem, or mystery, now begins to appear. Why should the possible
worlds partition into only the two kinds, and not into the three kinds? It seems
on the face of it to offend against a principle of plenitude with respect to
possibilities, namely that we should allow any that we are not constrained to
disallow. Imagine it spatially: Here is a possible world wl that is G*IF. Here is
another, w2, that we can make as much like wl as possible, except that it is G*l
O. But there is no possible world anywhere that is just like one of these,
except that it includes just one element whose G* and F properties conform to
the pattern found in the other. Why not? Or, to make the matter yet more
graphic, imagine a time element. Suppose our possible worlds are thought of
as having temporal duration. A mixed world would be brought about if wl
starts off as a G*/F world at some given time, but then at a later time becomes
a G*/O world. For then, overall, it would be mixed, and the supervenience
claim would be falsified by its existence. This kind of world then cannot
happen, although there can be worlds that are like it in respect of the first part
of its history, and equally worlds that are like it in respect of the second part of
its history.
This is the ban on mixed worlds: it is a ban on interworld travel by things
that are, individually, at home. The problem that I posed is that of finding out
the authority behind this ban. Why the embargo on travel? The difficulty is
that once we have imagined a G*/F world and a G*/O world, it is as if we have
done enough to also imagine a G*/FvO world, and have implicitly denied
ourselves a right to forbid its existence. At least, if we are to forbid its
existence, we need some explanation of why we can do so. The positive part
of my contention was that in the moral case, projectivists can do this better
than realists. In the next section I rehearse briefly why this still seems to me to
be so, if we make some important distinctions, and then I turn to consider
related examples. And in time we have to return to the difficult claim (?), to
assess its role in this part of metaphysics.

II

Necessities can range from something very strict, approximating to 'analyti-

cally true', through metaphysical and physical necessity, to something approxi-
mating to 'usually true'. Then anyone who sympathizes a little with the puzzle
in an (S)/(P) combination can quickly see that it will remain not only if there is
one fixed sense of necessity and possibility involved, but also in a wider class
of cases. For whenever the supervenience claim (S) involves a strong sense of
necessity, then it will automatically entail any version with a weaker sense of
necessity. Hence, we will get the same structure at the lower level, when the
possibility is affirmed in that corresponding, weaker, sense. Thus if (S) took
the form of claiming that 'it is metaphysically necessary that . . .' and (P) took
the form of claiming that 'it is physically possible that . . .', and if we also
136 Ethics

suppose (as we surely should) that metaphysical necessity entails physical

necessity, then we would have the (S)/(P) combination at the level of physical
necessity. On the other hand, we will not get the structure if the relation is
reversed. If the possibility claim (P) is made, say, in the sense of 'possible as
far as conceptual constraints go', this does not entail, say, 'metaphysically
possible'. And then if (S) just reads 'metaphysically necessary', there is no
mystery: we would have it just that it is metaphysically necessary, but not an
analytic or conceptual truth, that F supervenes upon G*; equally, it would not
be an analytic or conceptual truth that any given G* produces F, and there is
no puzzle there. For the puzzle to begin to arise, we need to bring the modali-
ties into line.
I mention this because it affects the moral case quite closely. Suppose we
allow ourselves a notion of 'analytically necessary' applying to propositions
that, in the traditional phrase, can be seen to be true by conceptual means
alone. To deny one of these would be to exhibit a conceptual confusion: a
failure to grasp the nature of the relevant vocabulary or to follow out immedi-
ate implications of that grasp. In a slightly more modern idiom, to deny one of
these would be 'constitutive' of lack of competence with the vocabulary. We
may contrast this with metaphysical necessity: a proposition will be this if it is
true in all the possible worlds that, as a matter of metaphysics, could exist. Of
course, we may be sceptical about this division, but I want to respect it at least
for the sake of argument. For the (S)/(P) combination in moral philosophy
provides a nice example of & prima facie case of the difference, and one that
profoundly affects my original argument. This arises because someone who
holds that a particular natural state of affairs, G*, underlies a moral judge-
ment is very likely to hold that this is true as a matter of metaphysical neces-
sity. For example, if I hold that the fact that someone enjoys the misery of
others underlies the judgement that he is evil, I should also hold that, in any
possible world, the fact that someone is like this is enough to make him evil.
Using 'MN' for metaphysical necessity, I would have both

ar

and I would evade the original argument by disallowing the metaphysical

possibility of a world in which people like that were not evil. This, it might be
said, is part of what is involved in having a genuine standard, a belief that
some natural state of affairs is sufficient to warrant the moral judgement. For,
otherwise, if in some metaphysically possible worlds people like that were evil
and in others they were not, surely this would be a sign that we hadn't yet
properly located the natural basis for the judgement. For instance, if I did
allow a possible world in which some people like that were not evil, it might
be because (for instance) they believe that misery is so good for the soul that it
 Supervenience Revisited 137

is a cause of congratulation and rejoicing to find someone miserable. But then

this fact becomes what I earlier called a releasing fact, and the real underlying
state of affairs is now not just that someone enjoys the misery of others, but
that he does so not believing that misery is good for the soul.
Because of this, the original puzzle does not arise at the level of metaphysi-
cal necessity. But now suppose we try analytic necessity. It seems to be a
conceptual matter that moral claims supervene upon natural ones. Anyone
failing to realize this, or to obey the constraint, would indeed lack something
constitutive of competence in the moral practice. And there is good reason for
this: it would betray the whole purpose for which we moralize, which is to
choose, commend, rank, approve, or forbid things on the basis of their natu-
ral properties. So we might have

But we would be most unwise to have

For it is not plausible to maintain that the adoption of some particular stan-
dard is 'constitutive of competence' as a moralist. People can moralize in
obedience to the conceptual constraints that govern all moralizing, although
they adopt different standards and come to different verdicts in the light of a
complete set of natural facts. Of course, this can be denied, but for the sake of
this essay I shall rely on the common view that it is mistaken. So since we deny
(N2) we have

We then arrive at an (Sa)/(Pa) combination, and my mystery emerges: why the

ban on mixed worlds at this level? These would be worlds possible as far as
analytic constraints go, or 'analytically possible' worlds. They conform to
conceptual constraints, although there might be metaphysical or physical bars
against their actual existence.
Of course, in a sense I have already proposed an answer to this question.
By saying enough of what moralizing is to make (Sa) plausible, and enough to
make (Pa) plausible, I hope to enable us to learn to relax with their combina-
tion. It is just that this relaxation befits the anti-realist better. Because the
explanation of the combination depended crucially upon the role of moraliz-
ing being to guide desires and choices among the natural features of the
world. If, as a realist ought to say, its role is to describe further moral aspects
of the world, there is no explanation at all of why obeying the constraint (Sa) is
constitutive of competence as a moralist.
Can this argument be avoided by maintaining (?)? No, because there is no
prospect of accepting (?) in a relevantly strong sense. For (?) to help, we
would need it to be read so that, necessarily (in some sense), if something is F
138 Ethics

and G*, and the G*-ness underlies its being F, then it is analytically necessary
that anything G* is F. And this we will not have in the moral case, for we want
to say that there are things with natural properties underlying moral ones, but
we also deny analyticities of the form (N a ). (?) would not help if the necessity
of the consequent were interpreted in any weaker sense. For example, we
might want to accept (?) in the form

Then there will be metaphysical necessities of the form of the consequent, that
is, of the form (N m ), but they will not help to resolve the original mystery,
since that is now proceeding at the level of analytical necessity. It is the
possibility, so far as conceptual constraints go, of mixed worlds that is to be
avoided.

III

The argument above works because we are careful to distinguish the status of
the supervenience claim, and in this case its extremely strong status, from that
of the related possibility claim. I have done that by indexing the modal opera-
tors involved: We have four different forms of modal claim: (S), (N), (P) and
(?), and each of them can involve analytic or conceptual necessity, ( a ), meta-
physical necessity ( m ), and we come now to physical necessity, (p). For now I
want to turn to consider nonmoral cases of the same kind of shape. These
examples are all going to start life as examples of the joint (S)/(P) combina-
tion. They may not finish life like that: it may become obvious, if it is not so
already, in the light either of the plausibility of (?) or of the difficulty over
banning mixed worlds, that either supervenience is to be abandoned, or (N)
accepted. But here are some test cases:
First example. Suppose that in w, a physical setup G* underlies some
particular mental state F. Suppose G* is possession of some pattern of neu-
rones or molecules in the head, and Fis having a headache. Nowhere in wl is
there anything unlike x, in being G* but not F. Next door, in w2, however,
there are things that are G* but not F. Now we are told that wl is acceptable,
and that w2 is acceptable. But nowhere is there a world w3 that is like wl but
that changes to become like w2, or that contains some particular individuals
who are like those of w2.
Second example. Suppose that in w1 a particular molecular constitution G*
underlies membership of a natural kind, F. G* consists of a complete physical
or chemical breakdown of the constitution of a substance (e.g., being com-
posed of molecules of H2O), and Fis being water. Nowhere in wl could there
be a substance with that chemistry, that is not water. In w2, however, this
combination is found. Once again, although each of these possible worlds
exists, there is no G*/FvO, or mixed world, in which some substances with
this chemical constitution are water and others are not.
 Supervenience Revisited 139

Third example. Suppose that in wl a particular set of primary qualities,

particularly concerning refractive properties of surfaces, G*, underlies posses-
sion of a colour F. Nowhere in wl could there be things with that kind of
surface, without that particular colour. However, there are possible worlds
where this combination is found: G*/O worlds. Again, there are no mixed
worlds, where some things with the primary, surface properties are F, and
others are not.
In each of these cases we have the (S)/(P) combination. And I hope it is
obvious that each case is at least prima facie puzzling—enough so to raise
questions about whether the combination is desirable, or whether we should
make severe distinctions within the kinds of necessity and possibility involved
to end up avoiding the combination altogether. How would this be done?

First Example
How should we interpret the supervenience of the mental on the physical?
Perhaps centrally as a metaphysical doctrine, so we shall accept (Sm). Should
we accept (Sa)?We should if we can find arguments as strong as those in the
moral case for claiming that it is constitutive of competence in the mental
language that we recognize the supervenience of the mental on the physical.
But I doubt if we can do this. For whether or not we are philosophically
wedded to the doctrine, we can surely recognize ordinary competence in users
who would not agree. One day Henry has a headache, and the next day he
does not. Something mental is different. But suppose he simply denies that
anything physical is different (giving voice to Cartesianism). Is this parallel to
the error of someone who makes the same move in a moral case? I do not
think so: Henry is not so very unusual, and if his error is shown to be one
because of the 'very meaning' of mental ascriptions, then whole cultures have
been prone to denial of an analytic truth. In other words, it seems to me to be
overambitious to claim that it follows, or follows analytically, from change in a
mental state that there is change in an underlying physical state. It makes
views conceptually incoherent when enough people have found them per-
fectly coherent (consider, for example, changes in God's mind).
Let us stick then with (Sm). It would seem to me plausible, if we accept
this, to accept the correlative necessities (Nm) and (?m). We would then be
forced to deny (P m ), and we just do not get involved with the problem of
banning mixed worlds. (Nm) does the work for us, by disallowing the meta-
physical possibility of G* without F. However, there is the famous, or notori-
ous, position of Davidson to consider, which accepts some form of super-
venience of the mental on the physical, but also denies the existence of lawlike
propositions connecting the two vocabularies.4 Davidson is not very explicit
about the strength of necessity and possibility involved in his claims, but it can
scarcely be intended to be weaker than joint acceptance of (Sm) and of (P ).

4. D. Davidson, 'Mental Events', in Essays on Actions and Events (Oxford: Clarendon

Press, 1980).
140 Ethics

And even if supervenience is taken as a matter not of metaphysical, but just of

sheer physical necessity (in our physical world, there is no mental change
without physical change, even if there could be), it does not matter. For from
(Sm) we can deduce (Sp), so we have the (S)/(P) combination, at the level of
physical necessity. So, according to me, the position ought to be odd, and
indeed it is. Why is it physically impossible for there to be a world that
contains some wl characters with headaches, and some w2 characters in the
same physical state but without headaches? Once we have allowed the physi-
cal possibility of the w2_type, how can we disallow the physical possibility of
them mingling with w1 types?
It does not appear to me that light is cast on this by Davidson's reason for
allowing (Pp). This reason is that in some sense the mental and the physical
belong to different realms of theory: ascriptions of mental properties answer
to constraints different from ascriptions of physical ones, and hence we can
never be in a position to insist upon a lawlike correlation of any given physical
state with any given mental state (this is not just the variable realization point:
here we are told not to insist upon a physically necessary physical-to-mental
correlation; prima facie we might be allowed to do that in various cases, even
if we could never insist upon lawlike connections the other way round). I do
not accept that this is a good argument, for there can certainly be interesting
laws that connect properties whose ascriptions answer to different constraints:
temperature and pressure, or colour and primary properties, for instance.
However, I do not want to insist upon that. For there remains the oddity that
if Davidson's reasoning is good, it should apply equally to the supervenience
claim. How can we be in a position to insist upon anything as strong as (S p ), let
alone as strong as (Sm)? The freedom that gives us (Pp) is just as effective here.
I may coherently and effectively 'rationalize' one person as being in one
mental state and another as being in another, obeying various canonical princi-
ples of interpretation, regardless of whether they are in an identical G* state:
I might just disclaim interest in that. Of course, if (Np) were true, it would be
different, but it is precisely this which the anomalous character of the position
denies.
So if the mental reality is in no lawlike connection with the physical, as (Pp)
claims, I can see no basis for asserting that nevertheless it supervenes upon it.
But, again, the word 'reality' matters here. One way of thinking that the
mental must supervene upon the physical, in at least the sense of (S m ), is by
convincing ourselves that the physical reality is at bottom the only reality:
molecules and neurones are all that there are. And then there might be
something about the way mental vocabulary relates to this—relates to the
only reality there is—that justifies both (Sp) and (Pp). Perhaps there is some
argument that obeying the supervenience constraint is required for conceptual
coherence, or at any rate for metaphysical coherence; and perhaps Davidson's
argument for (P ) can be put in a better light than I have allowed. I do not
want to deny this possibility. But I do want to point out that once more it is
bought at the cost of a highly anti-realist, even idealist, view of the mental.
The 'truth' about the mental world is not a matter of how some set of facts
 Supervenience Revisited 141

actually falls out. It is a matter of how we have to relate this particular

vocabulary to the one underlying reality. If we thought like this, then we
would begin to assimilate the mental/physical case to the moral/natural case.
At any rate, it provides no swift model for arguing that anti-realism is the
wrong diagnosis of the (S)/(P) combination in that case.
Why do I say that this is an idealist or anti-realist direction? Because the
constraint on our theorizing is not explained by any constraint upon the way
the facts can fall out. It is constrained by the way we 'must' use the vocabu-
lary, but that 'must' is not itself derived from a theory according to which
mental facts and events cannot happen in some given pattern; it is derived
from constraints on the way in which we must react to a nonmental, physical
world. I regard this as a characteristically idealist pattern: the way the facts
have to be is explained ultimately by the way we have to describe them as
being. Thus I would say that the explanation of moral supervenience is a
paradigmatically anti-realist explanation. By way of contrast, and anticipating
the third example, we can notice how there cannot be a strong, analytic
version of the doctrine that colours supervene upon primary properties, pre-
cisely because it is so obvious that the only conceptual constraint upon using
the colour vocabulary is that you react to perceived colour in the right way.
Somebody who thinks that a thing has changed colour, but who is perfectly
indifferent to any question of whether it has changed in respect of any primary
property (or even who believes that it positively has not done so), is quite
within his rights. His eyesight may be defective, but his grasp of the vocabu-
lary need not be so.

Second Example
In order to avoid unnecessary complexity, I should enter a caveat here. I am
going to take being composed of H2O as a suitable example of G*; an example
that is of the kind of complete physical or chemical basis that results in stuff
being of a certain kind, such as water. I am going to take it that this is known
to be the case. So I shall not be interested in the kind of gap, which can in
principle open up, in which people might allow that something is H2O and is
water, allow that wateriness supervenes upon the chemical or the physical, but
deny that some other specimen of H2O is water. This is a possible position,
because it is possible to disbelieve that the facts registered by something's
being H2O exhaust the physical or chemical facts that may be relevant to its
kind. I am going to cut this corner by writing as though it is beyond question
that molecular constitution is the right candidate for a complete underlying
property—a G* property. I don't think that this affects the argument, al-
though it is a complex area and one in which it is easy to mistake one's
bearings.
Once more, it is natural to take the various claims involving the relation-
ship of H2O and water in a metaphysical sense. It is also natural, to me at any
rate, to assert (Sm) only if we also assert (Nra), and (? m ). Being water super-
venes upon being H2O only because anything made of H2O must be water.
142 Ethics

And if we had an argument that it does not have to be water, perhaps because
we imagine a world in which countervailing circumstance makes substances
composed of that molecule quite unlike water at the macro-level (and more
unlike it than ice or steam), then we would just change the basis for the
supervenience. We would have argued for a releasing property, R, and the
true basis upon which being water supervenes would be G* (being H2O) and
being ~ R.
Might someone believe that the (S)/(P) combination arises at some level
here? The argument would have to be that in some strong sense it must be
true that being water supervenes upon physical or chemical constitution; but it
need not be true, in this equally strong sense, that H2O is the particular
underlying state. Now, I do not think there is any very strong sense in which
being water must be a property underlain by a physical or chemical basis. Of
course, we are familiar with the idea that any such property must be a matter
of chemistry. But there is no good reason for saying that people who fail to
realize this are incompetent with the kind term 'water'. They just know less
about the true scientific picture of what it is that explains the phenomenologi-
cally important, macro-properties of kinds. They are not, in my view, in at all
the same boat as people who fail to respect the supervenience of the moral on
the natural. This is because the latter fault breaks up the whole point of
moralizing, whereas ignorance of the way in which wateriness is supervenient
on the chemical or physical does not at all destroy the point of classifying
some stuff as water and other stuff as not: uneducated people still need to
drink and wash. However, it is now commonly held that there is no absolute
distinction here: Quine has taught us how fragile any division would be be-
tween conceptual and 'merely' scientific ignorance. So someone might hold
that there is an important kind of incompetence, half conceptual but perhaps
half scientific, that someone would exhibit if he failed to realize the super-
venience of being water upon chemistry or physics, and that this is a worse
kind of incompetence than any that would be shown by mere failure to realize
that it is H2O which is the relevant molecule. So we might try a notion of
'competently possible worlds', (c), meaning those that are as a competent
person might describe a world as being: then we would have an (SC)/(PC)
combination. Should this tempt us to an anti-realist theory of 'being water'?
Saying that there are no mixed possible worlds in this sense means only
that any competent person is going to deny that there are worlds in which
some things of a given chemical or physical structure are water and others of
the same structure are not; but that competent people might allow worlds in
which things are H2O but are not water. The first is a kind of framework
knowledge, which we might expect everyone to possess; but competence to
this degree need not require the specialist piece of scientific knowledge, which
we might not expect of everyone, and which might even turn out to be false
without affecting the framework. We might even suppose that supervenience
claims have, characteristically, this framework appearance, and suggest that
this is why they do not trail in their wake particular commitments of the (N)
form. And now the counterattack against my argument in the case of morals
 Supervenience Revisited 143
5
and mind gathers momentum. For if an (SC)/(PC) combination works in a
harmless case like this, then the shape of that combination cannot in general
suggest anti-realism, and something must be wrong in the arguments so far
given.
One reaction would be to allow the parallelism, and to grasp the nettle.
When I said that we could relax with the (Sa)/(Pa) combination in morals, I
tried to explain this by saying that the role of a moral judgement is not to
describe further moral aspects of reality; it is because the vocabulary must fit
the natural world in certain ways that the combination is explicable. I might
try the same move here: it is because 'wateriness' is not a further aspect of
reality (beyond its containing various stuffs defined in chemical ways) that the
combination is permissible at the level of 'competently possible' worlds. But I
think this will strike most uncommitted readers as weak: anti-realism has to
fight for a place these days even in the philosophy of morals, and is hardly
likely to seem the best account of the judgement that I have water in my glass.
I think a better reaction is to remember well all that is meant by the notion of
a 'competently possible world'. Remembering this enables us to say that an
(SC)/(PC) combination is harmless, and implies no problem of explanation that
is best met by anti-realism. This is because the 'ban on mixed possible worlds'
that it gives rise to is explicable purely in terms of beliefs of ours—in particu-
lar, a belief that we suppose competent people to share. We believe, that is,
that no two things could be identical physically without also forming the same
stuff or kind and we believe that all competent people will agree. While we
suppose this, but also suppose that competent people may not agree that if a
thing is H2O then it is water (because this requires a higher level of special-
ized, as opposed to framework, knowledge), then we have 'competently possi-
ble' worlds of the two kinds, and the ban on mixed worlds. But this has now
been explained purely by the structure of beliefs that can coexist with compe-
tence. There is indeed no further inference to a metaphysical conclusion
about the status of wateriness, because the explanation which, in the other
cases, that inference helps to provide is here provided without it. To put it
another way, we could say that in the moral case as well, when we deal with
analytically possible worlds, we are dealing with beliefs we have about compe-
tence: in this case the belief that the competent person will not flout super-
venience. But this belief is only explained by the further, anti-realist nature of
moralizing. If moralizing were depicting further, moral aspects of reality,
there would be no explanation of the conceptual constraint, and hence of our
belief about the shape of a competent morality.
It cannot be overemphasized that my original problem is one of explana-
tion. So it does not matter if sometimes an (S)/(P) combination is explained
in some ways, and sometimes in others. I do not suppose that there is one
uniform pattern of explanation, suitable for all examples and for all strengths
of modality (particularly if we flirt with hybrids like the present one). The

5. I owe this objection to Michael Smith. I am also greatly indebted to conversations with
David Bostock and Elizabeth Fricker.
144 Ethics

explanation demanded in the moral case is, according to me, best met by
recognizing that moralizing is an activity that cannot proceed successfully
without recognition of the supervenience constraint, but this in turn is best
explained by projectivism. In the present case, the best explanation of why
competent people recognize the supervenience of kinds upon physical or
chemical structure is that we live in a culture in which science has found this
out. I don't for a moment believe that this suggests any metaphysical conclu-
sions. If this is right, it carries a small bonus. It means that the argument in
the moral case does not depend upon drawing a hard and controversial
distinction between 'conceptual' and other kinds of incompetence. It merely
requires us to realize that there can be good explanations of our beliefs about
the things that reveal incompetence. Anti-realism is one of them, in the case
of morals, and awareness of the difference between framework scientific
beliefs and specific realizations of them is another and works in the case of
natural kinds.

Third Example
The previous case posed the only real challenge that I know to the original
mystery. By contrast, the case of colours reinforces the peculiarity in the case
of morals. For it would be highly implausible to aim for colour/primary prop-
erty supervenience as an analytic truth or one constitutive of competence with
a colour vocabulary. Intuitively, we feel that it is very nice and satisfying that
colours do indeed supervene upon primary properties, and that there would
be scientific havoc if they did not. But anybody who believes that they do not
(mightn't God live in a world where displays reveal different colours to him,
although there are no physical properties of surfaces of the things displayed?)
can nonetheless recognize colours and achieve all the point and subtlety of
colour classification.
Recent empirical work casts doubt even on the fact that 'everybody
knows' that colours of surfaces are caused by the wavelength of reflected
light. Other relational properties may matter. So it is wise to be cautious
before putting any advanced modal status on supervenience or necessitation
claims in this area. Certainly we expect there to be some complete primary
property story, G*, upon which colour supervenes as a matter of physical
necessity. But then we would also immediately accept the corresponding the-
sis (N ), and there is no problem about mixed worlds. Similary, if we bravely
elevated the supervenience (S) into a metaphysical thesis, there would be no
good reason why (N) should not follow suit. (N) will not rise into the realms of
analytic necessity, but then neither will (S). So at no level is there a mystery
parallel to the one that arose with morals, and with Davidson's position on the
mental and the physical. Of course, an (S)/(P) combination could be manufac-
tured at the level of 'competently possible worlds', as in the last example, but
once more it would avail nothing, because it would be explained simply by the
shape of the beliefs that we have deemed necessary for competence.
 Supervenience Revisited 145

IV

I have now said enough by way of exploring the original argument and its near
neighbours. It would be nice to conclude with an estimate of the importance
of supervenience claims in metaphysics. Here I confess I am pessimistic. It
seems to me that (?) is a plausible doctrine, and in every case in which we are
dealing with metaphysical or physical necessity, it seems to me that we could
cut through talk of supervenience and speak directly of propositions of the
form (N). This makes it clear, for example, that we may be dealing with
'nomological danglers' or necessities that connect together properties of very
different kinds, and it may lessen our metaphysical pride to remember that it
is one thing to assert such necessities but quite another thing to have a theory
about why we can do so. Like many philosophers, I believe many super-
venience claims in varying strengths; perhaps unlike them, I see them as part
of the problem—in the philosophy of mind, or of secondary properties, or of
morals or kinds—and not as part of the solution.

Addendum
This essay and its predecessor have probably attracted more discussion than
any others in this volume. But I am heartened by what I detect as a general
impatience with easy acceptance of supervenience claims, either in the moral
sphere, or in the philosophy of mind, or in the relations between the special
sciences.6 So at least my concluding paragraph has found an echo. Indeed,
some philosophers have voiced mistrust of the whole notion, and it is easy to
understand why. Supervenience is supposed to give us a great deal: the covari-
ance of one set of properties (the F, or further properties) with another set
(the G, or grounding properties); the dependence of the F properties on the
G; these consistently with the absence of reduction of the F to the G proper-
ties; and, finally, all in a metaphysically innocent way. It is this last that makes
the package almost too good to be true. Kim reminds us that when the same
structure was discussed in the philosophy of biology, earlier in the century,
Samuel Alexander, a leading emergentist, recommended that we accept emer-
gence (in effect, the same thing as supervenience) with natural piety, and my
1973 essay pointed out the dark side of that: unless we can explain the super-
venience, there is nothing metaphysically innocent about it at all.7 Indeed,
without further explanation we might be left with a nomological dangler, or
6. For instance, Stephen Schiffer, in Remnants of Meaning (Cambridge: MIT Press, 1987)
claims that appeals to a 'special primitive' relationship of supervenience is 'obscurantist in the
extreme'. Jaegwon Kim, in 'Supervenience as a Philosophical Concept' (Metaphilosophy, 1990)
gives an extremely useful view of the various concepts that have been distinguished, and high-
lights the explanatory task that supervenience claims trail with them.
7. See Kim, p. 15.
146 Ethics

worse if the supervenience is a priori, since then we are left with a nasty
'opaque isolated logical fact'. My strategy was to urge that the demand for an
explanation is met much better by a quasi-realist story than by any other.
A number of writers challenged this perception. Some challenged whether
supervenience of the moral on the natural is, as I claimed, analytic. One kind
of counterexample might be divine command moralities, in which value is
supposed to supervene upon the existence of a divine fiat, which presumably
might deem some things good and then bad without there being natural
change. 8 But apart from its doubtful coherence, I think the position would
best be characterized as having a rather over-generous conception of the
natural, by presenting divinities as well as humans to think of. Extreme
axiarchism is a more threatening position, since it seems to reverse the order
of dependency, making the natural world dependent on the ethical order.
Again, the coherence of the thought is not all that evident, but a better
response is to hand (here I am indebted to an unpublished paper, 'Moral
Covariance', by Glenn Branch). We can jettison the 'dependency' element in
supervenience without losing my argument. That is, so long as we have it
analytic that the ethical covaries with the natural, or in Davidson's terms that
F predicates cannot distinguish between any entities that G predicates cannot
distinguish between, then we have the same demand for explanation, whether
or not we think that Gs sometimes depend upon Fs. The whole business of
dependence is very fraught in any event, as essays 11 and 13 testify, so it is
probably wise to keep away from it as long as possible. This is one respect in
which essay 6 may actually have posed the problem better than essay 7, whose
asymmetric notion of 'underlying' introduces an element of dependency
which, while present in most concepts of supervenience, may be better put on
one side for the sake of this problem (although it will not leave the stage
entirely, since it is worth realizing that it may pose a further explanatory
demand of its own). That is, as well as explaining the status we give to the
covariance of F predicates and G predicates, we must also explain why we
think that the F properties are there 'in virtue of an object's G properties,
and this may not be all that easy. A quasi-realist will see both covariance and
the asymmetry of dependency as a reflection of the fact that valuing is to be
done in the light of an object's natural properties, and without that constraint
nothing recognizably ethical could be approached at all.
Other writers have missed the point of the problem in other ways. Susan
Hurley, for example, cites the fact that 'theories must supervene on descrip-
tions of what they are theories about' and claims that this shows that there is
no need for noncognitive underpining for supervenience, which is as well
explained by her own theory-oriented view of ethics.9 But the claim backfires,
for the truth of theories need not at all supervene (let alone analytically) upon
what they are theories about. The underdetermination of theory by data can
be put exactly as the claim that it is logically possible that an identical body of
8. James Klagge, 'An Alleged Difficulty Concerning Moral Properties', in Mind, 1984.
9. S. Hurley, 'Objectivity and Disagreement', in Morality and Objectivity, ed. Ted Honde-
rich (London: Routledge & Kcgan Paul, 1985), pp. 56-57.
 Supervenience Revisited 147

data might in one world be truly explained by one theory, and in another
world by a different and incompatible theory.
Hurley goes on to 'make the point more formally' by making a distinction
that seems to be equivalent to that between weak (intraworld) and strong
(interworld) supervenience: 'in each logically possible world there may be
some theory that does the requisite job, while there is no one theory such that
in every logically possible world that theory does the requisite job'. 'Doing the
job' here means being true by a coherentist conception of truth, so we have
the idea that a theory may be true in one world but not in another, although in
every world there will be some true ethical theory. This is no more than setting
up the problem, giving us the distinction between

(1) Necessarily, if anything x has property F in A, then there is a G such

that x has G, and there is a theory true in A entailing that everything
G has F.

and

(2) Necessarily, if anything x has property F in A, then there is a G such

that x has G, and there is a theory true at all worlds, entailing that
everything G has F.

The ban on mixed worlds is given by the first, and the contrast with necessita-
tion claims by its contrast with the second. The cost is that once we allow
different theories of GIF relations true at different worlds, my question simply
re-emerges: why shouldn't one of these theories be true of some objects in
some world, for example at some place or time, and another be true of other
objects at other places and times? Hurley tells us that there are to be no mixed
worlds, but completely fails to engage the project of explaining why this has to
be so, which is the problem I posed.
In general, I remain fairly happy with the argument. It does, I think,
depend very heavily on the outside analyticity claim. If it is less than 'constitu-
tive of competence' with the concepts that supervenience is recognized, the
explanatory possibilities multiply. For example, 'Cornell' realism (see the
addendum to essay 9) seems at first sight to have no trouble explaining su-
pervenience: if ethical properties are natural properties, then of course they
covary. But Cornell realism, I argue, is less successful in explaining what it is
to have an ethical perspective on a natural property, and therefore not at all
well equipped to say why it is crucial to that perspective that supervenience be
respected. After all, water is H2O, but it is not part of competence with
attributions of wateriness that we respect a chemical supervenience claim. We
make a scientific error, perhaps, but no conceptual error if we believe that
wateriness depends on some other, perhaps unknown, properties.
A possible reaction to my problem is to press for 'strong', or interworld,
supervenience, which is here called necessitation, whereas I pose the puzzle in
terms of the weak, or intraworld, notion, which is why the ban on mixed worlds
148 Ethics

comes out oddly. Necessitation gives us the alternative posed in the first part of
this essay, that there is some ground G* such that, necessarily, if anything is
G*, it is F. It may be, as I hint in the essay, that this is a better notion to work
with, and if so my problem disappears. But something in the texture of ethics is
left out. For it is not evidently 'constitutive of competence' to recognize any
instance of the conditional 'if anything is G*, it is F' as necessary. It is not even
clear that the competent must recognize that there are instances of such necessi-
ties. Yet supervenience (like universalizability) latched onto something quite
obvious about ethics. This obvious thing seemed better captured by weak
supervenience than by necessitation.
What this leaves us, however, is a rival explanation to mine. The analytic
claim can be put weakly: it is entirely intraworld, as in (Sa) (p. 137), and this
befits a condition on competent moralizing—which, after all, concerns what is
true in this world rather than what might be true in exotic possible worlds. But
then the proposition that there are no mixed worlds need not be analytic: it is
no doubt true, and is recognized as true by the most capable people, such as
philosophers. But the question of why philosophers think it true is answered
another way: it is because we philosophers believe that there are instances of
the necessitation, and that is where the authority behind the ban lies. Because
if there cannot be things that are G* and not F, then there cannot be worlds in
which some G* things are and others are not F.
If that is right, my original puzzle gets a solution without invoking any
projective theory. But the explanatory demand must turn to the necessita-
tions, and the realist must say why, for her, they do not amount to danglers
best accepted in a spirit of natural piety. I know why I accept them, if I can
find my ethical imagination simply baffled at trying to contemplate how not to
approve or disapprove of things with particular natural properties. But the
realist still seems left drawing a blank cheque on the synthetic a priori, which
strikes me as worse, especially when, far from having examples of these
necessities, we are so hard pressed to find examples where the unmodal
conditionals are even true.
 8
Errors and the Phenomenology of Value
Oh Is there not one maiden breast
That does not feel the moral beauty
Of making worldly interest
Subordinate to sense of duty?
W. S. Gilbert
The Pirates of Penzance

John Mackie described himself as a moral sceptic, and he described his theory
of ethics as an error theory. The ordinary user of moral language wants to
claim something that, according to Mackie, cannot be claimed without error:
he wants to claim 'something that involves a call for action or for the refrain-
ing from action, and one that is absolute, not contingent upon any desire or
preference or policy or choice, his own or anyone else's' (p. 33).l Again,
someone in moral perplexity may want to know whether a course of action is
wrong 'in itself (p. 34), and 'something like this is the everyday objectivist
concept', which is erroneous. For, according to Mackie, ordinary judgements
and perplexities include an assumption that there are objective values, in a
sense in which he denies that there are. This assumption is ingrained enough
to count as part of the meaning of ordinary moral terms, but it is false.
Mackie did not draw quite the consequences one might have expected from
this position. If a vocabulary embodies an error, then it would be better if it
were replaced by one that avoids the error. Slightly more accurately, if a
vocabulary embodies an error in some use, it would be better if either it, or a
replacement vocabulary, were used differently. We could better describe this
by saying that our old, infected moral concepts or ways of thought should be
replaced by ones that serve our legitimate needs but avoid the mistake. Yet
Mackie does not say what such a way of thought would look like, and how it
would differ in order to show its innocence of the old error. On the contrary, in
the second part of the book he is quite happy to go on to express a large
number of straightforward moral views about the good life, about whether it is
1. Unless otherwise stated, page references are to Ethics: Inventing Right and Wrong (Lon-
don: Penguin Books, 1977).

149
150 Ethics

permissible to commit suicide or abortion, and so on. All these are expressed
in the old, supposedly infected, vocabulary. Mackie does, of course, notice the
problem. He explicitly asks (p. 49) whether his error theory rules out all first-
order ethics, and when he returns to the question (p. 105) there is a real threat
that ideally there would be no such activity as first-order moralizing. The
threat is averted, supposedly, only by introducing the general Humean theme
about the social function of morality: 'Morality is not to be discovered but to
be made: we have to decide what moral views to adopt, what moral stands to
take' (p. 106). Yet from the standpoint of an error theory, it is quite extraordi-
nary that we should have to do any such thing. Why should we have to choose
to fall into error? Surely it would be better if we avoided moral (erroneous)
views altogether and contented ourselves with some lesser, purged commit-
ments that can be held without making metaphysical mistakes. Let us call
these 'shmoral' views, and a vocabulary that expresses them a 'shmoral vocabu-
lary.' Then the puzzle is why, in the light of the error theory, Mackie did not at
least indicate how a shmoral vocabulary would look, and why he did not
himself go on only to shmoralize, not to moralize. And in my view this is
enough of a puzzle to cast doubt back on to the original diagnosis of error. In
other words, it would obviously have been a silly thing to do, to try to substi-
tute some allegedly hygienic set of concepts for moral ones; but that in itself
suggests that no error can be incorporated in mere use of those concepts.
In reply to this it may be said that appearances notwithstanding, Mackie
did actually go on only to shmoralize. He rids himself of the error but uses the
Humean reconstruction of practical needs and practical reasoning to advocate
various shmoral views. These are only accidentally expressed in a vocabulary
that looks like that of ordinary moralists: the identical shape of the words does
not signify identical concepts, although there is sufficient overlap in function
between moralizing and shmoralizing to justify retention of the same words.
This is certainly possible. But it leaves an acute problem of identifying just
where shmoralizing differs from moralizing: what shows us whether Mackie is
moralizing or shmoralizing? Does it determine the issue that he will say things
like 'there is no objective prescriptivity built into the fabric of the world'?
Troubles multiply. First, it is clear that not all moralists will deny this (many
moralists will not even understand it). Second, it seems gratuitous to infer that
there are two different activities from the fact that there are two or more
different theories about the nature of the activity. It would be much more
natural to say that Hume and Mackie moralize, just as ordinary people do, but
with a developed and different theory about what it is that they are doing. The
error theory then shrinks to the claim that most ordinary moralists have a bad
theory, or at least no very good theory, about what it is to moralize, and in
particular that they falsely imagine a kind of objectivity for values, obligations,
and so on. This may be true, but it does not follow that the error infects the
practice of moralizing, nor the concepts used in ways defined by that practice.
Here, however, a fairly blanket holism can be introduced to rescue
Mackie, or at least to urge that it is profitless to oppose him. Our theories
infect our meanings; so a different theory about the nature of the activity of
 Errors and the Phenomenology of Value 151

moralizing will yield a different meaning for the terms with which we do it;
hence Mackie is right that the ordinary meanings do embody error. It be-
comes profitless to split things in two, so that on the one hand there is the
error-free practice, and on the other hand a multiplicity of possibly erroneous
theories about its nature. Indeed, the split appeals no more than the despised
analytic-synthetic distinction, and if the opponents of an error theory need
that, they will gain few supporters.
It is important, and not just to this philosophical issue, to see that this
defence fails. To answer it, distinguish between the activity or practice of moral-
izing and the 'full meaning' of moral terms, where this is determined as the
holist wishes, by both the practice and whatever theory the subjects hold about
the nature of their practice. Then the holist may have the thesis about 'full
meaning', with the consequence that Hume and Mackie may give a different
full meaning to their terms, simply through having a different theory of their
point and purpose. But it will not follow that their practice will differ from that
of other people. Hence, it will not follow that other people's practice embodies
error. For it is in principle possible that we should observe the practice of some
subjects as closely as we wish, and know as much as there is to know about their
ways of thinking, commending, approving, deliberating, worrying, and so on,
yet be unable to tell from all that which theory they hold. The practice could be
clipped on to either metaphysic. The holist will have it that this alters meanings
throughout. But we can give him that, yet still maintain that no difference is
discernible in the practice, and therefore that no error is embodied in the
practice of those who hold the wrong theory. To use a close analogy, there are
different theories about the nature of arithmetical concepts. Hence a holist may
claim that a subject will give a different total meaning to numerals depending
on which theory he accepts, and this difference will apply just as much when the
subject is counting as when he is doing metamathematics. All that may be true,
yet it would not follow that any practice of counting embodies error. That
would be so only if one could tell just by observing it which of the competing
metamathematical theories the subject accepts. In the arithmetical case this
would not be true. Similarly, I maintain, in the moral case one ought not to be
able to tell from the way in which someone conducts the activity of moralizing
whether he has committed the 'objectivist' mistake or not; hence any such
mistake is better thought of as accidental to the practice.
Obviously there is an answer to this. It is that the objectivist error does so
permeate the practice that you can tell, from the way people moralize, that
they are in its grip. It is as if a strict finitist theory of, say, arithmetic led
someone to deny that you could count certain sets that others can happily
enumerate. But which features of the practice show this? They are to be
features that lie beyond the scope of what I have called 'quasi-realism': the
enterprise of showing how much of the apparently 'realist' appearance of
ordinary moral thought is explicable and justifiable on an anti-realist picture. 2
1. 'Rule Following and Moral Realism', in Wittgenstein: To Follow a Rule, ed. S. Holtzman
and S. Leich (London: Routledge & Kegan Paul, 1981), and chapter 6 of Spreading the Word
(Oxford: Clarendon Press, 1984).
152 Ethics

According to me, quasi-realism is almost entirely successful, and I do not

think John Mackie provided reasons for thinking otherwise. In other words,
proper shmoralizing is proper moralizing.

II

So far, I have tried to show that there is something fishy about holding an
error theory yet continuing to moralize, and I have argued that the 'holistic' or
Quinean defence of such a position would fail. The argument can now move
in different directions. Let us call the Humean picture of the nature of moral-
ity, and of the metaphysics of the issue, projectivism. On this view we have
sentiments and other reactions caused by natural features of things, and we
'gild or stain' the world by describing it as if it contained features answering to
these sentiments, in the way that the niceness of an ice cream answers to the
pleasure it gives us. Then we could say that Mackie is right about the meta-
physical issue, and ought to have been more thoroughgoing in replacing moral
terms and concepts by different ones—in other words, that the projectivist in
ethics should conduct his practical reasoning in a different way: his shmoraliz-
ing would not be moralizing. Let us call this a revisionist projectivism. By
contrast, there is the quasi-realist identification of shmoralizing with moraliz-
ing. In effect, the skirmishes in part I of this essay urge that quasi-realism be
taken seriously, because even projectivists are going to find themselves indulg-
ing in a practice that is apparently identical with moralizing. Of course, in
opposition to each of these views, there is the realist charge that projectivism
is false in any case; finally there is the 'quietist' view, urged by Professor Hare,
for instance, that no real issue can be built around the objectivity or otherwise
of moral values.
If we are to say that the practices characteristic of moralizing are or are
not available to a projectivist, we should be careful to identify the practices at
issue. Elsewhere I try to show how the realist-seeming grammar of moral
discourse can be explained on that metaphysic.3 This involves, for instance,
addressing the Geach-Frege problem of accounting for the unasserted occur-
rence of sentences using moral terms, explaining the propositional form that
we give to moral utterances, explaining why we may legitimately worry
whether one of our moral views is correct and hence explaining the role of a
concept of truth in ethics, and so on. If this work is successful, there is no way
of arguing that the grammar of moral discourse either refutes projectivism or
forces it to take a revisionist course. This means, of course, that Mackie
cannot properly use these aspects of our practice in support of the error
theory. And sometimes he does just this. For instance, he cites Russell's
feeling that on a particular moral issue (opposition to the introduction of bull-
fighting into England) one does not just express a desire that the thing should
not happen, but one does so while feeling that one's desires on such a matter
3. See also essay 10.
 Errors and the Phenomenology of Value 153

are right.4 Mackie thinks that this is a claim to objectivity, and as such errone-
ous. The quasi-realist will see it instead as a proper, necessary expression of
an attitude toward our own attitudes. It is not something that should be
wrenched out of our moral psychology; it is something we need to cultivate to
the right degree and in the right places to avoid the (moral) defect of indiffer-
ence to things that merit passion. This actually illustrates a central quasi-
realist tactic: what seems like a thought that embodies a particular second-
order metaphysic of morals is seen instead as a kind of thought that expresses
a first-order attitude or need. Perhaps the nicest example comes from coun-
terfactuals that seem to assert an anti-projectivist mind-independence of
moral facts: 'even if we had approved of it or enjoyed it or desired to do it,
bear-baiting would still have been wrong' can sound like a second-order
realist commitment directly in opposition to projectivism. But in fact, on the
construal of indirect contexts that I offer, it comes out as a perfectly sensible
first-order commitment to the effect that it is not our enjoyments or approv-
als to which you should look in discovering whether bear-baiting is wrong (it
is at least mainly the effect on the bear).
For the rest of this essay I shall suppose that this aspect of quasi-realism is
successful. So projectivism can accommodate the propositional grammar of
ethics; it need not seek to revise that. On the contrary, properly protected by
quasi-realism, projectivism supports and indeed explains this much of our
ordinary moral thought. But in my experience this explanation is apt to leave
a residual unease. People feel uncomfortable with the idea that this is the true
explanation of our propensity to find and to respect values, obligations, du-
ties, and rights. This unease is perhaps rather like that of nineteenth-century
thinkers who found it so difficult to do ethics without God. It is located in a
tension between the subjective source that projectivism gives to morality and
the objective 'feel' that a properly working morality has. It is this objective
feel or phenomenology that people find threatened by projectivism, and they
may go on to fear the threat as one that strikes at the core of morality. We may
scoff at those who thought that if God is dead, everything is permitted. But it
is harder to really shake off the feeling that if duties, rights, and so forth come
down to that—to the projectivist earth—then they do not have quite the
power or force, the title to respect, that we were brought up to believe.
It is, I think, particularly the side of morality associated with obligation
that is felt to be subject to this threat. Obligation needs to be 'peremptory and
absolute', as George Eliot famously said; it often needs to be perceived as
something sufficiently external to us to act as a constraint or bound on our
other sentiments and desires. The chains and shackles of obligation must
come from outside us. Can anything both be felt to have this power, and yet
be explained as a projection of our own sentiments? The charge will be that
projectivism falsifies this aspect of morality; projectivism will be unable to
endorse this kind of perception of obligation, but must explain it away as a
phenomenological distortion. This perception will be seen as the result of an
4. P. 34.
154 Ethics

error, and realist opponents of projectivism will join with revisionists to urge
that it marks a point at which quasi-realism fails. 5 The realists will trust the
phenomenology, and the revisionists will regret it. We can notice in this con-
nection that when Mackie identifies the error of ordinary thought, he often
points to the 'intrinsic' or 'absolute' to-be-done-ness that certain actions are
felt to possess.6 It is not just the 'intrinsic' value of happiness or pleasure,
because it is less surprising that these values should receive a projective expla-
nation. It is as if the objectivist's error is to think of certain things as obliga-
tory in a way that has nothing to do with us, and about which we can do
nothing: a way that could in principle stand opposed to the whole world of
human desire and need.
Now, admittedly, it might seem from this that the error is to adopt a
deontological rather than a teleological first-order morality. But surely this is
wrong, for Mackie did not want the error to be purely one of adopting a
defective or nonconsequentialist first-order morality. Doing that may be a
natural consequence of a metaphysical mistake, but it is not in itself an 'error'
intrinsic to the very nature of morality. I think instead that Mackie chose the
word, and chose to concentrate upon, obligation because of the absolute and
external 'feel' that he wanted to indicate, and that he felt was not explicable or
defensible on a projective metaphysic. And if he were right, then by threaten-
ing this part of the feeling of obligation, projectivism would indeed threaten
one of the most important and characteristic parts of morality. But is there any
reason to believe that he is right?
The issue will look rather different, depending on whether the difficulty is
supposed to concern the explanation of moral psychology, or its justification.
Consider a very pure case of someone in the grip of a duty. Mabel and Fred
want to marry each other. The opportunity is there, the desires are aflame,
the consequences are predictably acceptable or even desirable. There is only
one thought to oppose it: they have a duty to do otherwise, so it would be
wrong. And this feeling that it would be wrong can wrestle with and some-
times even overcome all the rest. Isn't this mysterious? Called conscience, it
used to be mysterious enough to suggest an internal voice of God standing
outside the natural world of sentiments and desires. On the present line of
thought, it is mysterious enough to suggest perception of an external or objec-
tive moral fact, also standing outside the natural world of sentiments and
desires. Unfortunately, neither of these explanations is more than a gesture. It
is trivial to point out the gaps they leave. But there is a better explanation:
Fred has been brought up in a certain way, and a consequence of this upbring-
ing is that he looks on certain courses of action with horror. He will keep his
self-respect, be able to live with himself, only if he conducts his life in a
particular way, and this prompts a range of feeling that is sufficiently strong to
oppose immediate desire and that gains expression when he describes the
conduct as 'wrong.' Whether it was a good thing that Fred was brought up like
5. Sec for instance the first paragraph of John McDowell's 'Values and Secondary Qualities',
in Morality and Objectivity, ed. T. Honderieh (London: Routlcdge & Kegan Paul, 1985).
6. The Miracle of Theism (Oxford: Clarendon Press, 1982) pp. 104, 115 ff.
 Errors and the Phenomenology of Value 155

that is a matter of judgement, but it can hardly be doubted that it is a good

thing that people should sometimes feel like that, for otherwise they are more
likely to do the most awful things. It is of course a brute fact about human
beings that our sources of self-respect are malleable in this way, but that is a
matter of common observation. Equally, it is a matter of common observation
that there are cultural ways of reinforcing such feelings in elements of the
population that may be in particular need of them: traditionally soldiers and
girls get strong injections of honour and duty.
At the level of explanation, then, it is hard to see why there is any problem
for the projectivist. Indeed, it is hard to see how there could be. For many of the
ingredients of his account will be needed by any other account. For instance, his
observations on the plasticity of our sensibilities, and on the various devices
that lead people to respect different sets of obligations and to value different
aspects of things, will simply be copied by a realist, who will need to say that our
perceptions of moral facts are similarly trained and adapted. As usual, how-
ever, the extra ingredients the realist adds (the values or obligations which, in
addition to normal features of things, are cognized and the respect we then feel
for these cognized qualities) are pulling no explanatory weight: they just sit on
top of the story that tells how our sentiments relate to natural features of things.
If Fred poses a problem, then, it cannot be one of the explanation of moral
psychology but must be one of justification.
If Fred is rational, can his virtue survive his own awareness of its origin
and nature? If Mabel throws into her wooing a whole projective plus quasi-
realist explanation of what Fred is doing when he maintains that it's wrong,
and if Fred is rational, will this not destroy his resolve? Shouldn't he think
something like this: that although he has been brought up to use moral catego-
ries and to think that there are moral obligations and so forth, there are none
really—they are a fiction, or a useful, regulative myth: hence, forget them?
Once again we are reminded of those thinkers who felt that if there were no
God or no afterlife, then it would be rational to ignore the claims of morality
whenever self-interest suggested it. Their anxiety was grounded on a mistake
about rationality, for the altruistic or principled man is no more nor less
rational than the self-interested—he is just different in ways that affect his
happiness and the happiness of communities composed of people like him.
Rationality in itself does not force one sensibility or another on us, just because
we have some belief about the origin of that sensibility. This is obvious if we
take a parallel: Mabel may be tempted to laugh at Fred's moustache; Fred
may seek to dissuade her by telling a projectivist story about the judgement
that something is funny, but there is no reason for him to succeed. Finding
things very funny is perfectly compatible with believing that it is a tendency to
laugh which we project on the world when we do so. It is not uniquely rational
to try to smother our sense of humour because of this belief about its nature.
So Mabel is not irrational if she accepts Fred's theory of laughter and contin-
ues to laugh at his moustache, and by analogy he may be perfectly rational to
accept the projectivist account of morality, and to maintain his resolve just as
forcefully as before.
156 Ethics

I say that he may be rational to do this. But it is possible that he is not, for
an explanation of the origin of a sentiment can diminish its force. For exam-
ple, psychologists sometimes connect humour with sublimated or concealed
aggression. Believing this explanation, and being ashamed of aggressive in-
stincts, it would be rational for me to find fewer and fewer things funny. The
explanation coupled with other values undermines the sentiment. Similarly, a
morality might contain values whose effect, coupled with a projective explana-
tion, is to diminish a subject's respect for some obligations. For example, a
child may be brought up to believe that things really matter only in so far as
God cares about them; learning not to think of conscience as the voice of God
would couple with this attitude to diminish the force with which the child feels
obligations. Or someone might suppose that only commitments that describe
the constitution of the real world have any importance and that all others are
better ignored: a projective explanation of morality may then diminish the
attention that person is prepared to pay to it. 7 This latter attitude is actually
quite common. For example, when people feel uncomfortable about trying to
impose a morality on other people, what troubles them is the idea that moral
commitments lack real, objective truth values certified by an independent
reality. The hope of rehabilitating morality by making it an object of percep-
tion or reason, and thereby giving it a better claim on our attention, bears
witness to the same idea. In each case, however, it is not the explanation of
the practice per se that has the sceptical consequence, it is the effect of the
explanation on sensibilities that have been brought up to respect only particu-
lar kinds of thing. So when people fear that projectivism carries with it a loss
of status to morality, their fear ought to be groundless, and will appear only if
a defective sensibility leads them to respect the wrong things.
So far I have considered this problem only as it affects obligations. But
similar remarks can be apposite in connection with values. It is not initially so
surprising that we can go on valuing the good things of life while knowing that
the valuing is an expression of our own subjective sentiments. This need be no
more odd than that we should go on finding things funny, or painful, or
worthwhile, or beautiful, although God is dead, or although we accept subjec-
tive responses as the source of these reactions. However, David Wiggins has
found a problem even here for the position that he called noncognitivism,
which shares with projectivism the Humean theme that 'ends are supplied by
feeling or will, which are not conceived either as percipient or as determinants
in any interesting way of perception'.8 The core of the charge is, I think, that
projectivism cannot coexist with the way in which we perceive values as
residing in things outside ourselves. It is not entirely clear, because Wiggins
associates with projectivism the repugnant (first-order) doctrine that the only
things that possess any intrinsic value are human states of consciousness. But a
projectivist's sensibility need not, and in my view should not, take this shape.
He can admire features of things regardless of their effects on us: his first-
7. Mackic mentions this kind of psychology on p. 24.
8. 'Truth, Invention and the Meaning of Life', British Academy Lecture (Oxford: Clarendon
Press, 1976).
 Errors and the Phenomenology of Value 157

order morality need be anthropocentric no more than it need be egocentric.

Remember here that a projectivist who avails himself of quasi-realism can
assert those tantalizing expressions of apparent mind-independence: it is not
my sentiments that make bear-baiting wrong; it is not because we disapprove
of it that mindless violence is abominable; it is preferable that the world
should be a beautiful place even after all consciousness of it ceases. The
explanation of what we are doing when we say such things in no way impugns
our right to hold them, nor the passion with which we should do so. But if we
dissociate ourselves from this target, then at this point Wiggins seems to
threaten projectivism no more than the attack deflected in the last paragraph.
It might be that there are people who cannot 'put up with' the idea that values
have a subjective source; who cannot put up with the idea that the meaning of
their life and their activities is ultimately something they confer, and that even
critical reflection on how best to confer them conducts itself in the light of
other sentiments that must be taken simply as given. But this will be because
such people have a defect elsewhere in their sensibilities—one that has taught
them that things do not matter unless they matter to God, or throughout
infinity, or to a world conceived apart from any particular set of concerns or
desires, or whatever. One should not adjust one's metaphysics to pander to
such defects.
There is still that nagging feeling that on this metaphysic 'there are no
obligations, and so on, really' (otherwise, why call the position anti-realist?).
But urging this as a problem confuses two different contexts in which such a
remark might occur. Protected by quasi-realism, my projectivist says the
things that sound so realist to begin with—that there are real obligations and
values, and that many of them are independent of us, for example. It is not the
position that he says these for public consumption but denies them in his
heart, so to speak. He affirms all that could ever properly be meant by saying
that there are real obligations.9 When the context of discussion is that of first-
order commitment, he is as solid as the most virtuous moralist. It is just that
the explanation of why there are obligations and the rest is not quite that of
untutored common sense. It deserves to be called anti-realist because it avoids
the view that when we moralize we respond to, and describe, an independent
aspect of reality. Again, mathematics provides a useful model for understand-
ing this. There are anti-realist views of what we are doing when we practise
arithmetic. But they need not and should not lead to anyone wondering
whether 7 + 5 is 'really' 12, for that would be an expression of first-order
doubt that would not be a consequence of the second-order theory. Arithmeti-
9. Compare Evans on the unintelligibility of one way of thinking of colours as real ('Things
Without the Mind', in Philosophical Subjects, ed. Z. van Straaten [Oxford: Clarendon Press,
1980]). I want to maintain that any genuinely anti-projective attempt to think of obligations or
values as 'real' either is similarly unintelligible or marks a mistake about explanation. This is why
I would deny that there is an aspect of moral phenomenology that gives morality an objective
appearance that quasi-realism must regard as illusory (as McDowell claims in note 4 to 'Values
and Secondary Qualities'). For there is nothing in the appearances of morality to force us to make
the mistake about explanation. Obligations and so forth appear in exactly the way I would
predict.
158 Ethics

cal practice would remain as solid and certain as could be, but explained
without reference to an independent mathematical reality.

III

Thus far I have been using quasi-realism to protect the appearance of moral-
ity: to urge that there is no error in our ordinary ways of thought and our
ordinary commitments and passions. This enterprise will interest a projectivist
most, because it defends him against the most forceful attack he faces, which is
that he cannot accommodate the rich phenomena of the moral life. But realist
opponents of projectivism need to notice quasi-realism as well, since otherwise
they do not know how to launch an attack on projectivism. They would not
have correctly located its strengths or weaknesses. Nevertheless, they could
concede that its defence is successful on these fronts, yet still maintain their
hostility. They can urge that the metaphor of projection fails, or is better
replaced by a comparison between our knowledge of ethics and our knowledge
of other things, such as mathematics or colours. It is this latter comparison that
I now wish to explore. It is not, in my view, right to suppose that there is
immediately an issue between two rival theories of morality. This is partly
because some of the writers I shall mention, who might seem to be offering a
perceptual account of morality, are at least half-inclined to deny that they wish
to offer a theory at all, although that does leave the status of some of their
remarks regrettably unclear. At any rate, as I see it there are in the beginning
two invitations, but they are not so much rivals as complementary to each
other. The one is to explore the idea of a projection upon the world of a
sentiment that we feel; the other is to explore the idea of a perception of a real
property, but one that is intimately related to our own sensibilities. These
mark different directions of exploration, and it should not be obvious at first
sight which will prove the more profitable. I believe that at the end the first
provides illumination where the second runs into obstacles, disanalogies, and
an ultimate inability to say anything. I also believe that the first can explain
and soothe away the fears that lead people to the second—the fear I addressed
in part II, for example, that without obligations of a reality to which a person
cannot aspire, everything is permitted. I shall try to make good these claims by
presenting the 'perceptual' direction in the light of the writings of David
Wiggins, Thomas Nagel, and John McDowell, and more recently Hilary Put-
nam,10 but as I mentioned, I am conscious that it is not easy to extract one
theory, or just one theory, from those writings. However, at least they suggest
a direction of thought, and it is this direction which I want to block.
The opposition understands that projectivism is an explanatory theory that
maintains that moral values are projections of sentiment because we have a
10. Wiggins, op. cit.; T. Nagel, 'Subjective and Objective', in Mortal Questions (Cambridge:
Cambridge University Press, 1979); J. McDowell, 'Are Moral Requirements Hypothetical Impera-
tives?' (Proceedings of the Aristotelian Society, Supplementary Volume, 1978); H. Putnam, Rea-
son, Truth and History (Cambridge: Cambridge University Press, 1981).
 Errors and the Phenomenology of Value 159

better explanation of moral practices if we see ourselves as responsive only to

a value-free world. But according to the opposition, a number of consider-
ations make this an insufficient basis for projectivism. Disquiet can perhaps
be focused under three headings.
(1) Consider secondary properties. Colours (etc.) are real properties of
objects, and this is true even if the best causal explanation of how we detect
them proceeds by mentioning primary properties. Colours really exist, al-
though the reality that contains them is not independent of the fact that there
also exist human modes of perception.
(2) The thesis just put forward will appear surprising only because (i) of a
prejudice that only primary properties, or the properties of some 'ultimate'
scientific theory of things, are real, or (ii) we forget the truth that the world
cannot be 'prised away from' our manner of conceiving it, nor from our
interests and concerns when we do so. Since neither of these motives is
legitimate, there is no obstacle to (1), and to using the parallel with colours to
allow a reality to values and so forth.
(3) It is true that a training of a particular kind is needed to enable people
properly to perceive values and so forth, but this is harmless: people need
training to detect, for example, features of tunes or shades.
I do not suppose that each of the writers I have mentioned would assent to
each of these. For example, although the work of Nagel is prominent in
opening up the idea of a reality that is yet subject-dependent, his own work on
moral motivation is much more concerned with rationality than with any
analogy to the perception of secondary qualities. And Wiggins thinks that the
question of the truth of moral commitments looks very different if we consider
values and if we consider obligations. But I shall put questions of attribution
to one side, simply taking these three themes to form the core of a perceptual
model of moralizing that at least appears to be a rival to projectivism. Is it a
rival, and if it is, then how are we to tell which is better?
Wiggins writes that he has 'long marvelled' at the fact that philosophers
have dwelt frequently upon the difference between 'good' and 'red' or 'yel-
low'. I do not think he should have, unless indeed it is marvellous that philoso-
phers should emphasize things that are banal and basic. At any rate, it is very
easy to rattle off significant differences between secondary properties and
those involved in value and obligation. Here are half a dozen.
(a) Moral properties supervene upon others in a way quite different from
any in which secondary properties do. It is a scientific fact that secondary
properties supervene upon primary properties. It may even be a metaphysical
fact, at least inasmuch as it would offend deep metaphysical commitments to
imagine secondary properties changing while primary properties do not. But it
is not a criterion of incompetence in the ascription of secondary properties to
fail to realize that they must supervene upon others. On the other hand, that
moral properties supervene upon natural ones is not a scientific fact, and it is
criterial of incompetence in moralizing to fail to realize that they must do so.11
11. See also essays 6 and 7.
160 Ethics

(b) The receptive mechanisms whereby we are acquainted with secondary

properties are well-known objects of scientific study. For example, the kinds
of damage to the retina or the ear or the taste buds that result in defective
perception of secondary qualities can be studied. These studies are not at all
similar to studies of defects of character that lead to moral blindness: these
latter studies have no receptive or causal mechanisms as their topic. This is
just as well, for we need to put things in a particular moral light after we are
told about their other properties; we do not also have to wheel a particular
sensory mechanism up against them. Connected with this, and with (a), is the
thought that if our secondary-property-detecting mechanisms fail, we might
expect to know that immediately: it presents itself as a loss of immediately felt
phenomenal quality, just as it does when light fails or we stick cotton wool in
our ears. There is no such loss when we become, say, corrupt. We cannot
become corrupt overnight, and usually we cannot tell when we have done so.
Indeed, it would be a hallmark of many kinds of moral blindness that this is
so. The really coarse man thinks that he is perfectly in order, but that other
people are too fastidious (recognizing that you have become really coarse is in
this way self-refuting: the realization itself shows some residual delicacy).
(c) It is not altogether simple to characterize the 'mind-dependence' of
secondary qualities. But it is plausible to say that these are relative to our
perceptions of them in this way: if we were to change so that everything in the
world that had appeared blue came to appear red to us, this is what it would
be for the world to cease to contain blue things, and come to contain only red
things. The analogy with moral qualities fails dramatically: if everyone comes
to think of it as permissible to maltreat animals, this does nothing at all to
make it permissible: it just means that everybody has deteriorated. 12
(d) The way in which moral practices vary with the forms of life of a
society is not at all similar to the way, if any, in which perceptions of secondary
qualities can vary with those forms of life. Roughly, we expect such percep-
tions to vary in acuity depending on whether the property perceived is impor-
tant to a culture. But once a predicate is located as expressing such a property,
there is no prospect of finding that it has a radically different extension,
whereas many things are evaluated quite differently in different groups or at
different times. Similarly, apart from rare borderline cases, there is nothing in
secondary quality ascription parallel to the 'essentially contested' character of
many moral verdicts.
(e) It is up to a subject whether he cares about any particular secondary
property in any way. If morality consisted in the perception of qualities, there
would be a theoretical space for a culture that perceived the properties per-
fectly, but paid no attention to them. But however it is precisely fixed, the
practical nature of morality is clearly intrinsic to it, and there is not this
theoretical space.
(f) Evaluative predicates are typically attributive: a thing may be good qua
12. I stressed this in 'Rule Following and Moral Realism' in Wittgenstein: To Follow a Rule,
ed. S. Holtzman and S. Leich (London: Routledge & Kegan Paul, 1981). McGinn concentrates
upon the point in The Subjective View (Oxford: Clarendon Press, 1983), p. 150.
 Errors and the Phenomenology of Value 161

action of a commander-in-chief, but bad qua action of a father, just as a man

may be a good burglar but a bad batsman. Secondary properties just sit there:
a red tomato is a red fruit and a red object just bought at the grocer's.
(Wiggins notices this asymmetry after the passage quoted.)
Of course, the extent to which these constitute disanalogies can be de-
bated. But perhaps by way of illustrating their strength in the moral case, we
can notice that sometimes they will not present such a clear picture. For
example, it is very doubtful whether they apply with equal force to the percep-
tion of physical beauty. For at least (a), (c), (e), and (f) can be queried in this
case. And this in turn connects with the sense we can have that sometimes the
beauty of a thing must be perceived and cannot be told. Whereas when it
cannot be told how good something was, this is always because some other fact
about it resists communication—how happy we were, or how brave we
needed to be. So, unlike John Mackie, I incline to find the projective nature of
morality much better motivated than the projective theory applied to aes-
thetic evaluation. But applied to ethics, the cumulative effect of this consider-
ations seems to me to be great enough that expecting a theory of moralizing to
look very much like a theory of secondary quality perception appears a severe
error of philosophical taste. Nevertheless, we cannot depend entirely upon
this cumulative effect. For it will be retorted that mention of secondary quali-
ties just provided an illustration of a combination, a shape of theory, that can
also apply to ethics, however different the subject matter is in other respects.
This is the combination or shape of theory illustrated by (1), (2), and (3). So
the disqualification of secondary properties wins one battle, but it does not by
itself win the war against a 'perceptual' direction.
I will now try to show that once they are properly distanced from other
perceptual analogues, (1), (2) and (3) provide no theory of ethics at all, let
alone one capable of standing up against projectivism. The first thing to
realize is that there is nothing to prevent a projectivist from speaking of the
perception of moral properties, of the world containing obligations, and so
on. We speak of the perception of every single category of thing and fact that
we ever communicate. We speak of perception of numerical truths, truths
about the future, truths about the past, possibilities, other minds, theoretical
entities of all kinds. We speak of perception whenever we think of ourselves as
properly indicating the truth: in other words, whenever we feel able to say
that 'if it hadn't been the case thatp, I would not be committed t op'. But this
is not the end of epistemology, but its beginning, for the theorist's job is to
reflect upon our right to hold such conditionals. Merely reporting that we hold
them is not doing this. Now, in the ethical case, the projectivist, protected
again by quasi-realism, has a story to tell about this: he can explain why
people who are satisfied that their moral sensibilities are functioning well
express themselves in this way. But genuine cases of perception standardly
demand stories with different ingredients. 'If it hadn't been the case that the
shape was square, I would not have believed that it was' can be said because
we are causally affected by shapes and can use those effects to deliver verdicts
on them. 'If it hadn't been red, I would not have believed that it was' can be
162 Ethics

said because I know enough of my normality in relation to other people to

know that only when a thing disposes most people, in good light conditions, to
say that it is red, do I say that it is red. And of course I can be wrong about
that on an occasion of bad light or bad brain state or whatever.
The important point is that speaking of moral perception by itself provides
no theory whatsoever of such conditionals. It provides only a misleading sense
of security that somewhere there is such a theory. The theory is not causal, as
in the case of shape, nor can it be a matter of conformity with a community,
for that just misplaces moral reality, which is not created by community con-
sensus, as (c) reminds us. So what is it? It just doesn't exist. But this means
that the invitation to explore the perceptual direction has simply petered out.
It is as if someone thought that they could seriously provide a theory of
mathematical truth that based itself on the idea that we perceive that 7 + 5 =
12, and then simply turned its back on the disanalogies between such knowl-
edge and ordinary sense perception. It is obvious that until the question of the
status of these conditionals, and our right to believe them, is prosecuted
nothing has been said, or at any rate nothing that cannot be tacked on to the
end of any genuinely successful account of arithmetic. Similarly with ethics.
The nub of the matter, then, is that the projectivist provides an explana-
tion making moralizing an intelligible human activity with its own explanation
and its own propriety, and the opposition provides none, but gestures at an
evidently lame analogy. John McDowell has countered both by claiming that
the explanatory pretensions of projectivism are 'spurious', and by mounting
an opposition case for being able to do something better.13 I take this last
claim first. In effect it uses the 'interest-relative' nature of explanation to cite
contexts in which proper explanations of various verdicts can be given by
citing supposedly projected states of affairs. 'Why did I find that frightening/
funny/appalling?' It can satisfy the interest behind such questions to answer
'Because it merited fright/mirth/horror'. 'Why do we find human happiness
good?' 'Because it is good'. Citing the supposedly projected state of affairs
here plays a part in an explanation, and one that in certain contexts can meet
the need behind the question.
This is true, but by itself it is quite inert. Compare: 'why do we say that the
cube root of 1728 is 12?' 'Because it is 12'. At least if the motive behind the
question is fear that this is an anomalous, surprising thing for us to say, then
the answer can allay it: we are, as it were, only running true to form in such a
verdict. We are not in the grip of strange or local arithmetical error. This
provides an explanation relative to an interest in whether the thing that we say
shows us making a mistake: the reply says that it does not. Similarly in the first
cases: a suspicion that there is something odd about, say, finding the dark
frightening can be allayed by saying that it is what you would expect, that
darkness merits fear. But of course allowing all this goes no way to disallowing
another, wider, explanatory interest that these answers quite fail to engage.

13. McDowell, 'Values and Secondary Qualities,' p. 118ff'.

 Errors and the Phenomenology of Value 163

This questioner may be asking why we find something frightening because he

finds any such reaction puzzling: why do human beings ever feel fear, or get as
far as supposing that anything merits fear? No doubt there is an answer to
hand: one which talks of the behavioural consequences of the emotion, and
their evolutionary advantages to creatures that have it. In a similar vein we try
to place the activity of moralizing, or the reaction of finding things funny, or
the practice of arithmetic. In particular we try to fit our commitments in these
areas into a metaphysical understanding of the kinds of fact the world con-
tains: a metaphysical view that can be properly hostile to an unanalysed and
sui generis area of moral or humorous or mathematical facts. And relative to
this interest, answers that merely cite the truth of various such verdicts are
quite beside the point. This, again, is because there is no theory connecting
these truths to devices whereby we know about them—in other words, no way
of protecting our right to the conditionals I identified.
Could it be held that this explanatory interest is somehow unjustified: that
explanations of a certain type cannot be had, or that the desire for them is the
desire for an illusory, 'external' viewpoint outside of all human standpoints
and perspectives? This is the justification for not having or wanting to have an
explanatory theory along my lines at all. There are two reasons to resist this
'quietist' idea (again, I hesitate to attribute it directly, because the opponents
to projectivism that I have mentioned tend to ride both the perceptual, ex-
planatory line and the suggestion that we need no line at all, in uncomfortable
tandem). The first reason for rejecting it is that we know that it is a common
human option to moralize about more or fewer things in greater or lesser
strengths. The scope of morality can wax and wane, and this makes it urgent to
find an explanation of the practice that goes some way to defining its proper
scope. Secondly, there can never be an a priori right to claim that our activity
in making judgements X permits of no explanation (except the gesture that
says that we perceive X-type states of affairs). You just have to try the various
explanations out. And of course it is particularly perverse to say that any
explanatory attempt in a direction must fail when many appear to have suc-
ceeded well. (I myself think that there is precious little surprising left about
morality: its metatheory seems to me pretty well exhaustively understood. The
difficulty is enabling people to appreciate it.) Could it be said that although
these wider explanatory interests are legitimate, they mark a boundary be-
tween the philosopher and the natural scientist? The evolutionary explanation
of the emotion of fear is not only empirical, but marks a recognizable divide
between any enterprise of understanding fear as we all feel it and know it and
understanding it discursively, in terms of its origins or function. Can the phi-
losopher rest with the phenomenology, and dismiss the rest as sociology, psy-
chology, or someone else's science? The trouble then is that the philosopher
gets to say nothing: Hobbes and Hume and Mackie become classified as
natural scientists, and the only philosophical activity left is playing variations
on the theme of everything being what it is and not another thing. The philo-
sophical spade becomes by definition the one that is turned on the first shove.
164 Ethics

IV

There is one final question I would like to raise, but not to settle. So far I have
discussed the metaphysics as if it were exclusively a second-order issue, with
no necessary consequences for first-order moral theory. But we saw in part II
that when Mackie characterizes the mistake that according to my kind of
projectivism need not be made, he finds it natural to describe it by using a
deontological moral vocabulary. And it is, I think, not a mistake to expect that
a projective theory will consort with consequentialist first-order views. Since
those views are generally downgraded today, it will be important to get that
connection a little bit further into focus, lest projectivism be damned by
association.
It should be said at the outset that there is no essential connection between
projectivism and a consequentialist view in ethics. It could be that all human
beings found it natural to feel certain sentiments, which gain expression as
approval, when faced with some features of action, although those features
have no consequences that explain the approval. This would be parallel to the
way in which certain gestures or timings of actions are hugely funny, although
for no apparent reason. If we had this kind of propensity, it would not alter
the metaphysics—it would not in itself make a realistic theory easier to define
properly, or more likely to be true. But we would say that those features are
good (or right, or whatever) and perhaps we would be unable to envisage
admirable moralities that did not do so: we would have a deontological ethics.
As a metaphysical view, projectivism explains what we are doing when we
moralize. It does not follow that it can explain, or be asked to explain, all the
features of the particular way we moralize. First-order quirks would be as
mysterious to a Humean as they are to anyone else. Nevertheless, it is natural
to associate projectivism with consequentialist moralities, in the following
way. A projectivist is unlikely to take the moral sentiments as simply given.
He will fill out the story by attempting an explanation of the practice of
moralizing. This turns to its function, and particularly to its social function. In
Mackie's terms, morality is an invention that is successful because it enables
things to go well among people with a natural inheritance of needs and desires
that they must together fulfil. Moral thought becomes a practice with a pur-
pose. Saying this goes beyond the metaphysical view, as I have tried to ex-
plain, but it is a natural addendum to it. And if it is right, there must be at
least a limit to the extent to which moral thought can oppose consequentialist,
teleological reasoning. It will be unclear how wholeheartedly a moralist who
understands this second-order theory can endorse deontological views that
stand in the way of all human purpose or fulfilment. Perhaps this is part of the
trouble with Fred and Mabel. Perhaps Fred has a psychology that motivates
him one way, when his and Mabel's happiness would be found another way.
So should he not regard this as an encumbrance: isn't he the victim of an
upbringing, and should he not see his particular psychology as a defect,
whether or not he can effectively work to change it?
 Errors and the Phenomenology of Value 165

This is another version of the problem of part II, except that this time it is
the peculiarly deontological cast of mind that is threatened. But Fred need not
regard himself as a victim, so long as he can endorse the general policy of pro-
ducing human beings whose motivational states are like his. What we then have
is a 'motive consequentialism'—a grown-up brother of rule-utilitarianism.14
The motivations people obey are good in proportion as the consequences of
people being like that (and knowing that other people are like that) are good.15
Actions are then judged either in the light of the motivations that prompted
them, or in the different dimension of their actual effects in the world, depend-
ing on the purposes for which we are judging them. But the position does not
collapse into ordinary act-consequentialism, because for well-known reasons
one would expect a society of people motivated solely by consequentialist
considerations to do pretty poorly. Nor need any such position share the other
prominent feature of utilitarianism that causes dislike: the idea that all values
are ultimately commensurable. The features of human life that we value, and
that would be drawn into any remotely plausible sketch of human flourishing,
very probably represent a bundle of ultimately incommensurable goods,
among which there is no systematic way of making choices. In any case, there is
ample room for a projectivist to respect the reasons that make this seem plausi-
ble. His explanatory project can start from the heterogeneity of ways in which
life can flourish or fail. On the whole, then, I regard the alliance with consequen-
tialism as a strength; to put it another way, it is an alliance with only the best
features of that direction in ethical thought. Of course, there may be features of
some people's moralities that even this diluted motive-consequentialism can-
not well explain, and these it will regret. But I hope I have said enough to show
that none of them could possibly count as integral to moral thought itself.
14. And, fairly clearly, the one that Hume endorsed. Talking of motives is better than talking
of rules (rule-utilitarianism can be charged with 'rule worship' when it tries to give the verdict to a
rule rather than to utility in a hard case. But what charge is there of motive worship?), and as
explained in the text, consequentialism is not subject to at least some of the main objections to
utilitarianism.
15. Rule-utilitarianism is falsely supposed to collapse into act-utilitarianism partly through
neglecting this qualification (e.g., B. Williams, Utilitarianism For and Against [(Cambridge:
Cambridge University Press, 1973), p. 118 ff]). I am contesting what Williams calls the 'act-
adequacy premise'. The consequences of a rule being embedded in a society go well beyond the
consequences of definite commissions or omissions for which the rule is responsible. There is also
the consequence of mutual knowledge that the rule is likely to order action. To illustrate the effect
of this, consider a rule that promises made to dead people should be respected. The main part of
the good such a rule does lies not in any surplus utility of acts performed in accordance with it, but
in the dignity with which one can approach old age or death in a society where it is known that
people have such respect. This value resides not in acts, but in states of mind for which respect for
the rule is responsible.
15. There is a subtle discussion of this relationship in A. Gibbard, Wise Choices, Apt Feelings
(Cambridge: Harvard University Press, 1990), esp. chapter 17.
 9
How To Be an Ethical Anti-Realist

Some philosophers like to call themselves realists, and some like to call them-
selves anti-realists. An increasing number, I suspect, wish to turn their backs on
the whole issue.1 Their strengths include those of naturalism, here counseling
us that there is none except a natural science of human beings. From this it
follows that there is no 'first philosophy' lying behind (for instance) physics, or
anthropology, enabling the philosopher to know how much of the world is 'our
construction' (anti-realism) or, on the contrary, 'independent of us' (realism).
This naturalism bestows small bouquets and small admonishments to each
of the previous parties. The anti-realists were right to deny that there exists a
proper philosophical (a priori) explanation of things like the success of phys-
ics, which some people were acute enough to discern from their armchairs,
while others did not. A scientist can say that there was a certain result because
a neutrino or electron did this and that, but a philosopher has nothing to add to
this. If she tries to say, 'Not only did the result occur because of the neutrino,
but also because neutrino theory depicts (corresponds with, matches, carves at
the joints) the world,' she adds nothing but voices only a vain, and vainglori-
ous, attempt to underwrite the science. This attempt may have made sense in a
Cartesian tradition, when the mind's contact with the world seemed so prob-
lematical, but its time has passed. On the other hand, anti-realists, sensing the
futility of this road, stress instead the dependence of the ordinary world on us,
our minds and categories, and again the additions they offer are unaccept-
able. 2 Characteristically, if realism fails because it is vacuous, anti-realism fails
because it strays into mistakes—making things dependent on us when they
obviously are not, for example. 3 Again, and perhaps even more clearly, it is
plausible to see anti-realism as attempting to theorize where no theory should
be—in this case, making the unnatural, Cartesian mind into a source of
worlds. These theories are naturally described as 'transcendental', and the
word reminds us that for all his hostility to rational psychology, Kant himself
failed to escape this trap.
1. For example, see Arthur Fine, 'Unnatural Attitudes: Realist and Instrumentalist Attach-
ments to Science', in Mind, 1986.
2. On Putnam in this connection, see Ruth Garrett Millikan, 'Metaphysical Anti-Realism',
in Mind, 1986.
3. My favorite example is Putnam, Reason, Truth and History (Cambridge: Cambridge
University Press, 1981), p. 52.

166
 How To Be an Ethical Anti-Realist 167

The transcendental aspect can be seen if we put the matter in terms of

what I call 'correspondence conditionals.' We like to believe that if we exer-
cise our sensory and cognitive faculties properly and end up believing that p,
then p. What kind of theory might explain our right to any such confidence? If
p is a thesis from basic physical theory, only the theory itself. To understand
why, when we believe that neutrinos exist, having used such-and-such informa-
tion in such-and-such a way, then they probably do, is just to understand
whatever credentials neutrino theory has. That is physics. Any attempt at a
background, an underwriting of the conditional from outside the theory, is
certain to be bogus.
When considering such global matters as the success of our science or the
nature of our world, it seems that naturalism ought to win. But in local areas,
it seems instead that the battle can be joined. In this essay I would like to say
in a little more detail why I think this is so. The main problem which I leave
dangling is that of seepage, or the way in which anti-realism, once comfort-
ably in command of some particular area of our thought, is apt to cast imperial-
istic eyes on neighboring territory. The local anti-realist faces the problem of
drawing a line, which may prove difficult, or that of reneging on naturalism
and allowing that global anti-realism must after all make sense. The second
part of my essay is an exploration of this specific problem.
Why can battle be joined in local areas? What I said about physics might
be retorted upon any area. To understand how, when we believe that twice
two is four, we are probably right requires arithmetical understanding. To
understand why, when we believe that wanton cruelty is wrong, we are also
right requires ethical understanding. Where is the asymmetry?
Let us stay with the example of ethics. Here a 'projective' theory can be
developed to give a satisfying way of placing our propensities for values.
According to me, the surface phenomena of moral thought do not offer any
obstacle to this theory. They can be explained as being just what we should
expect, if the projective metaphysics is correct. (I call the doctrine that this is
so 'quasi-realism'—a topic I return to later.) I have also argued that this
package contains various explanatory advantages over other rival and alleged
rival theories. The projectivism is not, of course, new—the package is in-
tended indeed to be a modern version of Hume's theory of the nature of
ethics, but without any commitment to particular operations of passions such
as sympathy. Emotivism and Hare's prescriptivism are also immediate ances-
tors. Anything new comes in the quasi-realism, whose point is to show that,
since projectivism is consistent with, and indeed explains, the important sur-
face phenomena of ethics, many of the arguments standardly used against it
miss their mark. These arguments allege that projectivism is inadequate to
one or another feature of the way we think ethically; the quasi-realism retorts
that it is not, and goes on to explain the existence of the features. Such
features include the prepositional as opposed to emotive or prescriptive form
and the interaction of ethical commitments with ordinary prepositional atti-
tude verbs, talk of truth, proof, knowledge, and so forth. Here, it is the
relationship of that programme to naturalism that is to be determined.
168 Ethics

The first link is this. I think that naturalism demands this view of ethics, but in
any case it motivates it. It does so because in this package the fundamental
state of mind of one who has an ethical commitment makes natural sense.
This state of mind is not located as a belief (the belief in a duty, right, value).
We may end up calling it a belief, but that is after the work has been done. In
fact, we may end up saying that there really are values (such as the value of
honesty) and facts (such as the fact that you have a duty to your children). For
in this branch of philosophy, it is not what you finish by saying, but how you
manage to say it that matters. How many people think they can just announce
themselves to be realists or anti-realists, as if all you have to do is put your
hand on your heart and say, 'I really believe it!' (or, 'I really don't')? The way
I treat the issue of realism denies that this kind of avowal helps the matter at
all. The question is one of the best theory of this state of commitment, and
reiterating it, even with a panoply of dignities—truth, fact, perception, and
the rest—is not to the point.
The point is that the state of mind starts theoretical life as something
else—a stance, or conative state or pressure on choice and action. Such
pressures need to exist if human beings are to meet their competing needs in a
social, cooperative setting. The stance may be called an attitude, although it
would not matter if the word fitted only inexactly: its function is to mediate
the move from features of a situation to a reaction, which in the appropriate
circumstances will mean choice. Someone with a standing stance is set to react
in some way when an occasion arises, just as someone with a standing belief is
set to react to new information cognitively in one way or another. It matters to
us that people have some attitudes and not others, and we educate them and
put pressure on them in the hope that they will.
So far, two elements in this story are worth keeping in mind, for it will be
important to see whether a projective plus quasi-realist story can do without
them. These are: (1) the fundamental identification of the commitment in
question as something other than a belief and (2) the existence of a neat,
natural account of why the state that it is should exist.
Obviously, the emergence of cooperative and altruistic stances is not a mere
armchair speculation. It can be supplemented by both theoretical and empirical
studies.4 It is noteworthy that the account will insist upon the nonrepresenta-
tive, conative function for the stance. The evolutionary success that attends
some stances and not others is a matter of the behaviour to which they lead. In
other words, it is the direct consequences of the pressure on action that matter.
Evolutionary success may attend the animal that helps those that have helped
it, but it would not attend any allegedly possible animal that thinks it ought to
help but does not. In the competition for survival, it is what the animal does that
matters. This is important, for it shows that only if values are intrinsically

4. R. Axelrod, The Evolution of Cooperation (New York: Basic Books, 1984).

 How To Be an Ethical Anti-Realist 169

motivating is a natural story of their emergence possible. Notice, too, the way
the evolutionary success arises. Animals with standing dispositions to cooper-
ate (say) do better in terms of other needs like freedom from fleas or ability to
survive failed hunting expeditions by begging meals from others. No right, duty,
or value plays any explanatory role in this history. It is not as if the creature with
a standing disposition to help those who have helped it does well because that is
a virtue. Its being a virtue is irrelevant to evolutionary biology. There is no such
naturalistically respectable explanation.
The commitment may have psychological accretions consistently with this
being its core or essence. The precise 'feel' of an ethical stance may be a
function of local culture in its scope, or of some of its interactions with other
pressures and other beliefs. A pressure toward action can be associated vari-
ously with pride, shame, or self-respect, and there is no reason to expect a
simple phenomenology to emerge. The essence lies in the practical import,
but the feelings that surround that can vary considerably. There is no reason
for a stance to feel much like a desire, for example. Consider as a parallel the
way in which a biological or evolutionary story would place attraction be-
tween the sexes, and the culturally specific and surprising ways in which that
attraction can emerge—the varieties of lust and love (whose imperatives often
do not feel much like desire either, and may equally be expressed by thinking
that there are things one simply must do; I say more about this later). So, if a
theorist is attracted to the rich textures of ethical life, he need not, therefore,
oppose projectivism. No 'reduction' of an ethical stance to one of any other
type is needed.
Now contrast the kind of evolution already sketched with any that might
be offered for, say, our capacity to perceive spatial distance. Again, what
matters here is action. But what we must be good at is acting according to the
very feature perceived. A visual-motor mechanism enabling the frog's tongue
to hit the fly needs to adapt the trajectory of the tongue to the place of the fly
relative to the frog, and an animal using perceived distance to guide behaviour
will be successful only if it perceives distances as they are. It is because our
visual mechanisms show us far-off things as far-off and near things as near that
we work well using them. That is what they are for. We can sum up this
contrast by saying that although the teleology of spatial perception is spatial,
the teleology of ethical commitment is not ethical. The good of spatial percep-
tion is to be representative, but the good of ethical stances is not.
The possibility of this kind of theory, then, provides the needed contrast
between the general case of science, where an attempt to provide a further,
background theory is transcendental, and the local particular case of ethics,
where there are natural materials for such a story ready at hand. It also means
that philosophers wanting a general realism versus anti-realism issue cannot
take comfort from the local case; the materials to generate theory there exist,
as it were, by contrast with anything that can be provided in the general case.
These simple naturalistic points are not always respected. Consider, for
example, the position associated with John McDowell and David Wiggins.
This goes some way in the same direction as projectivism, at least in admitting
170 Ethics

that a person's ethical outlook is dependent on affective or conative aspects of

his makeup. But it takes those aspects as things that enable the subject to do
something else—to perceive value properties. It is only if one is moved or
prone to be moved in a certain way that one sees the value of things, just as it
is only if one is prone to be moved in some way that one perceives the sadness
in a face. 5 This is supposed to do justice to the obvious point that sentiments
have something to do with our capacity to make ethical judgements, yet to
retain a 'perceptual' and cognitive place for moral opinion.
Let us suppose that this is a substantial theory and different from proj-
ectivism (in the light of what is to come, neither supposition is beyond doubt).
The view is substantial if it holds that changes in one's sensibilities enable one
to do something else: literally to perceive ethical properties in things. Or if the
'something else' is not literal perception, then at least its kinship with percep-
tion must be very close—so close that it cannot be explained as projection of a
stance. For the view is no different from projectivism if this 'something else' is
nothing else at all, but merely a different label for reaching an ethical verdict
because of one's sentiments. In other words, it is only different from proj-
ectivism if this literal talk of perceiving plays a theoretical role, and not just a
relabeling of the phenomena. This is not at all obvious. Theoretically low-
grade talk of perception is always available. Everyone can say that one can
'see' what one must do or what needs to be done, just as one can 'see' that 17
is a prime number. When I said that it is not what one finishes by saying that is
important, but the theory that gets one there, this is one of the crucial exam-
ples I had in mind.
Literal talk of perception runs into many problems. One is that the ethical
very commonly, and given its function in guiding choice, even typically, con-
cerns imagined or described situations, not perceived ones.6 We reach ethical
verdicts about the behavior of described agents or actions in the light of
general standards. And it is stretching things to see these general standards as
perceptually formed or maintained. Do I see that ingratitude is base only on
occasions when I see an example of ingratitude? How can I be sure of the
generalization to examples that I did not see? (I could not do that for colour,
for instance. Absent pillar-boxes may be a different colour from present ones;
only an inductive step allows us to guess at whether they are.) Or do I see the
timeless connection—but how? Do I have an antenna for detecting timeless
property-to-value connections? Is such a thing that much like color vision?
Perhaps these questions can be brushed aside. But in connection with natural-
ism, the question to ask of the view is why nature should have bothered.
Having, as it were, forced us into good conative shape, why not sit back? Why
should this be merely the curtain-raiser for a perceptual system? It seems only
5. John McDowell, 'Non Cognitivism and Rule Following', Wittgenstein: To Follow a Rule,
ed. by S. Holtzman and C. Leich (London: Routledge & Kegan Paul, 1981). Also, Sabina
Lovibond, Realism and Imagination in Ethics (Oxford: Blackwell, 1983). Other writers influ-
enced by the analogy include Mark Plaits, The Ways of Meaning (London: Routledge & Kegan
Paul, 1979) and Anthony Price, 'Doubt about Projectivism', in Philosophy, 1986.
6. John Locke, An Essay Concerning Human Understanding, Book IV. chapter IV, pp. 6-7.
 How To Be an Ethical Anti-Realist 171

to engender dangerous possibilities. It ought to make us slower to act, for we

must process the new information perceived. Worse, it might be that someone
moved, say, by gratitude comes to see the goodness of gratitude and then has,
quite generally, some other (negative) reaction to what is seen. Perhaps typi-
cally, the conative pressure opens our eyes to these properties, about which
we then have a different, conflicting feeling. Or is it somehow given that what
comes back is what went in—that the property perceived impinges on us with
the same emotional impact required for perceiving it? How convenient! But
how clumsy of nature to go in for such a loop! And why did we not evolve to
short-circuit it, as projectivism claims? In other words, we have here the
typical symptoms of realism, which not only has to take us to the new proper-
ties but also has to take us back from them, showing how perception of them
contrives to have exactly the effects it does.
This extravagance came from taking literally the talk of perception made
possible by changes of sensibility. But the theory seems to be meant literally.
Wiggins, for example, thinks that although projectivism can be dismissed
(values 'put into [or onto, like varnish] the factual world'), the right view is
that there are value properties and sensibilities for perceiving them 'made for
each other' as 'equal and reciprocal partners'.7
Can this be understood? Projectivism, from which the theory is supposed
to be so different, can easily embrace one half of the doctrine—that the
properties are made for the sensibility. The embrace ought to be a bit tepid,
because we shall see better ways of putting the view that value predicates
figure in thought and talk as reflections or projections of the attitudes that
matter. But it is the other half, that the sensibilities are 'made for' the proper-
ties, that really startles. Who or what makes them like that? (God? As we
have seen, no natural story explains how the ethical sensibilities of human
beings were made for the ethical properties of things, so perhaps it is a
supernatural story.)
Wiggins, I think, would reply that nothing extraordinary or unfamiliar is
called for here. Refinement or civilization makes both sensibility and prop-
erty. It is the process of education or moral refinement that makes sensibilities
end up in good harmony with values. 'When this point is reached, a system of
anthropocentric properties and human responses has surely taken on a life of
its own. Civilization has begun'. The implicit plea that we get our responses to
life into civilized shape is admirable, but is it enough to locate a view of the
nature of ethics, or is there a danger of confusing uplift with theory? Cer-
tainly, it is true that when we have gone through some process of ethical
improvement, we can turn back and say that now we have got something
right—now we appreciate the value of things as they are, whereas before we
did not. This Whiggish judgement is often in place, but it is, of course, a moral
judgement. Is it not pertinent to explaining how sensibilities are 'made for'
values. Is it a good theoretical description or explanation of the fact that we
7. D. Wiggins, Truth, Invention and the Meaning of Life (British Academy Lecture, 1976),
p. 348.
172 Ethics

value friendship that, first, it is good and, second, civilization has 'made' our
sensibilities 'for' the property of goodness? It seems overripe, since it goes
with no apparent theory of error (what if our sensibilities are unluckily not
made for the properties?), no teleology, and no evolutionary background. Its
loss of control becomes clear if we think how easy it is to generate parallels.
Perhaps something similar made our arithmetical powers for the numbers, or
our tastes for the niceness of things. Or, perhaps, on the contrary, the talk of
our sensibilities being made for the properties is theoretically useless and the
more economical remainder is all that is really wanted.
Might there still be room for a view that the properties are 'made for' the
sensibility, which avoids projectivism? The analogy with colours, for all its
many defects, might be held to open such a possibility. But colour at this
point is a dangerous example. If we ask seriously what colour vision is made
for, an answer can be found—but it will not cite colours. Colour vision is
probably made for enhancing our capacities for quickly identifying and keep-
ing track of objects and surfaces, and this asymmetry with, for instance,
spatial perception remains the most important point of the primary-secondary
property distinction.
Any analogy with colour vision is bound to run into the problem of depen-
dency. If we had a theory whereby ethical properties are literally made by or
for sensibilities, ethical truth would be constituted by and dependent on the
way we think. This might not repel Wiggins. It agrees with the analogy with
colours, and in the course of discussing Russell's worry ('I find myself incapa-
ble of believing that all that is wrong with wanton cruelty is that I don't like
it'), Wiggins freely asserts that 'what is wrong with cruelty is not, even for
Bertrand Russell, just that Bertrand Russell does not like it, but that it is not
such as to call forth liking given our actual responses.'8 But is it? I should have
said not. It is because of our responses that we say that cruelty is wrong, but it
is not because of them that it is so. It is true that insertion of the 'actual' into
the sentence makes it wrong to test the alleged dependence by the usual
device of imagining our responses otherwise and asking if that makes cruelty
any better.9 But our actual responses are inappropriate anchors for the wrong-
ness of cruelty. What makes cruelty abhorrent is not that it offends us, but all
those hideous things that make it do so.
The projectivist can say this vital thing: that it is not because of our
responses, scrutinized and collective or otherwise, that cruelty is wrong. The
explanation flows from the way in which quasi-realism has us deal with
oblique contexts. It issues an 'internal' reading of the statement of depen-
dence, according to which the statement amounts to an offensive ethical view,
about (of course) what it is that makes cruelty wrong. Critics of this explana-
tion allow the internal reading, but complain that the quasi-realist is being
wilfully deaf to an intended 'external' reading, according to which the depen-
8. 'A Sensible Subjectivism', in Needs, Values, Truth (Oxford: Blackwell, 1987), p. 210.
9. The use of 'actual' to make rigid the reference to our present attitudes and thereby fend
off some natural objections to dispositional subjective analyses is exploited in this connexion by
Michael Smith.
 How To Be an Ethical Anti-Realist 173

dency is a philosophical thesis, and one to which the projectivist, it is said,

must assent.10 The crucial question, therefore, is whether the projectivist
wilfully refuses to hear the external reading. According to me, there is only
one proper way to take the question 'On what does the wrongness of wanton
cruelty depend?': as a moral question, with an answer in which no mention of
our actual responses properly figures. There would be an external reading if
realism were true. For in that case there would be a fact, a state of affairs (the
wrongness of cruelty) whose rise and fall and dependency on others could be
charted. But anti-realism acknowledges no such state of affairs and no such
issue of dependency. Its freedom from any such ontological headache is not
the least of its pleasures. A realist might take this opportunity for dissent. He
might say, 'I can just see that the wrongness of cruelty is a fact (perhaps an
external one) that needs an ontological theory to support it—no theory that
avoids providing such support is credible'. In that case I gladly part company,
and he is welcome to his quest—for what kind of ontology is going to help?
The Euthyphro dilemma bars all roads there.11
It is tempting to think: on this metaphysics the world contains nothing but
us and our responses, so the fact that cruelty is bad must be created by our
responses. What else is there for it to be dependent upon? The prejudice is to
treat the moral fact as a natural one, capable of being constituted, made, or
unmade, by sensibilities. The wrongness of wanton cruelty does indeed de-
pend on things—features of it that remind us how awful it is. But locating
these is giving moral verdicts. Talk of dependency is moral talk or nothing.
This is not, of course, to deny that 'external' questions make sense—the
projectivist plus quasi-realist package is an external philosophical theory
about the nature of morality. But external questions must be conducted in a
different key once this package is brought in. We may notice, too, how this
undermines a common way of drawing up the realist versus anti-realist issue,
according to which anti-realism asserts that truth in some or all areas is 'mind-
dependent' and realism denies this. For here is the projection, as anti-realist a
theory of morality as could be wished, denying that moral truth is mind-
dependent in the only sense possible.
The point can be made as follows. As soon as one uses a sentence whose
simple assertion expresses an attitude, one is in the business of discussing or
voicing ethical opinion. Such sentences include The fact that cruelty is wrong
depends on . . .' or 'Our refined consensus makes it true that cruelty is
wrong'. And so on. If one generalizes and says things like 'moral facts depend
on us,' the generalization will be true only if instances are true or, in other
words, if one can find examples of truths like those. Since these ethical opin-
ions are unattractive, they must be judged incorrect, as must generalizations
of them. If one attempts to discuss external questions, one must use a differ-
ent approach—in my case, a naturalism that places the activities of ethics in
the realm of adjusting, improving, weighing, and rejecting different senti-

10. Ouassim Cassam, 'Necessity and Externality,' in Mind, 1986.

11. As essays 1 and 3 attest.
174 Ethics

ments or attitudes. The projectivist, then, has a perfect right to confine exter-
nal questions of dependency to domains where real states of affairs, with their
causal relations, are in question. The only things in this world are the attitudes
of people, and those, of course, are trivially and harmlessly mind-dependent.
But the projectivist can hear no literal sense in saying that moral properties
are made for or by sensibilities. They are not in a world where things are made
or unmade—not in this world at all, and it is only because of this that natural-
ism remains true.
The charge that projectivism refuses to hear an explanatory demand as it is
intended can be returned with, I suggest, much more effect. I was severe earlier
with Wiggins's theoretical description of us as indulging in a kind of coordina-
tion of responses and properties as we become civilized. But it is telling that the
Whiggish appeal to a value ('civilization') is introduced at that point. For the
introduction of values into explanatory investigations is echoed in other writ-
ings in this tradition, notably in those of John McDowell.12 The strategy is that
in a context purportedly comparing explanations of a practice—the practice of
ethical judgement—we allow ourselves to invoke the very commitments of that
practice. Why are we afraid of the dark? Because it is fearful. Why do we value
friendship? Because it is good and we are civilized. Why do I dislike sentimental-
ity? Because it merits it. And so on.
The refusal to stand outside ethics in order to place it is supposed to tie in
with one strand in Wittgenstein. This is the thought that there is characteristi-
cally neither a reduction nor an explanation of the members of any major
family of concepts in terms of those of another. Ethical notions require ethical
sensibilities to comprehend them. Similarly, why should it not require an
ethical sensibility to comprehend an explanation of the views we hold? Only
those who perceive friendship as good will understand why we do so, and to
them it can be explained why we do so by reminding them that it is good, or
making them feel that it is so. The rest—aliens, outsiders, Martians—cannot
be given the explanation, but this is as it must be. What I said about the
explanation of our spatial capacities will make it apparent that the circularity
exists there in exactly the same way. Only those who appreciate distance can
understand the distance-centered explanation of visual perception.
This returns us to a theme that has been touched at many points in this
essay. The insistence on hearing explanatory demands only in a way in which
one can invoke values in answering them had a respectable origin. We agreed
earlier that the parallel would be true of thinking about the correspondence
conditionals in the case of physics. But I hope I have said enough to show that
nature and our theory of nature surround our ethical commitments in a way
that gives us a place from which to theorize about them. No thing and no
theory surround our physics. In other words, the difference in the ethical case
comes in the theses I labeled (1) and (2)—the brute fact that an external
explanatory story is possible. We already know that in even more local cases,

12. For instance in his 'Values and Secondary Properties', in Value and Objectivity, Essays in
Honour of J. L. Mackie, ed. T. Honderich (London: Routledge & Kegan Paul, 1985).
 How To Be an Ethical Anti-Realist 175

where what is at question is not 'the ethical' in a lump but particular attitudes
and their etiologies. Social anthropology is not confined to explaining the rise
of puritanism to puritans or the evolution of polygamy to polygamists. Simi-
larly, nothing in Wittgenstein offers any principled obstacle to explaining the
general shape and nature of ethical attitudes and their expressions in projec-
tive terms.
Indeed, much in Wittgenstein is sympathetic to doing so. Not only is
Wittgenstein himself an anti-realist about ethics, he is in general quite free in
admitting propositions or quasi-propositions whose function is not to describe
anything—the rules of logic and arithmetic, for instance. It is clear that what
he wants to do is to place mathematical practice, not as a representation of the
mathematical realm, but as 'a different kind of instrument', commitment to
which is not like central cases of belief but much more like other kinds of
stance. It is also interesting that some of the apparently irritating or evasive
answers he gives when faced with the charge of anthropocentricity are exactly
those that a projectivist can give if quasi-realism has done its work, and that
according to me, no other philosophy of these matters can give. For example,
when Wittgenstein approaches the question whether, on his anthropocentric
view of mathematical activity, mathematical truth is yet independent of hu-
man beings, he says exactly what I would have him say:
"But mathematical truth is independent of whether human beings know it or
not!"—Certainly, the propositions 'Human beings believe that twice two is
four' and 'twice two is four' do not mean the same. The latter is a mathemati-
cal proposition; the other, if it makes sense at all, may perhaps mean: human
beings have arrived at the mathematical proposition. The two propositions
have entirely different uses.13
The proposition expresses a norm that arises in the course of human activities,
but it does not describe those activities, and it has no use in which the correct-
ness of the norm (the truth of the proposition) depends upon the existence or
form of those activities. That question simply cannot be posed; it treats what is
not a dependent state of affairs belonging at all to the natural world as if it
were.
I have tried to show that naturalism, which turns away from realism and
anti-realism alike in the global case, turns toward projective theories in the
ethical case. This theory is visibly anti-realist, for the explanations offered
make no irreducible or essential appeal to the existence of moral 'properties'
or 'facts'; they demand no 'ontology' of morals. They explain the activity from
the inside out—from the naturally explicable attitudes to the forms of speech
13. Ludwig Wittgenstein, Philosophical Investigations (Oxford: Blackwell, 1953), p. 226. In
'Wittgenstein and Realism' (in Wittgenstein: Eine Neubewehrung, eds. J. Brandl and R. Haller,
Vienna: Holder-Richler-Temsky, 1990), I cite Wittgenstein's nondescriptive treatment of philo-
sophical propositions (grammatical rules), religious propositions (expressions of ways of life),
many apparent descriptions of oneself (avowals), expressions of certainty, as well as modal and
arithmetical commitments. In each case—and these are the only cases in which his later philoso-
phy is seen at full tilt—his quasi-realist leanings are obvious, so it is a pity that he is usually
pressed into the service of blanket minimalism or soggy pluralism.
176 Ethics
that communicate them, challenge them, refine them, and abandon them, and
which so mislead the unwary.
So far I have talked of the issue of mind-dependency in fairly abstract
terms, and relied upon a relatively subtle move in the philosophy of language
to defend my view. I now want to discuss these points in practical terms. It is
evident that a more fundamental mistake underlies some discomfort with
projectivism. The mistake is visible in Wiggins's critique of 'non-cognitive
theories' in his British Academy Lecture.14 It results in the charge that proj-
ectivism cannot be true to the 'inside of lived experience'. Other writers (I
would cite Nagel, Williams, and Foot) seem to illustrate similar unease. The
thought is something like this: it is important that there should be some kind
of accord in our thinking about ethical stances from the perspective of the
theorist and from that of the participant. Our story about ethical commitment
is to explain it, not to explain it away. But projectivism threatens to do the
latter (many people who should know better think of Hume as a skeptic about
ethics, and, of course, John Mackie saw himself as one). It threatens to do so
because it shows us that our commitments are not external demands, claiming
us regardless of our wills or in direct opposition to our passions. It makes our
commitments facets of our own sentimental natures; this softens them, de-
stroying the hardness of the moral must.
From the inside, the objects of our passions are their immediate objects: it is
the death, the loved one, the sunset, that matters to us. It is not our own state
of satisfaction or pleasure. Must projectivism struggle with this fact, or disown
it? Is it that we projectivists, at the crucial moment when we are about to save
the child, throw ourselves on the grenade, walk out into the snow, will think,
'Oh, it's only me and my desires or other conative pressures—forget it'?
It ought to be sufficient refutation of this doubt to mention other cases.
Does the lover escape his passion by thinking, 'Oh, it's only my passion,
forget it'? When the world affords occasion for grief, does it brighten when we
realize that it is we who grieve? (The worst thing to think is that if we are
'rational', it should, as if rationality had anything to tell us about it.)
There is an important mistake in the philosophy of action that, I think,
must explain the temptation to share Wiggins's doubt. The mistake is that of
supposing that when we deliberate in the light of various features of a situa-
tion we are at the same time or 'really' deliberating—or that our reasoning can
be 'modeled' by representing us as deliberating—about our own conative
functioning. Representing practical reasoning as if it consisted of contemplat-
ing a syllogism, one of whose premises describes what we want, encourages
this mistake. But just as the eye is not part of the visual scene it presents, the
sensibility responsible for the emotional impact of things is not part of the
scene it takes for material. Nor is our sense of humor the main thing we find
funny. This does not mean that our sensibility is hidden from us, and when we
reflect on ourselves we can recognize aspects of it, just as we can know when
we are in love or grieving. But it does mean that its own shape is no part of the

14. Truth, Invention, and the Meaning of Life, section 4, note 6.

 How To Be an Ethical Anti-Realist 177

input, when we react to the perceived features of things. Furthermore, even

when we reflect on our sensibility, we will be using it if we issue a verdict:
when we find our own sense of humor funny, we are not escaping use of it as
we do so.
This misconstruction leads people to suppose that on a projective theory
all obligations must be 'hypothetical' because they are properly represented as
dependent upon the existence of desires. But the lover who hears that the
beloved is present and feels he must go, or the person who receiving bad news
feels he must grieve, has no thoughts of the form 'if I desire her/feel sad then I
must go/grieve'. Nothing corresponds to this. The news conies in and the
emotion comes out; nothing in human life could be or feel more categorical.
In ordinary emotional cases, of course, a third party may judge that it is only if
he desires her that he must go; this is not so in ethical cases. One ought to look
after one's young children, whether one wants to or not. But that is because
we insist on some responses from others, and it is sometimes part of good
moralizing to do so.
Once these mistakes are averted, is there any substance left to the worry
about failure of harmony of the theoretical and deliberative points of view? I
think not. Sometimes theory can help to change attitudes. One might become
less attached to some virtue, or less eager in pursuing some vice, after thinking
about its etiology or its functioning. One might qualify it a little (we see an
example in what follows). But sometimes one might become more attached to
the virtue, and sometimes everything stays the same. Does the story threaten
to undermine the promise that the stances cited in this theory of ethics make
good natural sense (does it take something divine to make the claims of obliga-
tion so pregnant with authority)? Not at all—I have already mentioned the
'musts' of love and grief, and those of habit and obsession are just as common.
There is one last charge of the would-be realist. This claims that proj-
ectivism must lead to relativism. Truth' must be relative to whatever set of
attitudes is grounding our ethical stances; since these may vary from place to
place and time to time, truth must be relative. The very analogies with other
conative states press this result: what to one person is an occasion for love or
grief or humor is not to another. Consider a young person gripped by the
imperatives of fashion. The judgement that people must wear some style, that
another is impossible, has its (naturally explicable and perfectly intelligible)
function; it appears quite categorical, for the subject will think that it is not
just for him or her that the style is mandatory or impossible (it was so in the
parents' time as well, only they did not realize it). Yet, surely this is a mistake.
The verdict is 'relative', having no truth outside the local system of prefer-
ences that causes it. The image is plain: a projectivist may inhabit a particular
ethical boat, but he must know of the actual or potential existence of others;
where, then, is the absolute truth?
The answer is that it is not anywhere that can be visible from this sideways,
theoretical perspective. It is not that this perspective is illegitimate, but that it
is not the one adapted for finding ethical truth. It would be if such truth were
natural truth, or consisted of the existence of states of affairs in the real world.
178 Ethics

That is the world seen from the viewpoint that sees different and conflicting
moral systems—but inevitably sees no truth in just one of them. To 'see' the
truth that wanton cruelty is wrong demands moralizing, stepping back into the
boat, or putting back the lens of a sensibility. But once that is done, there is
nothing relativistic left to say. The existence of the verdict, of course, depends
on the existence of those capable of making it; the existence of the truth
depends on nothing (externally), and on those features that make it wrong
(internally). For the same reasons that operated when I discussed mind-
dependency, there is no doctrine to express relating the truth of the verdict to
the existence of us, of our sentiments, or of rival sentiments.
What, then, of the parallel with the other emotions, or with the fashion
example? The emotions of grief and love are naturally personal; if the subject
feels they make a claim on others, so that those unstricken somehow ought to
be stricken, then she is nonrelativistically, absolutely wrong. Similarly with
fashion: the underlying story includes the need for a self-presentation that is
admirable to the peer group, and if what is admirable changes rapidly as
generations need to distance themselves from their immediate predecessors,
then the teenager who thinks that her parents were wrong to like whatever
clothes they did is mistaken in the same way as the subject of an emotion who
imputes a mistake to those who cannot feel the same. But the strongest ethical
judgements do not issue from stances that are properly variable. They may
sometimes be absent, from natural causes, as if a hard life destroys a capacity
for pity. But this is a cause for regret; it would be better if it were not so. In the
variations of emotion, and still more of fashion, there is no cause for regret. In
saying these things I am, of course, voicing some elements of my own ethical
stances, but, as I promised, it is only by doing this that ethical truth is found.

II

If projective theories have everything going for them in ethics, how much can
they jettison and still have something going for them? The two ingredients I
highlighted are the possibility of identifying the commitment in a way that
contrasts it usefully with belief, and a 'neat, natural account' of why the state
that it is should exist. In the case of ethics we have conative stances and a
visible place for them in our functioning. But what of other cases?
Colour commitments might attract attention, because not everybody will
be happy that the agreed story about what colour vision is and why we have it
leaves realism as a natural doctrine about colours. Here the second ingredient
is present. There is a neat, natural story of our capacity for colour discrimina-
tion, and in its explanatory side, both physically and evolutionarily, it makes
no explanatory use of the existence of colours. But there is no way woefully to
contrast colour commitments with beliefs. Their functional roles do not differ.
So there will be no theory of a parallel kind to develop, explaining why we
have propositional attitudes of various kinds toward colour talk, or why we
speak of knowledge, doubt, proof, and so forth in connection with them. If
 How To Be an Ethical Anti-Realist 179

anything can be drawn from a realism versus anti-realism debate over colour
(which I rather doubt), it would have to be found by different means.
Modal commitments are much more promising. Our penchant for necessi-
ties and possibilities, either in concepts or in nature, is not easy to square with
a view that we are representing anything, be it a distribution of possible
worlds or (in the case of natural necessity) a timeless nomic connection be-
tween universals.15
First, consider the case of logical necessity. A theory insisting on a
nonrepresentative function for modal commitment is clearly attractive. Here,
however, although I think the first desideratum is met—we can do something
to place the stance as something other than belief in the first instance—the
second is not so easy. The kind of stance involved is insistence upon a norm,
an embargo on a trespass. Saying that 2 + 2 is anything other than 4 offends
against the embargo, and the embargo in turn makes shared practices, shared
communication possible. So far so good, but what of a 'neat, natural theory'
of the emergence of the embargo? That shared practice should exist is good—
but do they so clearly depend upon such policing? If they do, it appears to be
because of something else: because we can make no sense of a way of thinking
that flouts the embargo. It introduces apparent possibilities of which we can
make nothing. This imaginative limitation is, in turn, something of which no
natural theory appears possible, even in outline. For when we can make sense
of the imaginative limitation, we do find it apt to explain away or undermine
the original commitment to a necessity. If it seems only because of (say)
confinement to a world in which relative velocities are always slow compared
to that of light that we find a relativistic view of simultaneity hard to compre-
hend, then that already shows how we would be wrong to deem the theory
impossible. If it is only because of the range of our colour vision that we
cannot imagine a new primary colour, then we would be unwise to rule out the
possibility that some natural operation might result in our admitting one.
Natural explanation is here the enemy of the hard logical must.
It is not obviously so in the case of natural necessity. Once more the
paradigm is Hume—not the Hume of many commentators, but the real
Hume, who knew that talk of necessity was irreducible but gave a projective
theory of it. The explanation here has us responsive to natural regularity, and
forming dispositions of expectation (we might add, of observing boundaries in
our counterfactual reasoning), which in turn stand us in good stead as the
regularities prove reliable. Here, once we accept the Humean metaphysics,
the naturalism seems quite in place. The upshot—talk of causation—is not
undermined but is explained by this interpretation. This accords exactly with
the case of ethics. There is a difference, however. I do not think metaphysical
obstacles stand in the way of the conception of nature that does the explana-
tory work in the example of ethics. But many writers have difficulty with the
conception of nature that is supposed to do it in Hume's metaphysics of

15. David Armstrong, What Is a Law of Nature? (Cambridge: Cambridge University Press,
1983), chapter 6.
180 Ethics

causation. Regularities—but between what? Events—but how are these to be

conceived, stripped of the causal 'bit' (to use the computer metaphor)?
Events thought of as changes in ordinary objects will scarcely do, for as many
writers have insisted, ordinary objects are permeated with causal powers.
Nothing corresponds to the easy, sideways, naturalistic perspective that strips
the world of values.
What is the option? All sides carry on talk of causation in whichever mode
they find best. The new realists like to produce apparent ontologies—
universals, timeless connections, and the rest. The Humean does not mind, so
long as the explanatory pretensions of these retranslations are kept firmly in
their place (outside understanding). Is there scope for a debate here? It is a
place where the ghosts are hard to lay, and I for one do not like being there
alone in the gloom.

Addendum
Since this essay was written, an influential marker in the debate has been its
nominal mirror-image, entitled 'How To Be an Ethical Realist', by Richard
Boyd (in Essays on Moral Realism, ed. Geoff Sayre-McCord, Ithaca: Cornell
University Press, 1988). Readers of that paper may expect this essay to be a
reaction not so much to Oxford, Wittgensteinian, realism, but to the newer,
naturalistic, Cornell variety. That is unfortunate, but in addition to essay 11 in
this volume, I can here add some brief remarks about that theory.
Cornell realism highlights the possibility of a property identity: to be good
is to be something-or-other natural. The theory may cite sophisticated clusters
of natural properties, or whatever natural properties best satisfy some alleged
folk theory of The Good, but its essence is equally visible if we consider a
simple equation: to be good is to produce human happiness, for instance. The
equation is protected from Moore's open question argument, by being pre-
sented not as any equation of meaning, but as a substantive metaphysical
identity, conceived along the lines of 'water is H2O' or 'heat in gases is molecu-
lar motion'.
I find this approach puzzling, because it is unclear what probems it solves.
It does not speak to moral psychology, because it does not tell us what the
difference is between those who equate creating human happiness with the
good and those who do not. That is, it does not address the question of
whether seeing the creation of happiness under the heading of the good is
possession of an attitude, or a belief, or something else, or whether it connects
with action in one way or another, or with emotions such as guilt and shame
(for a parallel point in connexion with modality, see the Addendum to essay
3). It does not speak to issues of meaning, because we do not know whether
those who resist the equation flout some (hidden) principle of meaning that
governs ethical concepts, and it does not speak to issues of proof and objectiv-
ity, because we do not know whether denying the equation is making an
objective or cognitive mistake or something else. If the equation does not help
 How To Be an Ethical Anti-Realist 181

with issues of moral psychology, or motivation, or meaning, or epistemology,

what good does it do? It seems to be an agreeable afterthought of those
particular people whose ethics allows them to accept the creation of human
happiness as a standard.
One way of missing these questions is to think that realism, at least so far
as a predicate is concerned, is essentially the doctrine that the predicate refers
to a property, or a real property. If we have reference to a real property, then
any sentence in which the predicate is concerned can be given a standard
truth-theoretical semantics; its truth is guaranteed in the way that any more
straightforward truth is guaranteed, and the contrasts that anti-realism tried
to open up are smoothed out. Michael Dummett has expressed this concep-
tion of realism, but he himself appears to go on to point out that unless
reference plays some role in the theory of meaning, it remains possible that
the meaning we give to some set of remarks deserves an anti-realist construc-
tion. Here it is a question of what states of mind end up with the theorist
announcing the identity of reference, and if those states are either covertly or
visibly 'noncognitive', no advance in a realist direction is made.
The crucial point, as far as 'reference' goes, is that in ordinary cases it ties
with meaning: to understand an expression is to have a grasp, explicit or
implicit, of what its reference is. Here, without further argument (which then
bears all the burden), there is no such equation, since people with aberrant
standards understand ethical terms, yet there is no apparent link to the proper-
ties that the naturalist highlights. One might be tempted to think that if there
is a real property identity, anyone missing it must be making an objective,
metaphysical error. But this is not so, since all it takes to miss it is a defective
ethical sensibility.
Once this is understood, property identity drops out of the picture. There is
no harm in saying that ethical predicates refer to properties, when such proper-
ties are merely the semantic shadows of the fact that they function as predi-
cates. A quasi-realist protection of ethical truth protects ethical predicates, and
if our overall semantic picture is that predicates refer to properties, so be it. But
ethical predication remains an entirely different activity from naturalistic predi-
cation, and this is only disguised by thinking of the world of properties as one in
which hidden identities may be revealed by the philosopher-as-scientist.
 10
Attitudes and Contents

I. General Considerations
G. F. Schueler's article puts in a forceful way various reservations about my
treatment of indirect contexts, on behalf of the position I have called 'quasi-
realism.'1 His opposition is, I think, as complete as could be: it is not only that
my treatment has been incomplete, which I happily concede, or that its formu-
lation has been defective, which I am prepared to believe, but also that
nothing like it could possibly succeed. That at least is the proper consequence
of some of his views—on logical form, on validity, and on the nature of
commitment. For example, if showing that an inference has 'the logical form'
or 'is an instance' of modus ponens involves taking it as 'the realist picture' has
it, then no attempt to explain it in other terms will be compatible with its
having that form. Again, if validity is ('as it is used in logic') denned in terms
of the impossibility of premises being true and conclusions false, then persons
reluctant to apply truth and falsity to any of the elements of an inference will
have to admit that the inference is not valid, as the term is used in logic.
Third, if 'talk of "commitments" is problematic for the antirealist', then anti-
realism will make no headway by thinking of a class of commitments more
general than those with representative or realistic truth conditions. Fortu-
nately, none of these contentions seems to me correct. Since the survival of
quasi-realism even in spirit demands their rebuttal, I shall start by considering
them in turn.
1. It is not too clear what it is for an argument to have the logical form of
modus ponens. If it is a diagnosis of its syntactical form, then obviously having
that logical form is compatible with any number of deep and different seman-
tics for the components. To show this, compare 'P, P—> Q, so Q' with the
implication taken as truth-functional, with the same seeming argument taken
as some suppose the English take it: P->Q is the commitment of one who
attributes a high probability to Q conditional upon P. Which is the true modus
ponens? If we plump for either exclusively, we face the uncomfortable conse-
quence that it becomes controversial whether natural English contains any
inferences of the form. If we embrace both, then being of the form modus
ponens is compatible with any number of deep and different explanations of
F. Schueler, 'Modus Ponens and Moral Realism,' in Ethics, 1988, pp. 501-517.

182
 Attitudes and Contents 183

the semantics of the components: how it comes about that we have here
elements describable as true or false, or a connective properly represented by
some or other. The same point could be made with any connective: know-
ing even when to interpret the negation sign of a logic as meaning negation is
no easy matter.2 If quasi-realism, in the form in which I tried to develop it, is
right, the 'deep' semantics of a surface example of modus ponens is to be
explained in a particular, and perhaps initially surprising, way. But it is modus
ponens, for all that. Or, if we say it is not, then we have no effective procedure
for telling when anything is.
2. Perhaps the best way to answer the restrictive view of validity is by
appeal to authority. One might cite imperative logic. Or one might cite the
approach to prepositional inference in terms of coherent subjective probabil-
ity functions, where validity corresponds to there being no coherent function
attributing a lesser probability to the conclusion than to the premises, and
coherence is denned in terms of immunity to Dutch book.3 (This is the ap-
proach that would best marry with the probabilistic view of conditionals
above.) Or one could cite the view of Stig Kanger, that in interpreting a
deontic logic, the extension of the truth predicate to the formulae, which could
equally be regarded as imperatives or expressions of attitude, is a conventional
matter.4 A further reply would draw the usual distinction between an alge-
braic, mathematical, pure, or uninterpreted semantics—itself sufficient to
yield notions of satisfaction, validity, and completeness—versus an applied or
interpreted semantics, in which the valuation clauses reflect something about
the use or meaning of the connectives.5 Formal studies are content with the
first, so that truth-in-a-structure, or satisfiability defined in terms of (for in-
stance) sets of open sets in the topology of the real line, defines validity.6 But
even when we turn to the second, the question of priorities still arises. It does
not go without saying that we interpret the prepositional connectives by draw-
ing on an antecedent understanding of (classical) truth and falsity. Falsity and
negation go hand in hand, and it should not be obvious which is the dominant
partner. The view that it is by knowing how to use the connectives in proofs
that we come to understand them, and hence gain what understanding we have
of the truth tables, is perfectly open. In Prawitz's words, 'Presumably, the
observational consequences that can be drawn from the assumption that a
person knows the condition for the truth of a sentence can also be drawn from
the assumption that he knows how to use the sentence in proofs'.7 The whole
2. B. I Copeland, 'What Is a Semantics for Classical Negation?', in Mind, 1986, p. 478.
3. Hartry Field, 'Logic, Meaning and Conceptual Role,' in Journal of Philosophy, 1977.
4. Stig Kanger, 'New Foundations for Ethical Theory', in Deontic Logic: Introductory and
Systematic Readings, ed. R. Hilpinen (Dordrecht: Reidel, 1971), pp. 55-56.
5. M. Dummett, 'The Justification of Deduction,' in Truth and Other Enigmas (London:
Duckworth, 1978), p. 293. A. Plantinga, The Nature of Necessity (Oxford: Oxford University
Press, 1974), pp. 126 ff.
6. H. Weyl, The Ghost of Modality,' in Philosophical Essays in Memory of Edmund Husserl
(Cambridge, Mass.: Harvard University Press, 1940).
7. D. Prawitz, 'Meaning and Proofs: On the Conflict between Classical and Intuitionistic
Logic,' in Theoria, 1977.
184 Ethics

philosophy of intuitionistic interpretations of the logical constants and of those

who give priority to sequent calculi and natural deduction systems opposes the
simple assumption that an antecedent understanding of 'representative' truth
and falsity affords the only road to understanding validity. A more plausible
view, and one that nicely fits quasi-realism, is that attributions of validity and
application of the truth predicate go hand in hand: I expand on this below.
3. Schueler finds the very notion of a commitment 'problematic for the
anti-realist.' His argument that it is one of 'those terms which seem to entail a
realist picture' is this: 'If I am committed to, say, paying my nephew's way
through school, or to the claim that a Republican succeeded Carter, then this
seems something objective, forced on me by a promise I have made or other
views I hold'. Well it might be, if we could suitably cash the metaphor of
forcing, and suitably interpret objectivity—although whether the objectivity
forced by, say, promises has anything to do with realism is another matter. But
equally, the commitment might not be forced by anything, like the commit-
ment to go for a jog once a week or to improve one's golf. It does not matter,
because I use 'commitment' as a general term to cover mental states that may
be beliefs, but also those that gain expression in propositional form, but which
for various reasons philosophers such as Hume, Ramsey, Wittgenstein, Ste-
venson, Ayer, Hare, and I have seen in terms of such things as acceptance of
rules, changes in disposition, and possession of attitudes, which are worth
separating from beliefs. Some commitments will have nothing to do with
approval: these include the change in one who accepts a rule of inference, or
treats a proposition as necessary, or accords a high subjective probability to Q
upon P, and so on, as well as Schueler's example of his belief that a Republi-
can succeeded Carter. In the sphere of ethics, approval and attitude are natu-
ral terms to work with, but it would not matter if neither fitted exactly or if
better terms for the state in question existed. What is important is the theoreti-
cal issue of whether and why the state is worth distinguishing from belief, or at
least from belief with representational truth conditions thought of realistically.

II. Fast-Track and Slow-Track Quasi-Realism

So far, I have simply dissented from Schueler's reasons for general pessimism
about the approach, intending to show that the ideas behind quasi-realism
survive his onslaught. But he is on stronger ground in attacking the detail of
my treatment. I shall turn to that after taking stock for a moment.
The problem is that of the embedding of sentences that primarily express
attitude, in contexts that might appear to admit only sentences that, in some
contrasting way, express propositions. When I say that these sentences primar-
ily express attitude, I have never intended to deny that they can be regarded
as expressing beliefs or propositions. This opposition would be going beyond
anything I embrace. But I do mean that the right way of theorizing about
them identifies them, in the first instance, as expressing states of mind whose
function is not to represent anything about the world. They express something
 Attitudes and Contents 185

more to do with attitudes, practices, emotions, or feelings arising in contem-

plating some kinds of conduct, with goal seeking, with insistence upon norma-
tive constraints on conduct, and they express nothing to do with representing
the world. In the familiar metaphor, their 'direction of fit' with the world is
active—to have the world conform to them—rather than descriptive or repre-
sentational. I call someone who approves of both this contrast, and this direc-
tion of theorizing, a projectivist. Projectivism may seem to be automatically
opposed to the view that in saying that something is good (etc.) we give voice
to a real belief about it, and it is often so introduced (as labels like 'non-
cognitivism' suggest). But this opposition is not automatic. Subtlety with the
concept of belief, or with the concept of truth or of fact, may enable the
expressivist to soften this opposition. Theory may enable us to understand
how a commitment with its center in the expression of subjective determina-
tions of the mind can also function as expressing belief, or be capable of
sustaining the truth predicate—properly called 'true' or 'false'. I tried to
herald this development with the notion of a 'prepositional reflection' in the
older essay, 'Moral Realism' [essay 6], and it was the point of the last pages of
chapter 6, and of all of chapter 7, in Spreading the Word.8 It means separating
truth (in this application at least) from 'represents' and its allies, but nobody
has ever pointed out the harm in that.
It did, however, seem to me that before this happy result could be secured,
work had to be done. It had to be shown why a sentence with this role could
properly function in the ways ethical sentences do—why it sustains a fully
prepositional role. I now think we should distinguish both a slow track and a
fast track to this result. The slow track involves patiently construing each
propositional context as it comes along. This is the line I took in trying to meet
Geach's problem. Its advantage is that of honest toil over what might seem
like theft; its disadvantage, if Schueler is right, is that it does not work. But
before judging that, I should admit not only that it threatens to look Ptole-
maic but also that it seems not to correspond to any obvious cognitive pro-
cesses we go through. It is not as though construing (say) conditionals with
evaluative components comes harder to us than construing them with ordi-
nary components, and this will need explanation.
Fast-track quasi-realism would get there in better style. It would make
sufficient remarks about truth to suggest that we need a comparable notion to
regulate evaluative discourse (even though this is nonrepresentational) and
then say that our adherence to prepositional forms needs no further explana-
tion than that. The adoption of propositional form and style meets our need to
share and discuss and dissent from attitudes or other stances. It involves only
philosophers in error, and little more need be said. That sounds cavalier, but it
was the line of, for instance, Kant and Nietzsche and probably Wittgenstein,9
8. Simon Blackburn, Spreading the Word (Oxford: Clarendon Press, 1984).
9. I. Kant, The Critique of Judgement, trans. J. C. Meredith (London: Oxford University
Press, 1927). Although Kant believes that the judgement of taste is not a cognitive judgement (p.
41) and is determined by subjective sensation, he also thinks that since we wish to demand
similarity of feeling from others, we 'speak of beauty as if it were a property of things' (p. 52);
186 Ethics

none of whom found any particular trouble in imagining the emergence of a

predicate with a nondescriptive role. Nietzsche puts it roundly:
The pathos of nobility and distance, as aforesaid, the protracted and domi-
neering fundamental total feeling on the part of a higher ruling order in
relation to a lower order, to a below—that is the origin of the antithesis
"good" and "bad" (the lordly right of giving names extends so far that one
should allow oneself to conceive the origin of language itself as an expression
of power on the part of the rulers: they say "this is this and this," they seal
every thing and event with a sound, and, as it were, take possession of it).
Perhaps our general propensity to seal things with sounds needs no de-
tailed explanation or justification (cf. 'this is nice' as a way of voicing plea-
sure, and immediately giving rise to compounds, 'if it's nice, two would be
nicer', etc.). But compromises are possible: the fast track can benefit from
some of the security achieved on the slow, and the slow track can make use
of some of the short cuts of the fast. Or so I shall argue. Notice that,
whichever track we favor, the point is to earn our right to propositional
forms—including the use of a truth predicate. If this is done, any conven-
tional concept of validity tags along—there is a level of analysis at which
modus ponens and the rest are no different when their components are
evaluative from when they are not.

III. Embedding
A parallel to the idea that a certain sentence expresses an attitude ('Hooray
for the Bears') would be the obvious truth that some others express com-
mands ('Go to see the Bears') and questions ('Are the Bears doing well?').
Now, when imperatives and questions give rise to subordinate clauses, the
linguistic forms typically maintain an indication of the original mood, even if
there is another syntactic change: 'he told me to go to see the Bears', 'he
asked me whether the Bears are doing well', 'if / am to go to see the Bears, I
had better have some tea first'. Here the right thing to say is that the subordi-
nate clause maintains the mood of the original, but that it is not uttered with
the force that a direct utterance of the sentence has (nothing is commanded or
questioned). Nevertheless, mood is in some sense primarily an indicator of
force. It is only by understanding what a question or command is that one
understands the function of the interrogative or imperative mood.
There is a primafacie puzzle here: Mood is primarily an indicator of force;
force is lost in subordinate clauses, but mood is not. I do not think, however,
that the puzzle is very deep, although its formal representation can be diffi-
cult. The subordinate clause in 'he said that P' identifies which proposition he
Nietzsche, The Genealogy of Morals, first essay, II; L. Wittgenstein, Remarks on the Foundations
of Mathematics (Oxford: Blackwcll, 1956), p. 163, contains a particularly clear statement of the
view that statements of mathematics mislead philosophers by their descriptive form. See also
essay 9, note 13.
 Attitudes and Contents 187

asserted; the clauses in 'he told me to go to see the Bears' or 'he asked
whether the Bears are doing well' identify which order he gave or which
question he asked. Mood indicates that a question or command is still part of
the topic, even when the overall communication is not itself a question or
command. Technically, I therefore agree with Michael Pendlebury that mood
(or at least the presence of the indicator 'to go . . .' , 'whether . . .') affects
the sense of such clauses: the embedding does not cancel the semantic signifi-
cance of the mood indicator, which is to maintain some connection with an
original command or question.10 It can matter that a question or command is
still in this sense part of the topic. Perhaps the nicest illustration of this is the
difference between 'he knew that the Bears had won' and 'he knew whether
the Bears had won': the first gives us simply the content of his knowledge,
while the mood in the second shows that what he is said to have known is the
answer to a question, which might have been yes or no. Of course, saying that
in this way a command or question is part of the topic is not implying that one
was ever actually uttered—one can know the answer to questions that have
never been asked.
We do not have a mood that indicates in this way that an attitude is part of
the topic. The nearest approximation is in indirect reportage of wishes ex-
pressed in the optative: 'would I were in Grantchester!' can perhaps be re-
ported: he said that he would be in Grantchester, but there is at least a slight
sense of strain. Normally, if I make plain to you what I feel, say about the
Bears, I will most probably do so using a sentence with an 'expressive' predi-
cate: 'the Bears are great!' The report of what I said in indirect speech is then
easy: 'he said that the Bears are great'. According to projectivism, the item of
vocabulary shows that the original utterance was expressive of attitude. In the
subordinate clause, the item remains to make attitude the topic, just as overt
mood indicators do. The person who said that the Bears are great expressed
just that attitude about the Bears. Saying that this is what he did is not of
course endorsing or subscribing to the view, any more than reporting a com-
mand or question involves reissuing it in propria persona.
Suppose we spoke an 'emotivist' language, in which expressions of attitude
wore this function on their faces. We would not have the predicative form, to
keep such expressions in the indicative mood, but an ejaculatory mood, corre-
sponding to that of 'Hooray for the Bears'. It would then be necessary to have a
construction of subordinate clauses corresponding to words such as 'that . . .',
'to . . .', and 'whether . . .', that marks the original attitude as the topic.
There seems no problem of semantic principle about this: 'that!', 'whether!',
and so on might be introduced, so that 'he said that! hooray for the Bears' tells
us which attitude he expressed, 'he wondered whether! hooray for the Bears'
tells us which attitude he was pondering, and so on.
If natural languages have chosen not to register expressive force by a
particular mood, they may have chosen to do it in other ways. And taking

10. M. Pendlebury, 'Against the Power of Force: Reflections on the Meaning of Mood', in
Mind, 1986, pp. 361-73.
188 Ethics

other cases of mood and force as our model, there might be no great difficulty
about imagining it done by an expressive mood and yielding a smooth interpre-
tation of at least some subordinate clauses. Here there is room for the compro-
mise between fast- and slow-track quasi-realism: see how far you get in imagin-
ing an overtly expressive language developed in such ways, and diminish
(even if not to zero) the gap between what it achieves and what we do with
predication, and talk of truth.

IV. A Logic
In his original article, Geach concentrated upon the special case of the ante-
cedent of conditionals. My suggestion involved, first, describing what we are
up to in embedding what is primarily an expression of attitude in a context,
making it intelligible that the context should have a function. Second, it
involved giving sufficient semantic theory to show why we have the way we do
of meeting that need. Thus, in the case of Geach's original conditional, my
suggestion of what we are up to involved taking up an attitude to an involve-
ment of attitude with attitude, or attitude with belief. Such 'second-order'
stances seemed to me both needed in themselves and plausible candidates for
the import of a conditional with evaluative elements. If we use '=>' to signify
the involvement of one mental state with another, the result was that a simple
conditional, 'if lying is wrong, then getting your little brother to lie is wrong',
came out as:

where the \ . . A notation shows that our topic is the attitude or belief whose
normal expression occurs within the slashes. Involvement is not a logical
notion, but neither should it seem mysterious. I tried to explain it by introduc-
ing the idea of a sensibility as a function from belief to attitude, attitude to
attitude, and so on: it is what we would overtly talk about by saying things like
'I really approve of making approval of an action depend on its consequences'
or 'believing that should increase your approval of this'. Endorsing or reject-
ing such an involvement of commitments one with another is an important
thing to do; it is therefore not surprising that we have a simple English form
with which to do it.
Let us now consider modus ponens. We have

Schueler and others rightly raise doubts about the kind of inconsistency in
avowing the two initial attitudes and refusing endorsement of the conclu-
 Attitudes and Contents 189

sion.11 In Spreading the Word, I talked of a 'fractured sensibility' that could

not be a proper object of approval. Schueler reasonably asks why it could not
be an object of approval and whether in any case this smacks more of a moral
or evaluative problem than of a logical one. Yet modus ponens with these
components is surely logically valid, and a proper semantics for expressions of
attitude ought to explain why.
How do attitudes become things that enter into logical relationships, that
matter in the theory of inference? It is well known that logical relationships
between imperatives can be studied by thinking of joint satisfiability—seeing
whether there is a consistent world in which each of a set of imperatives is
obeyed. Similarly, deductive relationships between norms can be studied by
thinking of ideal or relatively ideal worlds in which the norms are met. If we
have here the basis for a logic, it extends to attitude. For H!p can be seen as
expressing the view that p is to be a goal, to be realized in any perfect world—
a world in which ~p is less than ideal, according to this commitment. The
contrary attitude B!p would rule p out of any perfect world, and correspond-
ing to permission we can have T!p, which is equivalent to not hooraying ~p,
that is, not booing p.
Putting attitude to the fore, instead of the more usual obligations and
permissions of deontic logic, promises two gains. The first is that writers on
deontic logic usually interpret 'Op' and 'Pp' as purely propositional by making
them describe what is obligated or permitted by some supposed background
set of norms (the most notable exception to this generalization is Hector-Neri
Castaneda).12 But this divorces them from their ordinary expressive use,
which is not to describe what some (possibly alien) system of norms yields but
to insist upon or permit various things. If the apparatus of deontic logic can be
taken over while this use is kept primary, so much the better. But there is
another gain in taking the portmanteau term 'attitude' rather than the particu-
lar, restricted notion of 'obligation' and 'permission'. This is that the logical
apparatus should apply wherever we have the idea of a goal or aim and the
corresponding idea of something to be avoided or not to be avoided. We need
not be in the realm of the obligatory, or of requirements, but merely in that of
the needed or even just the desirable. Consistency in goals is still a desidera-
tum whose logic needs development. And in fact the deductive apparatus of
deontic logic does not depend in any way on taking obligations and permis-
sions as fully fledged deontic notions. The same structure exists if 'Op' is
interpreted as any kind of view that p be true ideally and 'Pp' as any kind of
toleration of p.
There is nothing surprising about using realization of goals or ideals as the
final test for consistency. The ordinary way of finding whether recommenda-
tions are consistent is to imagine them carried out and see if that can be
consistently done. But Schueler rightly raises a problem that might affect the
extension to attitude. This is that consistency in attitude is not a particular
11. Bob Hale, The Compleat Projectivist', in Philosophical Quarterly, Spring 1986, pp. 65-
85, anticipates the difficulties with validity.
12. H-N. Castaneda, Thinking and Doing (Dordrecht: Reidel, 1975), esp. chapter 2.
190 Ethics

virtue. I may wish that p and wish that ~p without particular shame. I may
desire that;?, and desire that q, but not desire that p & q: I want to spend the
evening at the theater, and I want to read my book, but I do not want to read
my book at the theater. There is a sense in which my goals are inconsistent—
they cannot all be realized—but, if this does not matter, then it is not suffi-
ciently like the vice of inconsistency in belief to form the basis of a logic.
My comment on this is threefold. First of all, I think part of the objection
comes from confusing desires with wishes. Inconsistent wishes may not matter
because in wishing or daydreaming we are spinning fictions, and inconsistent
fictions do not matter. This is because there is no connection with action. But
for all that, inconsistency in real desire may matter. Incompatible and there-
fore unrealizable goals are bad in a way quite analogous to the way in which
inconsistent beliefs are bad. The latter cannot represent the world properly;
but the former cannot represent how to behave in the world properly: they
cannot mate together with beliefs, in the usual belief-desire psychological
framework, to direct effective action. The man who believes both that it is
raining and that it is not is badly placed to act if he wants, say, to avoid getting
wet. But so is the man who believes that it is raining but wants both to get wet
and not to get wet.
The second point to notice is that attitude and desire are capable of qualifi-
cation. I may be subject to some desires, or some pressures (tiredness, mood)
that suggest reading a book and others that suggest visiting the theater. Do I
want both to read a book and to go to the theater? It is a crude way of
representing my state. Perhaps I want to read a book inasmuch as I am tired,
and want to go to the theater inasmuch as I like company. I feel the different
pressures, but it is at least as natural to say that I don't know what I want to do
as it is to say that I want to do both. I can indeed say that I would like to do
both, but that takes us back to the realm of wishes (I would like them not to
conflict somehow). So one way of diminishing the attraction of inconsistent
desires is to remember the difference between full-scale, all-in desires, and
attractions or pressures that are not yet resolved.
Even if this point were contested, a third defense is waiting. Although I
have urged the advantage of thinking in terms of a catholic conception of
attitude rather than of strict deontological notions, we could restrict our-
selves to concepts of being for or against or neither for nor against things
where consistency does matter. If this is a more limited range than the full
spectrum of desire, this need not matter. If, for instance, it embraced only
desires that one was inclined to submit to public scrutiny or translate into
practical advice, then there would be a corresponding restriction of the inter-
pretation of notions such as 'goal' or 'ideal'. That is fine, provided the
relevant attitudes satisfy the constraints when it comes to interpreting the
logic in a domain such as ethics. Since ethics is at bottom a practical subject,
this is to be expected.
In the usual metaphor, the direction of fit between desires and the world is
opposite to that between beliefs and the world. The desire that p dictates
action if it is deemed likely, but avoidable, that p, whereas the belief that p
 Attitudes and Contents 191

needs abandoning if it is deemed likely that ~p. 13 But since belief and desire
do each have a direction of fit and a content, then each should be fitted to play
a role in a logic of consistency. A person may flout the demand of consistency
in complicated or demanding situations, but only at the cost of tension: his
goals cannot be realized, or if he has inconsistent beliefs, the world cannot be
as he represents it. It may be admirable that we sometimes get into states
where we feel that tension, but this is so for belief as well as desire. It could
not be admirable in general, and it could not be true in general, for these
states are essentially characterized by responsibility either to the world, in the
case of beliefs, or in our response to the world, in the case of attitude.
I therefore reject Schueler's contention that there is no legitimate notion
of inconsistency. But it remains to be seen whether it ratifies my assault on
modus ponens, or any other natural inference pattern. Meanwhile, there is
another natural worry about my proposal. As it stands it yields no smooth
extension to other propositional contexts. For instance, simple disjunction
with an evaluative component does not yield an obvious second-order atti-
tude. 'Either Johnny has done something wrong, or Freddy has' is not well
represented as //!(\B!/\ OR \B!F\) where 'OR' introduces a kind of disjunc-
tive relation between attitudes. Because even if the idea of a disjunctive
relation between attitudes makes sense, one might know that one of them has
done something wrong but quite disapprove of taking up a negative attitude
toward either of them—if neither has yet been proved guilty, for instance. The
stance that H!(\B!J\ OR \B!F\) expresses seems to be that of someone who
endorses only psychologies that contain at least one of the embedded attitudes
(this would be the natural interpretation of disjunction), but this is not at all
the same as the stance of someone who thinks that either Johnny has done
something wrong or Freddy has. One could interpret disjunction by first
translating it into the associated conditional and then using the account of
conditionals on that. But there is something ad hoc about such procedures.
They take the theory too far from anything that seems necessary for the
ordinary truth-functional disjunction, and their very unnaturalness raises
again the question of adequacy. Even if the notion of involvement gives a
reasonable surrogate for implication, there may be no such notion naturally
available in each case of potential embedding.
Suppose then we take the theory of inference as primary. If we ask what
these embeddings are for, the immediate answer is that they mediate infer-
ence. They show us the deductive relationships between our commitments,
and between our commitments and our beliefs. So rather than replace logical
constants, as in the approach I just gave, we might try to retain them and to
provide an interpretation of embeddings of attitude in the contexts the deduc-
tive system is to treat: in the first place, contexts provided by the truth
functors.
We know, or think we know, what the negation, disjunction, and conjunc-
tion of an ordinary proposition is. It needs showing that we have any right to

13. Michael Smith, 'The Humean Theory of Motivation', in Mind, 1987, pp. 36-61.
192 Ethics

extend those notions to cover expressions of a different kind. Thus, in the

language to come, H!p is to be treated as a well-formed formula capable of
entering the same embeddings as p. Even if this provides a language that is
formally workable, it still needs showing that it provides one which is
interpretable—in which Hp can still be regarded as fundamentally expressive
of attitude.
Consider first negation. What can ~H!A mean? Schueler might say: noth-
ing much to do with truth or falsity, and since that reversal is the fundamental
effect of negation, it cannot mean anything to apply the notion here. But I
have already remarked that it might go the other way round: falsity of p is the
truth of the negation of p. Ordinary negation is expressive of denial: ~p is
that proposition whose expression denies p. There is a clear corresponding
relation between attitudes: there exists that attitude that 'denies', or rejects,
having p as an aim or goal. If H!p expresses the attitude of endorsing the goal
p, —Hip then expresses that of opposition: tolerating ~p or allowing it as
consistent with an ideal world. So we can say that T!A is substitutable for
~H!A, and H!A for ~T!A. Such a conversion drives external 'negations' on
attitudes inward. So even if the original occurrence of the external negation
made us uneasy, the unease is dissipated formally by the conversion and
philosophically by recognizing sufficient analogy between conflict of attitude
and conflict of belief.
What of other truth-functional contexts? An important feature of infer-
ence using prepositional calculus embeddings is that they can all be repre-
sented by the normal forms of conjunction and disjunction. In a tableau
development each move either adds to the string (as when A, B, are appended
under A & B) or divides the string (as when A makes one branch, and B
another, under A v B). Now let us suppose that we are involved with an
evaluative commitment, H!A or T!A in a propositional calculus embedding.
We can see what this means if we can interpret the strings in which it issues.
This reduces then to the problem of interpreting these two elements in a tree
structure. But it is easy to see what a string represents if, underneath this
embedding, we get H1p occurring alone (e.g., under p & H!p we get H!p).
This means that the initial complex commits us to the attitude, and this is not
hard to interpret. Being against clergymen and for free love commits one to
being against clergymen, and the notion of consistent realization of aims or
goals shows us why.
There remains the case in which a tableau under a complex branches, and
H!p belongs to one of the branches. The interpretation is that one potential
route to drawing out the consequences of the complex involves this commit-
ment, although another may not. Thus p v H!q issues in a branch; it is the
commitment of one who is what I shall call tied to a tree—that is, tied to
(either accepting that p, or endorsing q), where the parentheses show that this
is not the same as (being tied to accepting p) or (being tied to endorsing q).
Rather, the commitment is to accepting the one branch should the other prove
untenable. The essential point is that this is a quite intelligible state to be in.
Philosophically, we justify the procedure by analogy with the ordinary notion
 Attitudes and Contents 193

of accepting a disjunction, which similarly ties one to a tree of possibilities,

and formally, the language admits of identical deductive procedures.
How does this relate to the original proposal for treating the conditional?
Under a material conditional A->B, we get the tree with ~A in one branch
and B in the other. Suppose then we treat Geach's conditional 'if lying
yourself is wrong, getting your little brother to lie is wrong' this way. Some-
one asserting it is tied to the tree of (assenting either to 'lying yourself is not
wrong' or to 'getting your little brother to lie is wrong'). What was right
about my original proposal is that being so tied is in this case characteristic
of a particular value system or set of attitudes. Only someone with a certain
view of the relation between doing things directly and doing them indirectly
is apt to assent to the conditional. This represents the reason for his being
tied. What was inelegant about the original proposal, I now think, was
putting that directly into the content of the conditional itself. The assent to
the conditional itself does not tell us why someone is tying himself to that
tree, it tells us only that he is tied and that we can use this fact in assessing
the consistency of his position. I return to this below, after detailing the logic
a little.
I do not want to claim finality for the semantics I shall now sketch, but it
illustrates how a logic might be developed, and it shows that notions of
inconsistency and satisfiability can be defined. (It also bears out Kanger's
remarks mentioned above.) It uses Hintikka's notion of a set of 'deontic
alternatives'.14 In Hintikka's semantics, the central notion is that of norms
obtaining in a possible world, and of the deontic alternatives to that world
being the possible worlds that are in accordance with those norms. Hintikka
compares this notion to Kant's 'Kingdom of Ends' (Reich der Zwecke): it
represents a 'mere ideal' (Kant: 'frellich nur ein Ideal') that is not realized but
that we nevertheless must be able to think of consistently.
In Hintikka's development we work in terms of a model system or set of
model sets. A model set is a partial description of a possible world or alterna-
tive. A set of sentences including oughts and permissions will be satisfiable if
it is embeddable in such a set. This means that a set of sentences L is satisfi-
able if and only if there is a model system S and a model set m E 5, such that L
is a subset of m. Logical truth of A is unsatisfiability of the negation of A; B is
a logical consequence of A if and only if A—>B is valid, that is, (A & ~B) is
unsatisfiable.
Where we are not concerned with attitudes, the notion of a model set m is
defined in a standard way:

14. J. Hintikka, 'Deontic Logic and Its Philosophical Morals', Models for Modalities
(Dordrecht: Reidel, 1969).
194 Ethics

if (Ax)p G m and if the free singular term b occurs in the sentences of m,

then p (b/x) G m;
p (alx) is the result of replacing the variable x by the singular term z
everywhere in p.

Henceforth I shall depart somewhat from Hintikka's terminology, in order

to separate some of the main ideas more obviously. Suppose that to a standard
first-order language we add operators H! and T! subject to the condition that
if A is a well-formed formula, H!A is well formed and T!A is well formed.
Suppose now we start with a set of sentences L, which may contain sentences
with these operators among them. We begin by defining a next approximation
to the ideal, L*, of L.

(Ii) If H!A G L, then H!A L*;

(Iii) If H!A G L, then A G L*;
(Iiii) If T!A L, then a set L* containing A is to be added to the set of
next approximations for L.
(liv) If L* is a next approximation to the ideal relative to some set of
sentences L, then, if A E L*, A E subsequent approximations to
the ideal L**, L*** . . .

We can say that a set of final ideals, {L*** . . .} of L, is obtained when further
use of these rules produces no new sentence not already in the members
L*** . . . of the set.
The set of sentences L may contain disjunctions or conditionals ready to
be treated as disjunctions in the deductive apparatus. We can say that to each
branch of a disjunction there corresponds a route to an ideal.
We can then define

A set of sentences L is unsatisfiable iff each route to a set of final ideals S

results in a set of sentences S one of whose members contains both a
formula and its negation.

These rules need a little gloss. Obviously, Iii embodies the aim that an
ideal relative to a starting set of attitudes is obtained by specifying that the
goals expressed are met. Rules Ii and Iii merely ensure that the attitudes
specified originally remain in the subsequent realizations (if it is good that
people are kind, it remains good in a world in which they are kind). The
statement T!A gets handled slightly differently. It sees A as compatible with,
but not mandatory to, perfection. When tolerations are in play we have to
consider both developments in which they are realized and also developments
in which they arc not. One next approximation for (T!p & T!~p) should
contain p, and another ~p, but the fact that they are inconsistent with each
other does not reflect back on the original sentence. So to assess consistency,
we need to think of formulae as producing a set of next approximations. The
 Attitudes and Contents 195

rule is that if T\A is present in a set, there must be a next set in which A is
present, although it is not to be in all. This means that we shall have to
consider sets of next approximations to the ideal and sets of final idealiza-
tions. Intuitively, what is to matter is whether each such set is consistent.
It may be that modifications of Liv would be desirable. What is obtained by
realizing a toleration might not automatically feed through to subsequent
approximations. The intuitive idea would be that something may be tolerated
now, whereas were some ideal to become realized, it would no longer be
tolerable: in that case, Iiv would need qualification.15
We want to iterate the procedure of generating a next ideal. This can be
done by repeated use of these rules. If L* is already a next approximation to
the ideal and contains a sentence A, then except where A derives from
realization of a toleration, it must transfer to further approximations to the
ideal L**. . . . This is not so in general, if A belongs to an original set L (for
it may be a pity that A: A & H! —A is consistent). Here the idea is that once
we are following out what is so in the progressive approximations to a
perfect world, any realized ideal remains realized. The denizens of paradise
do not move.
To get a feel for such a semantics, consider the formula H!p-*p. This is not
valid: {Hip, ~p} is satisfiable. The next approximation is {H!p, p}*, and this is
a final ideal and is consistent. (As the gloss of liv showed, p does not transfer
through to L*—as far as the original set goes, p may be a pity, and this is
reflected in its absence from the final ideal.) Now however consider H!
(H!p->p). This is valid, for T! (H!p & ~p) is not satisfiable. By Iv, (H!p, ~p)*
must be added as a next ideal. But the ideal under that is {Hip, p, ~p}**,
which is inconsistent. Here is the operation of Iiii: it was already in this
working out of the ideal that ~p, so it stays there when we further consider
the ideal obtained by realizing H!p, and it generates inconsistency.

15. In his 'Moral Quasi-Realism' in Reality, Representation, and Projection, eds. J. Haldane
and C. Wright (New York: Oxford University Press, 1993), Bob Hale shows that this is too strong
as it stands. It gives us that T!H!p yields H!p, whereas intuitively tolerating admiring p is not the
same as admiring p. On the other hand, we need that T(H\p & ~p) is inconsistent since
otherwise H\(H\p—>p) would not be valid. The rule that is wanted needs to be that when (but only
when) what you tolerate is a wide scope H! . . . this may fairly be taken as realized in assessing
you for consistency. I discuss this further in 'Realism: Quasi or Queasy', in the same volume.
Technically, the rule that emerges is that if L* is a next approximation relative to some set of
sentences L, then if L* contains H(X), then a subsequent approximation L** contains X and all
the other sentences of L*.
In chapter 5 of Wise Choices, Apt Feelings (Cambridge, Mass.: Harvard University Press,
1990), Allan Gibbard gives an elegant account of indirect contexts based on a possible world
semantics. Instead of the consistency of realization of a set of attitudes, Gibbard relies on the
consistency of a set of norms, and this undoubtedly makes it easier to generate a smooth logic,
since the multiple embeddings that loom rather large in this essay are not a problem. It would be
possible to urge that starting with an abstract structure, such as a set of norms that is itself already
in the space of logic, being subject to norms of consistency, is starting with more Fregean baggage
on board than one ought to be allowed. But Gibbard and I are agreed that in the end this is not an
important charge, since he gives a functional story for the emergence and propriety of norms of
consistency parallel to that which I give for sets of attitudes.
196 Ethics

Transferred to these terms, Hintikka's main worked example is this. Prior

once took it as a 'quite plain truth' of logic that

But {Hip, p H!q, T! ~q} is perfectly satisfiable. In tree form and using the
notion =0=> to signify adding the next approximation to the ideal, we get

The right-hand route is bound to contain the inconsistent set, but the left
yields none—reflecting the fact that if something that ought to be so is not,
obligations or norms or goals consequential on its being so need not be held
either. As in the first example, something here is a pity, and this is reflected by
the fact that H! [H!p & (p-+H\q)->H\q] is indeed valid.
The possibility of valid formulae with wide scope H! suggests a notion of
'deontic validity' (Hintikka's term): in other words, although A may be consis-
tent, H\A need not be. In turn this gives us a needed notion: a person may be
something worse than 'immoral', or possessing contingently defective atti-
tudes, but not be 'inconsistent' in the sense of believing anything logically false.
He may simply have ideals or goals which admit of no consistent realization.
This logic yields one reduction principle immediately: H! Hip yields Hip.
What about T!T! T!p? {T!T!p & H!~p} =0 {Tlp & H\~p}, and this too is
inconsistent. This reflects the 'one-dimensional' way in which realizations of
goals are treated: we look through H! and T! to see what happens when they
are realized, and this transparency extends to iterations of them. Many com-
plexities could be introduced at this point and in connection with iterations
generally.
The semantics also generates interesting sidelights on the original pro-
posal for treating modus ponens. Suppose we took as an example 'X is, good,
if X is good Y is good, so Y is good'. Then a treatment like my original
might render it

And this is indeed valid. But the satisfaction is short-lived, for if we turn
instead to 'Giving makes happiness; if giving makes happiness, then Christ-
mas is a good thing, so Christmas is a good thing' a parallel treatment
renders it
 Attitudes and Contents 197

This argument is invalid. There is no way of reimporting the original p into the
set of final ideals. Clearly a treatment that makes a big asymmetry between
these two arguments is suspicious. However, my old proposal did not quite
have this form—it involved no propositional calculus embedding of H!p. Now
that such a form is available, obviously we short-circuit these proposals simply
to get H!p, H!p^H\q, so H!q, and similarly for the second version. Each of
these is valid.
So is Schueler right that my original proposal shows no inconsistency in the
set containing the premises of a modus ponens inference, but a denial of the
conclusion? As I mentioned, the original proposal for conditionals took seri-
ously the idea that they create an indirect context—one where the propositions
or attitudes normally expressed become, in some sense, the reference or topic
of the utterance. In the present development conditionals are treated as disjunc-
tions and broken open for example by tableau methods. Is there essential
opposition here? Not necessarily. The issue is whether we can interpret endorse-
ment of (\A\=^>\C\)—the original interpretation—as equivalent in strength to
(~A v C)—the place conditionals now have in the logic. Only a little leeway
with 'endorsement' and 'involves' ('=>') is needed; as much, in fact, as gives the
material implication its usual right to be thought of as rendering a conditional.
Say: endorsing the involvement is tying oneself to the tree: in other words, tying
oneself to restricting admissible alternatives to those in which ~A, and those in
which C. You have one or the other. And the effect of this on the theory of infer-
ence, when A or C or both are evaluative, is brought out in the model theory.

V. Conclusion
Slow-track quasi-realism will want to say that these proposals analyze or give
us the logical form of the arguments we are considering. Fast-track quasi-
realism need not say this. It can say: 'all this is very interesting'. It shows how
little is involved if we imagine us jumping ship—changing from an expressivist
language to our normal forms. But it is unnecessary to claim that we make no
jump at all. That would involve, for instance, defending the claim that nega-
tion is absolutely univocal as it occurs in ~H!p and in ~p, and similarly for the
other constants. But this need not be claimed. All we have is sufficient similar-
ity of logical role to make the temptation to exploit ordinary propositional
logic quite irresistible—and that is what we naturally do. The expressivist
language serves as a model showing us why what we do is legitimate—but that
may be all. This is what I meant by saying that fast-track quasi-realism can
benefit from the security provided by slow-track. I like this methodology. We
bootstrap our way into appreciating how prepositional expression, the arrival
of indirect contexts, and the arrival of the truth predicate, meet our needs,
without in any way betraying the original, economical, metaphysical vision.
At times we may have taken steps, benefiting from what are only analogies
between these kinds of commitments and beliefs, in order to treat the former
as we treat the latter. But if so, these steps are little and natural.
 11
Just Causes

In this essay I reflect mainly upon some difficulties that have been urged
against a 'projective' explanation of ethical discourse. These difficulties in
turn introduce wider problems about explanation and natural kinds, and per-
haps indicate further things to be said about the ongoing realism versus anti-
realism debate in ethics and elsewhere.
There are two kinds of difficulty I have in mind, both urged by Nicholas
Sturgeon in his paper 'What Difference Does it Make Whether Moral Real-
ism Is True?'1 The first I shall call the argument from sense, and the second
the argument from explanation. The first points to some aspect of moral
communication that a theory, such as mine, might find it difficult to construe.
The second does the same for explanations.
The essential structure of the argument from sense is this: A projective
theory has two items at its disposal in its sketch of a moral psychology. There
is the input side, or natural features of the world to which the subject is
responsive in forming an ethical commitment. And there is the output side, or
kind of commitment formed—the attitude or pressure upon action or choice.
It is important to projectivism that it can usefully identify these, initially at
least, without relying upon a moral vocabulary. It must be able to pin down
the input without saying (for example) that it is a perception of justice or
obligation or value. And it should pin down the output without saying that it is
just the ethical response, the 'kind of reaction we have when we approve of
something'. I put in the caveat that this should be so initially, for it is not true
that this austerity has to go on forever. If a projectivist can, as I put it, earn
the right to think in terms of an ethical truth and ethical features of things,
then he can talk of us responding to them and responding ethically, when he
works within the system. It is just that he thinks he has an explanation of what
this amounts to that gives no ground to realism.
Now as we look at the individual members of a community, we will find a
great deal of variety on each count. I shall take the case of social justice as an
example. Then holding constant a given output, say, one of active approval
1. Nicholas Sturgeon, 'What Difference Does It Make Whether Moral Realism Is True?', in
Spindel Conference: Moral Realism, Southern Journal of Philosophy Supplement, 1986. Similar
points are made in a variety of ways by many contributors to Essays on Moral Realism, ed. Geoff
Sayre-McCord (Ithaca: Cornell University Press, 1988).

198
 Just Causes 199

and encouragement, we find that people differ in the kinds of arrangements

that prompt it. And holding constant a given input—say, belief that a society
is governed by Rawls's difference principle—we find a variety of outputs,
ranging from adoration to execration. We find too attitudes that some people
take, and that others never take: white-hot indignation, for instance. And
some people respond to features that others are pretty much incapable of
noticing: male patronization or racial stereotyping, for example.
Against this heterogeneous background, one constancy stands out and
now appears odd. This is that we allow almost anyone, whatever the input and
whatever the output, who so expresses herself, to be talking about justice. In
Professor Sturgeon's example, Socrates and Thrasymachus disagree about
what justice is—but they are not talking at cross-purposes. They are not doing
so, even if one of them finds in it arrangements that leave the other cold, and
finds in it an object of concern and respect while the other scorns it. The
challenge now is plain: how can the projectivist find a single topic for such a
debate? How can he explain the intuition that the two protagonists in such a
debate are both talking about justice, but disagree about its nature?
As Sturgeon notes, this a resurgence of an old objection to naturalistic
reductions in ethics. In this form of theory, if my standards for a thing being
just are that it is X, and yours are that it is Y, then in saying that it is just I am
saying that it is X, and in saying that it is not just you are saying that it is not Y;
the decisive objection is that both these remarks might be true, so that the
disagreement has disappeared. And one good reason for emotivism was that it
restored the clash, as one of attitude and policy. What is now claimed is that
this form of solution is itself subject to the same objection. It restores disagree-
ment in cases where the same reaction is in play: you being for something and
I being against it. But it can find no identity of content if both standards and
reactions are fluid. Yet this is apparently something we manage to do.
There are of course limits to our charity. Suppose you claim to find it
unjust that I start walking more often than not on the right foot, but further-
more add that this is one of those cases of injustice that actually deserves
either no attention or even commendation. We cannot well get on terms with
this: the diagnosis has to be either that you are joking, or that you mis-
apprehend the term. A Satan who wants evil to be his good is one thing; a
Satan who wants this, but also has bizarre standards for evil, perhaps applying
the term mainly to acts of fairness and benevolence, would be best reinter-
preted.2 Thrasymachus himself faces this peril: if he sticks to the view that
justice is what is in the ruler's interest,3 perhaps he is not best seen as discuss-
ing justice at all. Add the quirk that he is contemptuous of whatever it is that
he is thinking of, and the difficulty increases.
However, even if we chop off some of the more extreme cases by this
means, there will be plenty of variation both of standard and of response, and
2. Milton, Paradise Lost, Book IV, 1.108. It is notable that Satan actually makes this exclama-
tion in the course of bidding farewell to hope, fear, and remorse—it is not a mere change in
standards, or a kind of existentialist experiment with a new fashion in values, that is in play.
3. Republic 339a.
200 Ethics
the challenge to find the one content within it still remains. The other resource
the projectivist has is clearly more promising. This is to remember other
examples of our determination to find uniformity in sense, even in the face of
massive differences of cognitive shape. Consider a straightforward belief—say
that some tree is an elm. Any two people are likely to differ in some way in
the evidence that will prompt them to think this and in the significance they
attach to it. But we will again allow sameness of content, even embracing the
most idiosyncratic input side (the person who cannot tell elms from beeches)
and bizarre output side (the person who thinks that if it is an elm it will house
a witch). So there is the same challenge to pin down sameness of belief given
wide idiosyncracy of conceptual role.
There is a well-known solution to this problem. That is that a group
speaking a language holds itself together with what Putnam called 'division of
linguistic labour' and what in Spreading the Word I called 'deferential conven-
tions'.4 In the clearest cases, we institutionalize an authority, and anyone
using a term is liable to be regarded as having used it with the meaning vested
in it by the authority. The familiar cases include deference to a deposit of
usage or to scientific authority or dictionary authority. Anyone using a term is
therefore in principle up for being told what it means, and his own peculiar
configuration of knowledge and ignorance is not the arbiter of it. We might
say that in this picture sameness of meaning or content is something that is
imposed, although only on someone who himself abides by, or recognizes the
authority of, the convention. We would not find someone liable in this way if
they belonged to a genuinely different social group, and as in the interpreta-
tion of dialect we can show some flexibility here. But the reason for deeming
meaning to be common is clear enough: for the purposes of smooth communi-
cation we need to look past the idiosyncracy and take away as much of a
common, communicable content as we can. Of course, this imposition can
seem like a myth. We can remain impressed by the endless divergence of
cognitive architecture even within a small group at one time, and lose any
sense that we are right to impose a common meaning. But two thoughts
should dispel this vertigo. One is that we need the imposition for communica-
tion. And the other is that we probably need it, or something like it, for any
thought at all. If the content of a person's remarks were read straight back
from her actual dispositions to application and withholding, then any standard
of correctness and incorrectness vanishes, and with it the possibility of truth or
falsehood.5
If this is the way to think of sameness of meaning, how will it apply to
ethics? In ethics we have no single authority, and we are dealing with 'essen-
tially contestable concepts'—ones where we predict the differences of input
and output so far described. But the basic mechanism remains the same. We
need to see disagreement where it exists, and for what it is. We do so by
presuming a common, communicable topic. The 'default' position in any
4. Spreading the Word (Oxford: Clarendon Press, 1984), pp. 130 ff.
5. The modern locus classical is Saul Kripke's Wittgenstein on Rules and Private Language
(Oxford: Blackwell, 1982).
 Just Causes 201

conversation is that we are discussing the same issue, even if we have different
opinions about it. Only wild idiosyncracy overturns the default.
The question now is whether the projectivist can simply take this over. It
seems that he can. He should do so by imposing as much identity of content as
the need for communication and debate requires, hence, respecting the possibil-
ity of a common topic through different standards and responses. This means
that the initial presumption in any debate about justice will be well seen as
doubly quantificational. We presume that the subject responds to some aspect
of a society's structures with some response of favour or disfavour. The evi-
dence that can be cited may have to fall within a certain range, but it may not be
the same as mine. That is what makes arguing with each other worthwhile.
Similarly, the responses must fall within a certain range—they must have some-
thing to do with a taste or distaste, with encouragement or acquiescence versus
rejection and regret—but, again, they may not be the same as mine. Some
words delimit the range more closely than others. The so-called 'thick' ethical
terms—courageous, courteous, fair-minded, and so on—delimit the input side
closely, but leave open quite a variety of responses. Output-oriented terms—
outrageous, intolerable, squalid—express relatively specific responses, but
might be applied to a wide range of subjects.
A great deal of effort goes into locating the responses. Is what we value
what we desire to desire?6 Or what we fully or finally desire to desire to . . . ?
Or is it something else altogether? The perspective I am urging will not see
these issues as definitional. What may be needed is not a swift, a priori
boundary, but a longer exploration of the possible contours of desire, re-
morse, higher order desire, and other pressures on choice and action. Milton's
Satan can be represented as wanting to make evil his good because of his
actual self-disgust, indeed his suffering, at being forced (by the need to have
something over which to reign) to make such a choice. For this reason it
would be absurd to see him as using the term only in an 'inverted commas'
sense. He knows that it is evil he is choosing. But if his strategy is successful,
and he succeeds in driving out not only remorse but all the other elements that
enable him to see his own plight as desperate, then indeed the interpretation
would start to waver. He would no longer be seeing evil as his good, but
merely doing evil and seeing it as good.
We might be tempted to separate only some desires as values—ones that
arise from or survive some process of universalization, for example. But we
then face the need to conduct a dialogue with people who reject any such
process. We should not see them as having no values at all, thereby closing off
the vocabulary for dialogue or dispute.
For an example of indeterminacy of interpretation, consider the prisoners'
dilemma. Suppose we face the need to educate people to obtain the highest
social good from such situations, and do so by putting various pressures on
them in childhood. Suppose this works for someone sufficiently well that he

6. For a defence of this choice, see David Lewis, 'Dispositional Theories of Value', in
Proceedings of the Aristotelian Society, Supplementary Volume, 1989.
202 Ethics

feels motivated to cooperate. But suppose too that he desires to desire not to
cooperate (he thinks cheats get more out of life). We might say that he desires
not to have the cooperative values he has. Or we might say that he fails to act
on his real (egoistic) values. My attitude is that we do not have to choose:
whichever way we put it, the task of getting him back into feeling comfortable
with the cooperative choice (seeing it as better) remains the same. Saying
anything further about the responses, just as saying more about the standards,
will be entering into an essentially first order, moral debate. It will be a matter
of assessing the right importance to give to various kinds and configurations of
desire or other pressure, and views on that can differ. This in turn makes it
important to frame the conversation as one around a common topic, the aim
being to explore differences of sensibility, and to promote improvement and
agreement. Words like 'justice' stand exactly where they need to, as the focus
for the exploration.
This simple story may seem strange to anyone conditioned to expect a
theory of meaning to take a certain form. We may have been led to expect talk
of 'descriptive meaning', located by the input side, and 'prescriptive' or emo-
tive meaning, located by the output. Yet here is a view that acknowledges
neither. This seems to me pure advantage. Meaning is properly talked of only
where we have convention. But there is no convention governing the selection
of standards in ethics: someone who approves of the wrong things is not
unconventional in the way of someone who uses a word wrong. A group
cannot put pressure on an isolated moral view by claiming that it is a conven-
tion that other things are good or bad. Nor is there a convention governing the
actual output: someone who feels a different way about the importance of
something (that may be described as justice or virtue) cannot be brought into
line by citing meaning. This is why, in my view, Professor Hare's battle to
make universalizability, and hence perhaps utilitarianism, emerge from the
meaning of ethical terms is quixotic. The reason is that the upshot would be
only that there is a convention that anyone reasoning without (roughly) ab-
stracting away from her own position is not to be described as reasoning
'ethically'. If this is true, we may still prefer and campaign for other ways of
reasoning. But it is unlikely to be true, for that very reason: we will need to
emerge with such a preference or campaign. And the terms in which to do so
best are obvious: one side thinks that universalizing is the way to find out
which things are good, and the other side thinks something else is. Any theory
on which the terms have been ('by definition') coopted by just one side lacks
the ordinary means to locate this debate. Things would be easier if words
enforced ethics, but they will not do so, any more than reason will.
The points I have made also diminish the force of one attack often made
on projectivism. This is its difficulty in defining, in other than ethical terms
(approval, guilt, etc.), the state of mind projected. I reply that, because of the
mechanisms given above, we would not want a definition serving only to rule
out cases that we need to 'rule in'. It would be damaging if the rich textures of
ethical response were somehow visibly sui generis, incapable of being ex-
plained as developments from less particular responses of desire, or of willing-
 Just Causes 203

ness to bow to pressure or to conform to group observances. But there is no

reason to believe that the natural history of ethics witnesses any inexplicable
leaps of psychology.
It is always necessary, in discussing moral realism, to press the question of
how realists escape the traps laid for their opponents. Why exactly does the
existence of a 'real' moral property help to restore sameness of sense to
differently functioning participants in a debate? Thrasymachus and Socrates
have to be brought into some kind of relationship with any such property. But
their different standards and different views of the importance of the issues
they are discussing make that just as hard for the realist as for anyone else.
Suppose, for example, that justice in a state really supervenes upon the exis-
tence of the difference principle governing distributions of wealth, and sup-
pose further that rightness belongs to the person in whom justice trumps all
other practical considerations. Then consider Thrasymachus, who thinks, or
expresses himself as thinking, that justice is whatever is in the interests of the
strongest, and (let us say) that in any case it is right to pay no attention to the
claims of justice when you can get away with it. How does 'realism' help us to
see his words as possessing the same sense that they do in the mouth of a
Socrates? Prima facie he is talking of something other than the real thing,
something that arises differently and makes different claims altogether.
The usual answer draws on the analogy with natural kind terms. Two
theorists may equally be talking of water, although differing about its nature;
why should two theorists not be talking equally of justice, although differing
about its nature? But the solution to the problem in the case of natural kinds is
quite different. A man may be known to be talking of water, in spite of
making misguided claims, because we can trace the causal line: we know the
stuff that prompts his claims, and toward which his actions are directed. But
there is no analogous procedure in the case of justice—no independent trac-
ing of the focus of attention that can counterbalance the impression that he is
talking of something else entirely.
A 'naturalistic' realist, still working with the 'water = H2O model', must
reply to this objection by filling out the contact we have with the real proper-
ties. Perhaps there is some explanatory role for the property with which he
identifies goodness or justice, whereby we all touch the right property in spite
of other differences. But this suggestion falls to an unpleasant dilemma. There
are two ways we might be locked onto moral properties, through being in
contact with natural ones with which they are identical. We may be in contact
cognitively (thinking of such properties, or in terms of them), or it may be that
they, or more accurately instantiations of them, are in some favoured way
explanatory of our thought and practice. The water/H2O case indicates the
latter model. But it then meets the fatal objection that someone might surely
be thinking and talking of justice even in a world in which it is not to be found.
Suppose justice in society is the instancing of the difference principle. Sup-
pose no societies do this. Then no such instancing is causally or in any other
way responsible for any of my thoughts, and on the natural kinds model, just
as it is said that Twin Earthers do not think of water as they splash around in
204 Ethics

XYZ, so it would have to be said that nobody in a bad world can think or
dream of justice. The essence of this argument is that it is always tokens or
instances of properties, rather than properties tout court, that play explana-
tory roles independently of the content of thought. I can certainly change my
behaviour through thinking of a chiliagon. But 'chiliagonhood' or the prop-
erty of being a chiliagon cannot influence me otherwise, except through being
instanced. I cannot change my behaviour or thought 'because of the property
of being a chiliagon' unless this means either that I suddenly started to think in
terms of such a property, or was influenced by the existence of an instance, for
example by bumping into one.
On the other hand, if the contact is thought of more cognitively, so that
someone can indeed think of a justice that is never found, the issue of interpre-
tation remains moot. Suppose a Nozick strenuously insisting that justice in
society has to do with transfers of ownership, and has nothing whatsoever to
do with the embedding of the difference principle. What is the story saying
that he is 'really' well seen as thinking and talking of justice (the instancing of
the difference principle) in spite of his views? How do we see him as actually
concerned with this property, in spite of the disavowal? I see no independent
reason for doing this, and the realist falls victim to the same argument—from
different standards to different sense—that he used to overturn his Moorean
and Fregean ancestors. In short, for the realist to deem him to be talking of
the favoured property involves just as much of an imposition, or policy of
looking past the differences of standard and conceptual role, as it does for the
anti-realist: here as elsewhere the reality of the property is quite idle. In fact,
if anything it makes matters worse, since for the realist there may be a hidden
right or wrong, depending on the explanation of the subject's thought, where
for the anti-realist our purposes give us automatic standards of right imposi-
tion. For all I can see, on the realist picture, one person may be misguided in
standards and motivations, but 'locked onto' the right property; another with
the same functioning may be 'locked onto' the wrong one, so although we
engage them equally and accuse them both of mistakes about justice, we
might be wrong to do so. For the anti-realist there is no such possibility: the
interpretation is, as it were, downwind of our interpretative needs.
Although I think the above dilemma is fatal to this kind of realism, it does
introduce the whole issue of explanations involving moral properties, and
brings us to the second of Professor Sturgeon's objections. As examples in
which the moral feature of a situation is attributed an explanatory role, he
reminds us that poverty and injustice may be cited in the same breath as
explanations of human events such as revolutions and cites the plausible truth
that children thrive when treated with decency and humanity. And the pro-
jectivist, he believes, cannot tolerate this, since according to him there is no
feature there to play any causal role at all.
There are, I think, three compatible lines that a projectivist might use to
explain these explanations. The first would be to query whether the explana-
tions are elliptical. Someone citing injustice as the cause of revolution might
be adverting to the population's perception of injustice, or belief that they are
 Just Causes 205

victims of injustice. And there is nothing surprising in this having effects (if
the projectivist is still avoiding talk of perception and belief, then substitute
the population's attitude toward their arrangements). This is a plausible con-
struction of the revolutions case: at least, I cannot readily think of mecha-
nisms whereby injustice brings about revolutions except through the popula-
tion's awareness of it. However, the children case is different: we might
explain a child thriving by mentioning the decency and humanity with which
she was brought up, without supposing that the child perceived the upbringing
as decent and humane. The child might thrive because of it even at a time
when she lacks the concepts to recognize the upbringing as decent and hu-
mane. This brings us to the second projectivist line, which is that the explana-
tion points downward to the properties upon which the moral verdict de-
pends. According to me, an upbringing is decent and humane in virtue of
other features—meeting the child's needs, engaging with its attempts at ac-
tion and communication, and so on—and I may simply point toward those
other, causally powerful properties by using the moral predicate. An educa-
tion will have to be decent and humane in virtue of other properties, and
those other properties will bring it about that children thrive (indeed, that's
largely why they get the honorific titles).
This may seem ad hoc at first sight, so it is worth dwelling a little on other
cases where the same pattern is found. In what Jackson and Pettit call 'pro-
gramme explanations' we mention some supervening feature of a state in
order to point to the existence of some other, underlying property, whose
instancing is the causally powerful event.7 Thus Fred's coughing at a particular
place and time in the concert irritated the conductor. But it is quite proper to
explain the conductor's annoyance by mentioning that someone coughed. We
do not have to pick out the individual event: it may be enough to say that it
was an event of some sort. Or we may explain opium's effect, if not very
helpfully, by saying that it has a dormitive virtue, or in other words possesses
some property that causes sleep. The actual chemical property responsible
may be unknown to us, but by saying that there is one there we at least rule
out other kinds of explanation. Again, I may explain the appearance of a
letter on the screen when I press a key by saying that the computer is running
Word, and again it is plausible that this explanation works merely by telling us
that there is some property in the electronics correlating keystroke and letter.
The presence of quantification in these explanations is important, given
what I said earlier about the attribution of content to the moral remarks of
others. If you tell me that injustice caused the revolution, I understand that
there is some property that you take to give rise to injustice, and that caused
the revolution. I must, in my own assessment, separate the truth of the causal
story you are pointing toward, from my own verdict on whether it amounts to
injustice. Knowing your devotion to Rawls I may have a good idea of the first,
and I may agree, for instance, that the absence of the lexical priority of liberty
indeed caused the revolution. Perhaps I would not myself call that unjust, but

7. F. Jackson and P. Pettit, 'Functionalism and Broad Content', in Mind, 1988.

206 Ethics

I can assent to the explanation without endorsing the verdict you pass on the
feature. Of course, it can work the other way round, if I assent to the injustice
but deny the causation. And finally there is scope for misunderstanding, if I
am not privy to the kinds of feature you take to justify the verdict. We might
each say that injustice caused the revolution, but be completely at cross-
purposes: I am pointing to the absence of government by the differing princi-
ple, and you to the Right have in mind the punitive top taxation rate of 30
percent.
I cannot readily think of an explanation citing a moral property that will
not succumb to one of these two diagnoses. But I would like to introduce the
third, more speculative strategy, even if we do not have to rely upon it. This
allows that there exists a moral feature, injustice, and even allows that it can
itself be causally relevant, in spite of having a projected origin. It will not be
obvious that this position is available to the projectivist. But here is a sketch
of the way it might be. The first part is to establish our right to talk of the
moral feature or property. Now, if the projectivist adopts quasi-realism, he
ends up friendly to moral predicates and moral truth. He can say with every-
one else that various social arrangements are unjust, and that it is true that
this is so. Once this is said, no further theoretical risks are taken by saying that
injustice is a feature of such arrangements, or a quality that they possess and
that others do not. The first step, in other words, is to allow propositional
forms of discourse, and once that is done we have the moral predicate, and
features are simply abstractions from predicates.
The second part is to explain how such a feature might be causally rele-
vant. This is not easy, because we have deep problems in thinking out how any
features are causally relevant. Strictly speaking, we want to say that it is their
instances that are relevant, but when we think in detail of instantiation, we
tend to lose sight of the way the property itself relates to the causal powers of
the instance. This is the problem that prompts writers following Lewis and
Davidson to equate mental events with physical events, or in general 'super-
vening' events with subvening realizations, generally thought of in terms of
particular physical states or mechanisms. Thus in Jackson and Pettit's exam-
ple, the causally powerful event is a particular disturbance of Fred's coughing
and also the realization of someone coughing. In their handling of the case,
mentioning that someone coughed is a 'programme explanation' for us, the
causally relevant feature being its 'realization' on the occasion in a particular
sequence of events. The idea is that we look down from its being someone
coughing, to find its being that particular sequence of physical events, and it is
only then that we find causal power.
I do not regard this diagnosis as compulsory.8 We might well want to say
that the causally relevant feature was just someone coughing—why is this not
a property whose instancing actually made the conductor angry? One way, of
preserving our right to say this will be via identifying the event of someone
coughing with the event of molecules moving. But there may be other ways. I
8. See also essays 13 and 14.
 Just Causes 207

do not need to adjudicate that here. On any view, the conductor, we suppose,
is apt to explode when anybody coughs, so making somebody cough gives us a
control over him: it is a feature of the audience that we can pick out and work
upon, regardless of its particular 'realization', and when this is so we want to
talk about causal relevance. This is so whether the causal chain works via the
conductor's thought that someone is coughing, or whether the coughing sim-
ply acts on his emotional system regardless of his recognizing it as a cough.
The same might be true in the moral case. If we can talk of a feature of a
situation, via the development above, we can also use it in controlling events.
Thus, suppose a variety of social arrangements qualify, in our view, for injus-
tice. Suppose I want to promote a revolution. Working from within my moral
view I can issue the instruction that you bring about injustice, as a means to
doing it: I need not care which particular realization of events you go in for, if
I think that any will do equally. In the same way I can raise the temperature or
get someone to cough, without caring which particular distribution of kinetic
energies of molecules, or which particular person, realizes the state. So, simi-
larly, I may want you to be controlled by the moral reaction: to create a state
of affairs that merits disapprobation, since I believe that if you do that revolu-
tion will follow on, perhaps as others come to share the same reaction. To
sensibly issue the order I must suppose that your standards for injustice are
roughly the same as mine. It will not be much good if you think that socialist
Utopias are intrinsically unjust, whereas I had in mind something more to the
right. But similarly, it is no good using a badly calibrated thermometer to
control the central heating.
It may be instructive to compare the case of colour. We know enough
about the physics of colour to know that the best explanation of our colour
classification is the upshot of a complex function of the energy levels of the
three different kinds of receptor.9 Without endorsing a dispositional view of
the term, it is nevertheless true that as far as science is concerned, coloured
things share only a disposition to activate the same classification from (crea-
tures like) us. If this peculiar instrument, the visual system, did not exist, no
physicist would find anything but a shapeless disjunction of different physical
states in all the things we see as red. Nevertheless, we can certainly cite
colours in causal explanations: the butterfly escapes being eaten because its
colour matches the branch; the picture did not sell because the colours were
too bright. Colour may be what we need to use, as evolution does, to control a
certain effect, and again, where this is so we talk happily of causal relevance.
It should, however, be noticed that the only straightforward example of such
causal power is via perception (conforming to the first kind of case above).
It seems to me, then, that there is nothing in our explanatory use of moral
properties that lies beyond a projectivist's grasp. But by now many readers
will be wondering what remains that is distinctive of projectivism. In the old
days (they will complain) it was easy: we knew what emotivists and pre-
scriptivists stood for. But this new, conciliatory position is harder to pin down,
9. C. L. Hardin, Colour for Philosophers: Unweaving the Rainbow (Indiana: Hackett, 1986).
208 Ethics

harder to recognize as a position. My view has always been that it is a question

of explanation, of 'placing' our propensity for ethics within a satisfactory
naturalistic view of ourselves. I distinguish then between the ingredients with
which you start, and what you can legitimately end up saying as you finish. To
place ethics, I deny that we can help ourselves to moral features and explana-
tions from the beginning. We have to see them as constructions, or, as I call it,
projections, regarding ourselves in the first instance as devices sensitive only
to natural facts and producing only explicable reactions to them. The aim is to
explain and make legitimate the emergence of full-blown ethics on this aus-
tere basis. But there is no need to deny, as the error theorists did, that the full-
blown system is in order as it is. Nor is there any need to regret apparently
realistic features of it, if these can be earned from the slender basis. Among
these will be a propensity, when we are working from within the system, to use
ethical predicates in the ways that Sturgeon highlights.10 But, as I say, I can
see no sellout to any other metaphysic of ethics in the fact that we do so.
There is a final, much less easily dissected aspect of the matter that I
should like to end by mentioning. It is, I think, undeniable that moral anti-
realism is often seen as a dangerous doctrine, a more or less surreptitious
denial of the importance of ethics or of the possibility of genuine first-order
ethical theory. It is thought to consort with lack of real seriousness, just as
relativism is felt to undermine any real commitment. Nearly all writers who
currently incline to promote real first-order theory—Parfit, Foot, Nagel, Mc-
Intyre, Dworkin, and many others—find it important to eschew anti-realism,
and those who embrace it in one form or another—Mackie and Williams
come to mind—are less enthusiastic about the possibilities of substantive
ethical theory. I incline to a historical or cultural diagnosis of this piece of false
consciousness. It seems to me that there is always a temptation to absorb
ethics into the most convenient example of respectable inquiry. In the heyday
of theology, the will of God determined ethics. When 'logic' held sway, phi-
losophers conceived of themselves as investigating the logic of moral dis-
course, and in those who wished to promote substantive views (Hare is para-
mount here), the goal became one of showing how those views were really
consequences of that logic. When logic loses ground and naturalism is in the
ascendant, the model is that of the scientist: just as the analytic chemist breaks
open the hidden nature of water, so the philosopher as scientist breaks open
the hidden nature of the good. His equations—justice is the difference princi-
ple, goodness is happiness, or whatever—then derive authority from their
similarity to the respected scientific activities on which they are modelled. But
any authority so derived is hollow, of course, and the models serve merely to
disguise the fact that in advancing such equations we are just moralizing, no
more and no less. We may be doing it well or poorly, of course, but the activity
itself, rather than neighbouring paradigms, is the judge of that.
By all means we may urge and argue that all and only societies that embed
10. Those interested can follow the dialectic through a rejoinder by Professor Sturgeon, and
a reply by me, in Philosophical Studies, 1990.
 Just Causes 209

the difference principle are just, or that all and only actions that promote
happiness are good. But we should remember that only a loss in understand-
ing what we are doing comes from thinking of ourselves as thereby identify-
ing, in a scientific spirit, the property that is justice or goodness, as opposed to
describing, from a moral perspective, the properties of just and good institu-
tions or actions.
This page intentionally left blank
 Ill
MIND AND MATTER
This page intentionally left blank
 12
The Individual Strikes Back

In this essay I address some of the points Saul Kripke makes in his treatment
of the 'rule-following considerations' in the later Wittgenstein ('LW').1 There
are two different quarries to track down. There is the question of whether
Kripke's exegesis of Wittgenstein ('KW') is correct—whether KW is LW.
And there is the distinct question of the real significance of the consider-
ations, as they are put forward by KW. Kripke himself is carefully agnostic
about this second issue.2 KW is not Kripke in propria persona. And Kripke is
also careful to distinguish the exegetical issue from the question of signifi-
cance. The two issues only connect like this: If KW's arguments have some
property that we are convinced cannot belong to any argument that LW would
have used, we shall suppose that KW is not the real LW. And admirers of
Wittgenstein will suppose that significance is such a property: if KW's consid-
erations are faulty, then for that reason alone KW cannot be LW; LW would
not have used faulty considerations. But without commenting on this opti-
mism, I want to discuss two other properties that might distinguish KW from
LW: KW's use of scepticism and his attitude to facts. Each of these aspects
may legitimately raise worries about his identity with LW. And I shall also
offer some thoughts about the significance of KW's arguments for our concep-
tions of meaning.
Our topic is the fact that terms of a language are governed by rules that
determine what constitutes correct and incorrect application of them. I intend
no particular theoretical implications by talking of rules here. The topic is that
there is such a thing as the correct and incorrect application of a term, and to
say that there is such a thing is no more than to say that there is truth and
falsity. I shall talk indifferently of there being correctness and incorrectness,
of words being rule-governed, and of their obeying principles of application.
Whatever this is, it is the fact that distinguishes the production of terms from
mere noise, and turns utterance into assertion—into the making of judge-
ment. It is not seriously open to a philosopher to deny that, in this minimal

1. Saul Kripke, Wittgenstein on Rules and Private Language (Oxford: Blackwell, 1982). All
page references are to this work. Paragraph references in the text are to the Philosophical
Investigations.
2. Page 5.

213
214 Mind and Matter

sense, there is such a thing as correctness and incorrectness.3 The entire

question is the conception of it that we have.
KW pursues this issue by advancing a certain kind of scepticism. As
Kripke is well aware, this might provoke immediate protest. Surely LW is
consistently scornful of scepticism? But this reaction misunderstands the
function of scepticism in a context such as this. The function is not to
promote a conclusion about knowledge or certainty, but to force a reconsid-
eration of the metaphysics of the issue. That is, we begin with a common-
sense or unrefined conception of some kind of fact. We think we have an
understanding of what that kind of fact consists in. The sceptic tries to show
that on that conception we would have no way of distinguishing occasions
when the fact obtains from those when it does not. Now, we might conclude
from this that our conception was correct and that there is therefore a
definite kind of proposition whose truth value we can never reliably judge.
This would be a traditional sceptical conclusion. But we might alternatively
conclude that since we do know the truth about the kind of thing in ques-
tion, the conception was at fault. The things we know do not have the kind
of truth condition we took them to have; the facts are not quite the kind of
thing we took them to be. This is a metaphysical conclusion, and the
sceptical dialogue is an instrument for reaching it. LW may have had no time
for sceptical conclusions. But he may well have had thought processes that
can be revealed by using a sceptical instrument to reach a metaphysical
conclusion. This is what KW does. So, as far as the issue of using a sceptical
weapon goes, KW may well be LW.
It is clear that on this account there are two parts to the business; attacking
the old conception and producing a replacement. The negative part might be
successful, although the positive part is not. If LW was successful, then in the
positive part he produced a conception on which public rule-following (in
some sense of 'public') is possible, whereas private rule-following is not. I
shall be arguing that KW is not successful in doing this. He does not succeed
in describing what it is for there to be a rule in force, with the property that
this can obtain in a public case but not in the private case.

I. Sceptical Solutions
Kripke describes KW as adopting a 'sceptical solution' to the sceptical consid-
erations, modelled upon Hume's sceptical solution to his own doubts. To
assess this idea it is important to separate various strands in Hume's extremely
complex position. If KW is understood to be taking over the wrong parts of
Hume, he may too easily be rejected as a pretender. I suspect that parts of
Kripke's presentation will encourage this—particularly those where he talks
of the sceptical solution.
3. Crispin Wright comes close to denying it in Wittgenstein on the Foundations of Mathemat-
ics (London: Duckworth, 1980), e.g., pp. 21-22, where he seems to attribute to Wittgenstein an
error theory of determinacy of the correctness of saying anything.
 The Individual Strikes Back 215

Hume calls section V of the First Enquiry the 'Sceptical Solution of these
Doubts'. The doubts in question were introduced in Section IV; they concern
operations of the understanding. In particular they concern our ability to
reason a priori about what must cause what, and the impossibility of justifying
inductive reasoning. The sceptical solution of section V consists in denying
that processes of reasoning have the power and the place hitherto assigned to
them. They are replaced by processes of custom. This is why we have a
sceptical solution: Hume offers a view of ourselves which in part he shares
with traditional sceptics. The shared part is the denial that we can justifiably
reason to our beliefs. (Hume differs from tradition in his estimate of the
consequences of that.) When Kripke introduces the analogy with Hume, this
is what he first mentions. But this is not the aspect of Hume (nor the section of
the Enquiry) that actually matters. What matters is his reinterpretation of the
concept of causation—the topic of section VII of the Enquiry.4 It is here that
Hume has a (fairly) pure example of the process I described: a sceptical
argument forcing us to revise our conception of a kind of fact. It is here that he
parallels KW. But the reinterpretation does not deserve to be called a
'sceptical solution' to anything, nor did Hume so call it. It is at most a proposal
prompted by sceptical problems. But in principle it might have been prompted
by other considerations altogether. And in fact Hume's reinterpretation of
causation is motivated only partly by scepticism. It is at least as firmly seated
in the theory of understanding: problems with our Idea of the causal nexus. In
this part of Hume, scepticism is subsidiary, even as a tool.
Kripke can rejoin that Hume's reinterpretation of causation is, in itself,
deservedly called sceptical for a further reason. It denies that there is a 'fact'
of whether there is a causal connexion between two events. At least, that is
how Kripke takes Hume.5 But here too, there are subtleties in the offing, and
they matter to the parallel with KW. This is because LW's attitude about
'facts' is going to be crucial, and crucial to many philosophers' belief that KW
differs from him. The philosopher we have described ends up reinterpreting
some kind of fact. This leaves various options. The first might be called
'lowering the truth-condition'. This asserts that sentences in the area can only,
legitimately, be given such-and-such a truth-condition. This can be combined
with the view that in our ordinary thought we confusedly attempt to do more,
or misunderstand what we are actually doing enough to make mistakes (al-
though there is always a problem about how we can attempt to do more, if we
are supposed to have no conception of anything more to do). Lowering the
truth-condition is then a reforming view, and entails an 'error' theory of
ordinary thought. But it can be combined with the view that the lowering
really reveals what we meant all along; this gives us reductionism. Often it
does not matter very much which combination is offered, and indeed, since a
decision depends on a fine detection of ordinary meaning, we would expect
some degree of indeterminancy.

4. Similarly in the Treatise, esp. Book I, part 3, section 14.

5. Pages 68, 69.
216 Mind and Matter

Quite distinct from lowering the truth-condition, there is the option of

denying one altogether. The sentences in question are given some other role
than that of asserting that some fact obtains. This option is familiar from
expressive theories of moral commitment or from views that try to see arith-
metical theses as rules rather than descriptions or propositions, and so on.
Now I say that this option is distinct from lowering the truth-condition, and
indeed in its initial stages it certainly is. It is a confusion, for instance, to
muddle together expressive theories of ethics with naive subjective theories
that give moral commitments a truth-condition but make it into one about the
speaker. Arguments against this latter view often have no force against the
former. But Hume is responsible for the very complication that makes it so
hard to keep these options properly separated. This is the view that the mind
spreads itself on the world: the view I call projectivism. According to proj-
ectivism we speak and think 'as if the world contained a certain kind of fact,
whereas the true explanation of what we are doing is that we have certain
reactions, habits or sentiments, which we voice and discuss by such talk.
Hume was quite clearly a projectivist in moral philosophy, and it is plausible
to see his metaphysics of causation as in essence identical.6
Like the option of lowering the truth-condition, the option of denying that
there is one can be combined with either of two attitudes to our ordinary
practice. One would be that it embodies error. Ordinary talk is conducted as if
there were facts, when there are no such facts. The talk is 'fraudulent' or
'diseased'. The other option is less familiar, but much more attractive. It holds
that there is nothing illegitimate in our ordinary practice and thought. The
respects in which we talk as if there are, for instance, moral facts, are legiti-
mate. (I have called this view 'quasi-realism'.) If LW were a projectivist, he
would have to be this kind, for it preserves the doctrine that ordinary talk and
thought, before we start to philosophize about the nature of our concepts, are
in perfect order as they are.
Kripke acknowledges the strand in LW that matters here. He also realizes
that it goes far—far enough to stop LW from endorsing any such judgement as
this: 'there is no fact of the matter whether a term is rule-governed or not'.
But KW expresses himself differently: KW believes that there is 'no such fact,
no such condition in either the "internal" or the "external" world' (p. 69).
This talk will outrage friends of LW. KW denies the existence of a certain kind
of fact altogether; LW would never so express himself, ergo one is not the
other. But now recall that Hume himself says that 'those who deny the reality
of moral distinctions may be reckoned amongst the most disingenuous of
disputants'.7 Why? Because insisting on this reality is part of normal thinking.
It is part of the way of life, or way of thought or talk, that quasi-realism can
protect for us. So there must be room for a different version of LW. This one
would abandon his hostility to facts. He should accept that talking of facts is

6. For typical statements compare the appendix to the Enquiry, and Treatise, p. 167.
7. Enquiry Concerning the Principles of Morals, section I.
 The Individual Strikes Back 217

part of our legitimate way of expressing ourselves on the difference between

terms that are rule-governed and terms that are not. He has to do this if he
doubles for LW. For recall that the very passage (Investigations §137) that
begins these considerations is a version of the redundancy theory of truth.
And on that theory (whatever else it holds), there is no difference between
saying that it is true that p and that p or between saying that it is a fact that p
and that p. So anybody prepared to assert that terms are rule-governed—and
as I explained initially, that must mean all of us—can equally be heard saying
that there are facts, truths, states of affairs of just that kind.
Kripke says something strange about this. On p. 86 he imagines someone
saying that it cannot be tolerable to concede to the sceptic that there are no
truth-conditions or corresponding facts to make a statement about someone's
meaning true: 'Can we not with propriety precede such assertions with "It is a
fact that" or "It is not a fact that"?' But Kripke puts this complaint in the
mouth of an objector to LW. He then reminds us that LW accepts the redun-
dancy theory of truth, and says that this gives him a short way with such
objections. The dialectic seems to be the wrong way round. Because he ac-
cepts the redundancy theory, LW can assent to the locutions of fact and truth.
They add nothing to the fact that we judge other people not to be following
rules and applying terms in accordance with standards of correctness.
I don't know that the redundancy theory of truth should have this soothing
power. It may be that there are explanations of why we make some judge-
ments, and why we apply the calculus of truth-functions to the judgements we
make, that still leave us queasy when we consider whether there are facts in
the case. The quasi-realist construction of ordinary moral discourse is a good
example. Even if it protects everything we do, it can still leave us uneasy when
we contemplate the question: 'Yes, but are there any moral facts on this
account?' I shall not consider that question further in this paper. I just want to
note that on a redundancy theory, it is a bad place to become puzzled, and
that LW held a redundancy theory. Hence it seems that there is space for a
persona who profits from KW's arguments but draws a rather different conclu-
sion from them. With due humility, I shall call this character BW. He is going
to share a great deal of the argumentative strategy of KW. In part his differ-
ence is relatively cosmetic: the belief that scepticism is only in play, if at all, as
an instrument, and that the eventual conception of rule-following that must
emerge does not deserve to be called sceptical. But in part he differs more
substantially, because he hopes to preserve the implications of the redundancy
theory in LW, and he hopes to cement a more particular relationship between
LW and the real Hume, who matters to metaphysics. If he can preserve our
right to talk of the fact that words obey principles of application, this may be
as much because he forces us to revise our conception of a fact as anything
else. Alas, however, BW has one radical flaw. On his philosophy, there is no
particular reason to discriminate against the would-be private linguist. But
before advertising his intended end point further, it is necessary to review
some of the arguments by which he gets there.
218 Mind and Matter

II. The Critique of Rule-Following

It may seem outrageous to the touchy friends of LW that anything in Hume
could be a model for LW's attitude to rule-following. But perhaps we can stem
some of the outrage by reminding ourselves that it is an essentially normative
judgement that we are chasing. It is the judgement that something is correct
or incorrect. When this fact proves fugitive, as KW shows that it is, its norma-
tive nature is largely the problem. So it cannot be that outrageous to apply our
best explanations of value judgements to it, and this is what BW hopes to do.
But why is the fact fugitive in the first place?
In Kripke's development, we start by considering the understanding I had
of some term at some time in the past, which we can arbitrarily call 'yesterday'.
Suppose the term is some arithmetical functor, 'plus' or' + '. I understood it by
grasping a principle of application. We then consider my position today. When I
come to do a calculation, which we suppose I have never done before, I cer-
tainly believe myself to follow a principle. I believe myself to be faithful to
yesterday's rule, by adopting the same procedure or principle in determining
answers to problems expressed using the functor. Thus I believe that if I am
faithful to yesterday's principle, I should say '57 + 68 is 125'. Notice that this is
not quite the same as claiming that I should say this unconditionally. I may wish
to change my allegiances. There's no impropriety in deciding that yesterday's
principle of application is not the best one for this term, and to consciously start
to use it according to different rules. It is just that if I am faithful, then I ought to
give that answer. I most certainly should not say that 57 + 68 is 5. Nor should I
say that there is more than one answer to that problem, or that the problem is
indeterminate, so that there is no answer at all. The sceptical dialogue then
commences. The sceptic asks me to point to the fact that I am being faithful to
yesterday's rule by saying only one thing, and not these others. And this proves
hard to do. For any fact that I tell him about myself seems compatible with the
'bent-rule' hypothesis, as I shall call it: the hypothesis that the rule that was
really in force yesterday was a Goodman competitor. In other words, the kinds
of fact I am apt to allege are compatible with a story in which I really understand
by 'plus' a function with a particular singularity at x = 57 and y = 68. For
example, I might yesterday have meant what we should now express by saying

x + y = the sum of x,y, except when x = 57 and y = 68 and 5 otherwise

This hypothesis is not refuted by my present staunch denial that I had any
such thought in mind yesterday, or that if asked I would have used words like
these to explain myself. For the sceptic will urge that as well as having had a
bent interpretation of ' + ' yesterday, I could have had a bent interpretation of
these other terms as well. So pointing out that I would have presented no such
explanation to myself does not refute the sceptic. It would really be a question
of my using another rule to interpret the first (e.g., the rule for interpreting
various synonyms for ' + ' or for interpreting terms that would have occurred in
 The Individual Strikes Back 219

any explanation of the functor that I would have proffered). So the fact that
yesterday I would have said, for instance, 'x + y is always a number greater
than either x or y, with x = 57 and y = 68 no exception' is consistent with my
having been a secret bent rule user. The singularity, by current lights, would
have been there in the way I took the words involved in such an affirmation.
Perhaps 'exception' meant . . .
This argument (which is much more forcefully and thoroughly presented
in Kripke) undoubtedly corresponds to a central negative point of LW's. The
point is that taking a term in a certain way is something different from present-
ing anything as an aid to understanding it, or from accepting anything as aids
to understanding it. He says in §201 that he has shown 'that there is a way of
grasping a rule which is not an interpretation but which is exhibited in what we
call "obeying the rule" and "going against it" '! The negative point, that we
gain no approach to the required fact by embarking on a potential regress of
interpretations, is quite clear. The more positive claim, that the fact is exhib-
ited in what we call obeying a rule, must wait.
When presenting the sceptical challenge this way, we should not lose sight
of the fact that the case can be made without instancing a rival principle of
application, a bent rule, at all. A sceptic might just doubt whether there was,
yesterday or today, any principle at all behind my application of ' + '. Perhaps
all that happened was that I would look at things, such as triples of numbers,
and after a process that was phenomenologically just like one of being guided
by a rule, declare 'z = x + y' I would be in the same case as a lunatic who
thinks he is doing sums, when all that is happening is that he is covering pages
with symbols. Or, I would be like the man whom Wittgenstein considers at
§237, who with great deliberation follows a line with one leg of a pair of
dividers, and lets the other leg trace a path, but one whose distance from the
original line he varies in an intent but apparently random way. He might think
he is tracing a path determined by a rule relating its course to the first line.
But his thought that this is what he is doing does not make it true. In some
ways this is the primary weapon that LW uses against the private linguist. He
forces him back to saying that he has only his own conviction that he is
following a rule at all, and this private, phenomenological conviction that one
is following a rule is not enough to make it true.
I have followed Kripke in concentrating upon the normative aspect of the
fact we are looking for. So I agree with him that the answer to the problem is
not going to be given just by talking of dispositions we actually have. How-
ever, and crucially for what follows subsequently, I do not think the disposi-
tional account falls to all of Kripke's objections. The analysis he considers (p.
26) says that I mean some function by my functor if and only if I am
disposed, whenever queried about the application of the functor to a pair of
numbers, to give the answer that actually is that function, of them. Kripke
attacks this on the grounds that my dispositions are finite. 'It is not true, for
instance, that if queried about the sum of any two numbers, no matter how
large, I will reply with their actual sum. For some numbers are simply too
large for my mind—or my brain—to grasp' (pp. 26-27). So, according to
220 Mind and Matter
Kripke, my dispositions fail to make it true that I mean addition by ' + ' and
not 'quaddition'—a function that gives different results just when x and y are
so big that I cannot do sums involving them.
There are difficulties here. It is not obvious that dispositions in themselves
are either finite or infinite. The brittleness of a glass is a respectable disposi-
tional property. But there is an infinite number of places and times and strik-
ings and surfaces on which it could be displayed. Does this glass have a disposi-
tion that covers, for example, the fact that it would break if banged on a rock
on Alpha Centauri? What if scientists tell us that this glass couldn't get there,
because it would have decayed within the time it takes to be transported there?
Perhaps I am not disposed to give the answer faced with huge sums. But
perhaps also I have dispositions that fix a sense for the expression 'the answer I
would accept'. The answer I would accept is the one that would be given by
reiterating procedures I am disposed to use, a number of times. (The notion is
doubly dispositional.) The fact that I am not disposed to follow those proce-
dures that number of times seems like the fact that the glass cannot get to
Alpha Centauri. Now, a sceptic might maintain that we do not know of a
dispositional fact about me that is described in this way. Perhaps I am only
disposed to say that '3 + 5 = 8' when the calculation is not embedded in really
huge calculations. But this is just scepticism about dispositions. It is like suppos-
ing that the glass may be not brittle but 'shmittle', where x is shmittle if it
breaks when struck except. ... In effect, this is inductive scepticism about the
concept of a disposition, querying whether we can legitimately take disposi-
tions to cover what would have happened on unobserved occasions. So it
cannot be used to argue that even if we accept the concept, it permits no answer
to the problem of huge calculations. A similar complication might answer
Kripke's second objection (pp. 28-29). This is that what I mean cannot be read
off from what I am disposed to do, since I may be disposed to make mistakes.
The dispositionalist would have to read off what I meant from a table of
answers I actually give, and this might involve saying that I was computing
(correctly) a bent function—'skaddition'—and not making a mistake in at-
tempting to add. But this seems to ignore surrounding dispositions. Kripke
rightly dismisses any view that simply takes for granted a notion of the function
it was intended to compute, or that defines user's competence, since it presup-
poses the ideas we are looking for. But at least it is true that a calculator can
have, in addition to dispositions to give answers, dispositions to withdraw them
and substitute others. And it is possible that putting the errant disposition into
a context of general dispositions of this sort supplies the criterion for which
function is meant. The equation would be: By ' + ' I mean that function that
accords with my extended dispositions. An answer z = (x,y) accords with my
extended dispositions if and only if (i) it is the answer I am disposed to give and
retain after investigation, or (ii) it is the answer I would accept if I repeated a
number of times procedures I am disposed to use, this being independent of
whether I am disposed to repeat those procedures that number of times.
Kripke's point about mistakes can be illustrated if we consider a calculat-
ing machine. There is no physical or dispositional difference between a ma-
 The Individual Strikes Back 221

chine that is 'supposed' to compute addition but because of a mistake in the

hardware computes a bent function, and one that is designed to compute the
bent function in the first place. The two may be perfectly identical. This
strongly suggests that the notion of correctness, the notion of the rule to which
we are to be faithful, has to come from outside the thing itself. In the case of
the calculator, it certainly does. But this need not entirely destroy a disposi-
tional account, provided it can look for dispositions outside the simple disposi-
tion to give answers.
So is it a real truth that the right rule for ' + ' is in force among us? My own
answer would be that we do have dispositions that enforce this judgement.
They make it the only possible judgement about ourselves, when we describe
each other's thoughts. The concealed bent-rule follower is a theoretician's
fiction. Whenever we try to fill out the story of a person or a community that
really adopts a bent rule, it turns out that the singularity in the rule (by our
lights, of course) must affect the dispositions to behave that the community or
individual shows. I have argued this elsewhere in connexion with Goodman's
paradox.8 The concealed bent-rule follower is often thought of as though noth-
ing about him is different until the occasion of bent application arises. But this is
wrong. Someone who has genuinely misunderstood a functor is different, and
the difference can be displayed quite apart from occasions of application. Con-
sider, for instance, the bricklayer told to add bricks to a stack two at a time. If
this means to him 'add 2 up to 1000, and then 4', his reaction to the foreman may
be quite different. Perhaps he cannot carry four bricks at a time.
However, I am not going to pursue this defence of dispositions. I share
Kripke's view that whatever dispositions we succeed in identifying could at
most give standards for selection of a function that we mean. They couldn't
provide us with an account of what it is to be faithful to a previous rule. It is
just that, unlike Kripke, I do not think dispositions are inadequate to the task
of providing standards. Indeed, I think they must be adequate. For notice that
the problem of finitude applies just as much to any community as to any
individual. If the finiteness of an individual's dispositions leaves it indetermi-
nate whether he means one function or another by some functor, then so must
the finiteness of community disposition. So although communities induct their
members into ' + ' using community practices and go through all the correc-
tions and imitations that constitute community use, still, all that is consistent
with the 'skaddition' hypothesis too. We don't find communities disposed to
calculate numbers that are just too big. But prosecuting this point takes us
into BW's divergence from KW.

III. The Community and the Individual

The individual has a hard time against Kripke's sceptic. How does the public
fare better? The individual couldn't make the sceptic appreciate the kind of
8. Reason and Prediction (Cambridge: Cambridge University Press, 1973), chapter 4, and
Spreading the Word (Oxford: Oxford University Press, 1984), chapter 3.
222 Mind and Matter

fact it was—that he was being faithful to a principle, or rule, or previous

intention, when he gave some answer. The sceptic charges that there is no fact
of the matter whether the bent rule or the natural rule was the one intended,
or whether one principle of application or another was in force yesterday. And
if there is no fact of the matter of this sort, then there is really no fact of the
matter that any principle at all was in force. Any answer to the new sum can
be regarded as equally 'right', and that means that we cannot talk about
'right'. Faced with this impasse, the individual thankfully turns to the commu-
nity. He can point to his inculcation in a public practice, his gradual confor-
mity to patterns of behaviour accepted by others, and his acceptance as a
competent operator with ' + '.
The sceptic might allow all this to make the difference. But he has sud-
denly gone very soft if he does. He can easily specify bent principles, with
points of singularity where neither I nor anybody else used the terms yester-
day. If such points worried the individual, then they should equally worry the
community. So how does mention of the community give us the determinate
rule?
KW's answer is that in a sense it doesn't (see, for instance, p. 111). The
community is as much at a loss to identify the fugitive fact as the individual
was. The position is supposed to be no different metaphysically. The differ-
ence is that the community endorses or accepts the competent operator. They
or 'we' allow him to be using ' + ' to mean addition. He is 'seen' or dignified as
a rule-follower. We, the community, have justification conditions for doing
this. To gain the title, the individual's practice (on the finite number of occa-
sions when it will have been shown) must accord in some suitable way with the
community practice (on the finite number of occasions when it will have been
shown).
Now, merely citing the fact that we 'see each other as' obeying the same
principle of application makes no headway at all. Remember that the point of
the original worries about rule-following was not epistemological. The aim
was not to suggest that we cannot provide a foundation of some sort for the
judgement that a rule is in force, or that we cannot provide principles for
inferring such judgements from other, more basic principles. If this were the
aim, then replying that there is no inference and no foundation, but just a
basic fact that we make such judgements of each other, 'seeing each other as'
following principles might be relevant—at any rate, as relevant as this move is
in other areas of episternology. But making this point does nothing to suggest
how we are seeing each other when we say such a thing. We know well enough
what it is to see something as a duck, because we know what ducks are. But
we don't know what it is to see someone as obeying a principle of application,
unless we know what it is to follow one, and so far we still have no conception
of this. To put it another way, we do not know what a community would be
lacking if its members failed to see each other this way, or if they continually
saw each other in the light of potential bent-rule followers.
So the sceptic is still liable to feel short-changed. He has pointed out bent
rules that might have been in force yesterday, compatibly with all that the
 The Individual Strikes Back 223

individual could point to. He has pointed out that for all that the individual
could show, there might be nothing but his onward illusion that rules are in
force and that his dispositions to respond are correct or incorrect. The individ-
ual now links arms with others. The sceptic attempts to point out the same two
things to them: for all they can describe about themselves, bent rules (and
perhaps differently bent rules) might have been in force, underlying their
fortuitous coincidence of behavior over the finite samples they have come
across. Since this means that any answer to a new problem might be as correct
as any other, the sceptic suggests that again there is just the onward illusion
that there is correctness or incorrectness. The community replies that it has a
practice of dignifying its members as saying things correctly or incorrectly, and
in the light of this practice it says that all its members do mean the same, and
that what they mean provides a principle of application of a term.
We can see one way in which this could silence the sceptic. If the mutual
support itself provides the standard of correctness, then a community can
answer him. To understand this, consider the analogy with an orchestra. Sup-
pose that there is no such thing as an individual playing in accord with instruc-
tions coming down from the past, by way of scores or memory. Suppose too that
on an individual instrument there is no standard for the way a piece ought to go
(all melodies are equally acceptable). Then an individual cannot play well or
badly in isolation. Nevertheless, the orchestra may have standards of harmony
across instruments. And at a given time most instruments are playing, say,
notes from the chord of C major, then the individual who hits a dissonant C# is
incorrect by the standards of the orchestra. They can turn on him. Unfortu-
nately, this provides only a poor analogy for communities and their relations to
their own rule-following. For in the orchestra, harmony with others provides a
direct standard of correctness. This is not so with judgement. My community
may all suddenly start saying that 57 + 68 = 5, but this does not make me wrong
when I continue to assert that it is 125.1 am correct today in saying that the sun
is shining and daffodils are yellow, regardless of what the rest of the world says.
Obviously, any solution to these problems must avoid the disastrous conclusion
that it is part of the truth-condition of any judgement that a community would
make it (unless of course the judgement is itself not about the sun and daffodils
and so on, but about the community).
If the community cannot turn to the orchestral metaphor, then how have
they answered the sceptic? And why cannot their answer be taken over by the
isolated individual? Remember that there is a distinction between the overall
practice of a community, thought of as something defined by principles and
rules, and the exposed practice, thought of as only a partial, finite segment of
applications. If a community practices addition, meaning one thing and not
another by some functor ' + ', then the exposed practice will cover only a small
proportion of the applications whose correctness follows from the overall
practice. As the possibility of bent functions is supposed to show, the exposed
practice does not logically determine the question of which overall practice is
in force. Then we can imagine what we might call a 'thoroughly Good-
manned community' in which people take explanations and exposure to small
224 Mind and Matter

samples—yesterday's applications—in different ways. The sceptic who won

against Kripke's individual will now win against a community, showing that
they have no conception of the fact that makes it true that they do not form a
thoroughly Goodmanned community.
It may be helpful to think of it like this: The members of a community
stand to each other as the momentary time-slices of an individual do. So just
as the original sceptic queries what it is for one person-time to be faithful to a
rule adopted by a previous person-time, so the public sceptic queries what it is
for one person to be faithful to the same rule as that adopted by another. Now
if the public sceptic can be bypassed by, in effect, saying that this is what we
do—we see each other as mutually understanding the same rule, or dignify or
compliment each other as so doing, provided the exposed practice agrees well
enough—then the private sceptic can be bypassed in the same way. His doubts
admit of the same projective solution. When LW denies that 'we have a model
of this superlative fact' (§192), we can, as far as the metaphysics goes, shrug
and say that this is how we see ourselves. Then, as I have already explained,
the entire problem is to explain how we are seeing ourselves when we go into
this mode. In particular we need to cite some standard for saying this—and, if
LW is to win through, a standard that separates the public from the individual.
For when I write the sensation term in my diary I can and will see myself as
being faithful to a previous intention to apply it only in a determinate range of
circumstances. And, paying me the compliment (as it were), you can do so
too. It might be different if it somehow were natural for us to dignify fellow
members of the same public with the title of rule followers, and natural to
hesitate over any purely private attempt at self-description. But generally
speaking, there is little difference in naturalness. Until LW supposedly argues
that we shouldn't do so, most people would find it quite natural to believe that
a putative private linguist might be following or failing to follow previously
formed intentions to apply principles of application.
The dilemma so far is this. If the presence or potential presence of a
community of persons practising the same way enters as part of the truth-
conditions, part of the analysis, of what it is to follow a rule, the sceptic who
won against the private individual looks equally set to win against a commu-
nity that has the benefit of mutual support. But if mention of the community
comes as part of a Humean or projective solution, allowing us to bypass
troubles over our conception of the superlative fact, then a similar side-step is
in principle available to the individual. To split the individual from the public
we need enough understanding of how we are regarding ourselves to be able
to specify standards that need to be met by a candidate for rule-following, and
we have not yet got this.
The simple move, endorsed by KW and, for instance, by Peacocke, is to
say that in the case of the individual there is no distinction between it seeming
to him that a rule is in force, and there being a rule in force. The trouble with
this is that in any sense in which it is true (see next section), it is equally true

9. Enquiry, section I.
 The Individual Strikes Back 225

that there is no distinction between a community being thoroughly Good-

manned but seeming to itself to have a unified practice, and its actually having
a unified practice. And when the community says 'well, we just see ourselves
as agreeing (dignify or compliment ourselves as comprehending the same
rule)', the individual just borrows the trick, and compliments himself on his
rapport with his previous times. The community needs to say that the compli-
ment is empty in the case of one individual but meaningful in the case of a set
of them. How is this to be argued?

IV. Public and Private Practice

Peacocke writes:
In the end, Wittgenstein holds, the only thing that must be true of someone
who is trying to follow a rule, so long as we consider just the individual and
not facts about some community, is that he is disposed to think that certain
cases fall under the rule and others do not. But this is something which is
also true of a person who falsely believes that he is conforming to a rule. His
general argument is that only by appealing to the fact that the genuine rule-
follower agrees in his reactions to examples with the members of some
community can we say what distinguishes him from someone who falsely
thinks he is following a rule.10
The lynch-pin of this interpretation is the absence of a 'distinction', in the case
of an individual, between the case where he believes that there is a rule when
there is none and the case where he believes it and there is one. And my
question has been: do we think we have yet been given a 'distinction' between
a thoroughly Goodmanned community accidentally agreeing in exposed prac-
tice and a real community of understanding? And then, if the distinction is
given using materials from Hume—the projection upon one another of a
dignity—I ask what the standards are whereby it takes several people to do
this, when one cannot do it to himself. I shall now try to exhibit the force of
this question by considering the individual further. This is to illustrate the gap
between anything that has happened in LW, at least as far as §202, and the
application to exclude the private linguist.
Let us recall the basic point that giving an explanation, either in words or
using other aids, such as pictures or models, does not logically determine the
rule that governs one's understanding of a term: 'Any interpretation still
hangs in the air along with what it interprets, and cannot give it any support'
(§198). To avoid the threatening paradox that nothing can accord nor conflict
with a rule, because anything can be made out so to do, LW introduces the
positive suggestion that our rules are anchored in practice. In the remainder
of this paper I shall assume that this is along the right lines. That is, dignifying
each other as rule-following is essentially connected with seeing each other as

10. C. Peacocke, 'Rule Following: The Nature of Wittgenstein's Arguments', in S. Holtzman

and S. Leich, eds., Wittgenstein: To Follow a Rule (London: Routledge & Kegan Paul, 1981), p. 73.
226 Mind and Matter

successfully using techniques or practices. This at least begins to isolate the

nature of the judgement. It suggests a direction from which to find standards
for making the judgement, and in my view it connects interestingly with
pragmatic and coherence aspects of judging. Now, having introduced the
notion of a practice or technique. LW immediately goes on to draw two
famous conclusions: that to think one is obeying a rule is not to do so, and that
it is not possible to obey a rule privately. These are not the same, and the
second is not warranted. The question is, what kind of thing does a practice
have to be if it is to block the sceptical paradox? Perhaps, for instance, the
concept of a practice can include that of someone setting out to describe his
mental life, even on a highly private conception of the mental. The basic
negative point certainly does not rule that out. To do so we would need further
thoughts about what a practice must be, and the connexion it is supposed to
need with actual or possible publics. It is therefore a great pity that, with one
eye cocked on the later applications, commentators simply assume that by
§202 the publicity of practice has been satisfactorily argued (Peacocke, for
instance, just announces that 'by "practice" he here means the practice of a
community' [p. 72]). Of course, many believe that the later discussion of
sensation S and so on justifies the restriction. But that is another issue; to
invoke it is to abandon the hope that the rule-following considerations pro-
vide wider, more general thoughts from which the anti-private language con-
clusion independently emerges. To illustrate the gap, we can consider the
halfway house (to full privacy) provided by Robinson Crusoe cases.
Consider the example (due to Michael Dummett) of a bom Crusoe who
finds a Rubik's cube washed onto his island, and learns to solve it. The fact is
that he does it. He certainly doesn't solve it randomly, for he can do it on
demand. It is natural to say that he follows principles ('when there is a last
corner left, do x . . .'). Perhaps he has some rudimentary diagrams or other
mnemonics that he consults. With these he can do it, and without them he
cannot. Kripke considers what LW's attitude should be to such a case. The
attitude he offers, on behalf of KW, is that we can think of Crusoe, in such a
case as this, as following rules. But, 'If we think of Crusoe as following rules,
we are taking him into our community and applying our criteria for rule-
following to him'. The sceptical considerations are supposed to show 'not that
a physically isolated individual cannot be said to follow rules; rather than an
individual considered in isolation (whether or not he is physically isolated)
cannot be said to do so'.11 Again there is a nice parallel (although not quite the
one Kripke makes) with Hume. One conclusion that might be drawn from
Hume is not that a pair of events in a universe with no others cannot be
related causally (which is how Kripke takes Hume); rather, if we suppose that
they are, we are thinking of them as members of a (potential) family of other
regularly related events. We are not considering them in isolation, but, for
instance, arc thinking of what would happen if there were others like them.
Still, it is not clear what the compromise is. Certainly I, or we, are doing

11. Kripke says useful things about this, pp. 90 ff.

 The Individual Strikes Back 227

the thought-experiment. I have to consider whether Crusoe is a rule-follower

by using the normal, community-wide way I make the judgement. But that
would be true of any situation I seek to describe. And then, just as (contra
Berkeley) I might conclude that an island considered in isolation has a tree on
it, might I not conclude that Crusoe, considered in isolation, was following a
rule? How does the phrase 'considered in isolation' bring the community
further into the picture in this case than in the case of the tree? We are apt to
retort that Crusoe would have been a rule-follower in this situation whatever I
or we or any other community in the world had thought about it—just like the
tree. And the reason is that all by himself he had a technique or practice.
Now reconsider the private linguist, meaning someone who believes that
he has given an inner state a semantically essential role. Suppose he believes
that yesterday, in the presence of the state, he defined for himself a qualitative
similarity that other states might bear to it; that he is on the lookout for
recurrence and involved in the practice of judging whether such a recurrence
takes place. We seek to show that this is not a real practice. Let our private
linguist accept the basic point, admitting that the mere offering of words,
images, and so on does not determine a rule of application, or principle, that
is really in force. But he does not accept that his candidate practice is unsatis-
factory. What has LW got to show him?
'Whatever is going to seem right is right'—there is no distinction between
his seeming to himself to follow a rule and his genuinely doing so. It has often
been suggested that this charge is unargued, or, if argued, only supported by
overtly verificationist considerations. My endeavour has been to show how
difficult it is to release the public from its attack. In the light of this, let us
reconsider the projective and naturalistic elements that assisted the public:
the point that we naturally and perhaps usefully regard ourselves as mutual
possessors of the same understanding. We see ourselves so, and this attitude
is, in Humean vein, immune from sceptical destruction. Furthermore, there is
no lowering of the truth-condition of this judgement. It sits with its own
vocabulary and scorns any 'account' of it in other terms. It is just that if a
public failed to see itself in this light, it would mean that it could only see the
ongoing patterns of noise and reaction, in which no principle, no genuine
judgement, no truth and falsity, is visible.
My criticism of the flat reply to the sceptical problem ('this is how we see
ourselves') was, in effect, that it gave no account of what a community would
be missing if it failed to see itself in this light. Following LW we have accepted
that the clue to what is missing is to come from the notion of a practice. BW
therefore uses this thought and the fact that dispositions provide satisfactory
standards for the making of the judgement. As with other judgements with
normative elements, there is no attempt to make a lowering of the truth-
condition. The judgement can be seen perfectly well in terms of the projection
of an attitude (this cannot be uncongenial to LW). But that left us unclear
about the standards for applying it. BW therefore accepts most of what
emerges from LW, doing less violence to the redundancy theory and to the
preparedness to talk of facts, even when the underlying metaphysic is not what
228 Mind and Matter

that might suggest. But however well BW matches LW, Blackburn insists on
asking: is there any reason why the private linguist should not so regard him-
self? And in that case whose is the attitude 'whatever is going to seem right is
right'? Not the subject's own, for he dignifies himself as a genuine believer, as
having a principle of application and making a judgement with it. In doing so
he allows the possibility of mistake (it is not something there in the things going
on in his head or in his behaviour; it is something arising as a projection from an
attitude he takes up to his own projects). It is a component of his attitude that a
particular judgement might turn out better regarded as mistaken.
How can this attitude be appropriate? A technique is something that can
be followed well or badly; a practice is something in which success matters.
Now in the usual scenario, the correctness or incorrectness of the private
linguist's classification is given no consequence at all. It has no use. He writes
in his diary, and, so far as we are told, forgets it. So when LW imagines a use
made of the report (e.g., to indicate the rise of the manometer) he immedi-
ately hypothesizes a public use. He thereby skips the intermediate case where
the classification is given a putative private use. It fits into a project—a
practice or technique—of ordering the expectation of recurrence of sensation,
with an aim at prediction, explanation, systematization, or simple maximizing
of desirable sensation. To someone engaged on this project, the attitude that
whatever seems right is right is ludicrous. System soon enforces recognition of
fallibility.
I conclude then that it is no mistake to see the latter sections, from §240
onward, as integral to the anti-private language polemic. BW simply cannot
separate the private from the public with any considerations that are in force
earlier on. But I have tried to suggest other things as well. Following through
the problem of answering the paradox leads to sympathy with a basically 'anti-
metaphysical' conception of rule-following. We simply cannot deliver, in other
terms, accounts of what constitutes shared following of a rule, or what the fact
of a rule being in force 'consists in'. In my view this invites a projectivist
explanation of these kinds of judgement, although also in my view we cannot
conclude that it is improper to talk of 'facts of the case'. In any event, we are
left searching for standards whereby to make the judgement. It is possible that
those standards should exactly separate the public from the private (on some
vulnerable conception of the private, of course). But there is no particular
reason to expect them to do so. The problems with dispositions, either as
giving us the missing kind of fact, or as providing standards for allowing that a
rule is in force, failed to separate the public and the individual. So we cannot
now simply demand from the putative private-linguist an 'account of what the
distinction [between genuinely following a rule and only seeming to himself to
do so] amounts to'. He can reply, 'It amounts to all that it does in the case of
the public'. Just as a public dignifies itself as producing more than an intermi-
nable flood of words and noise, and sees itself as making mutually comprehen-
sible judgements, capable of truth and falsity, so does he. The public doubt-
less has a purpose in doing this, and is right to do it. When his putative
discriminations are part of a practice, so does he.
 13
Losing Your Mind: Physics, Identity, and
Folk Burglar Prevention*

In this essay I introduce a paradox in much current thinking about physics.

The paradox is that physics (and indeed any human thought) never identifies
causes—real causes. It may give us causal explanations, but these are only
ways of pointing toward the real causes, or features in virtue of which effects
follow causes. The paradoxical conclusion is that the predicates we use do not
express or refer to the causally powerful properties or states. We can head
toward the engine room, perhaps, but never get there. I diagnose current
eliminativism in the philosophy of mind as largely the upshot of the paradoxi-
cal way of thinking. If I am right, its proponents have done a great service,
giving us an outward and visible sign of an inner spiritual tangle. I start by
saying a little about eliminativism, by way of introducing the way of thinking
about physics that, according to me, it depends upon, and thence identifying
the paradox.
The prophets of eliminativism may follow one of two models in their
attitude to the categories whereby we understand ourselves and each other
in terms of what we feel, think, believe, intend, desire. One model is the
prophet Isaiah ('from the sole of the foot even unto the head there is no
soundness in it'), and the other is John the Baptist ('there cometh one
mightier than I after me'). The frequent comparisons of commonsense psy-
chology to caloric theory, phlogiston theory, or demonology suggest the first,
whereas the mere belief that in the future something better may turn up, by
some standards of accurate prediction, completeness, simplicity, or capacity
to deal with borderline cases, gives only the second. It is only on the first
model that there is something false in commonsense psychology. The second
attitude no more suggests that commonsense categories falsify things than
belief in Newtonian mechanics falsifies the folk view that it is harder to lift a
heavy stone than a light one, or belief in hydrodynamics falsifies the view
that there are rivers.
In this essay I oppose only the stronger doctrine. What is wrong with
*This essay owes an unusual amount to discussion and correspondence. I particularly want to
thank Ned Block, Mark Rowlands, Philip Pettit, and Marianne Talbot, as well as Jay Rosenberg
and the other symposiasts at the Greensboro Colloquium in 1988. An especial debt is due to
David Lewis, whose patient incredulity enabled me to avoid many errors.

229
230 Mind and Matter

strong eliminativism is not its saying that a particular mode of description may
one day be replaced. I have no quarrel (how could one quarrel?) with the bare
possibility of future modes of understanding that somehow improve on those
of folk psychology. Equally I have no quarrel with the bare possibility of
future modes of understanding that similarly improve on those of folk geogra-
phy or furniture classification. Again, many of the arguments against common-
sense categories detail particular hard cases of belief ascription and cases
where attribution of content is underdetermined. 1 I shall not be concerned
with these; there is, after all, no general doctrine that denies a term any
application because it is polycriterial or gives rise to difficult or borderline
cases. Finally, there is the issue of norms of rationality versus the natural—the
issues concerning correctness that burgeon into the rule-following consider-
ations; but I shall not be centrally concerned with these either. Instead I shall
argue that what is wrong with current eliminativism is its claim to see how
commonsense psychological description conflicts with science.

I. The Theory-Theory
To engage with eliminativism, I shall make one concession at the outset. This
is that 'commonsense psychology' should be identified as being, in some
suitable sense, a 'theory'. For the purposes of this paper I shall accept that the
terms of commonsense psychology get their meaning from an implicit func-
tional definition, or a network of sayings, platitudes even, connecting beliefs,
desires, and the rest with typical causes, typical interrelations, and typical
effects. Saying that commonsense psychology is a theory, in this weak sense, is
not inconsistent with allowing that it is used to do many things, nor with
allowing that we often know by observation, or direct access to our own
states, what we ourselves and others believe, desire, and so on.2 If we are to
reject the view that it forms a theory, this ought, I believe, to be on the
grounds that it suggests an unsustainable asymmetry between my own case
(observational, direct knowledge) and the third-person case (theoretical, indi-
rect knowledge), but in this paper I shall not pursue that problem.
For eliminativism, the term 'theory' in fact functions as a staging post:
because of some considerations, commonsense psychology is well regarded as
a theory, but theories are corrigible, hence it is corrigible, and the question of
its replacement may be raised. The view that terms are introduced by implicit
functional definition, which I shall call the role or network model of them,
does not itself immediately carry implications of corrigibility. Its doing so
depends on the corrigibility of the elements of the network—the generaliza-
tions that keep commonsense psychological terms in place and enable them to
1. Stephen Stich, From Folk Psychology to Cognitive Science (Cambridge, Mass.: Bradford
Books, 1983), especially chapters 4 and 5.
2. This point is well made by Patricia Churehland, 'Replies to Comments', in Inquiry, 1986,
p. 254. But I am now much more sceptical of the virtues of the 'theory' idiom: see my 'Theory,
Observation and Drama', in Mind and Language, 1992.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 231

be used in explanation and prediction. It is not obvious how corrigible these

are. A sample list includes such things as
• Barring self-control, anger causes impatience and antisocial behaviour.
• Barring a stronger contrary purpose, hunger causes eating.
• Barring pathological personality or circumstance, the apprehended loss
of loved ones causes grief.3
• People generally do what they say they will do.
• Intending P & Q is normally sufficient for intending P and intending Q
(from Fodor).4
Suppose we agree that such generalizations underlie commonsense prac-
tice, in the sense that if they are denied, commonsense psychology will have
no explanatory and predictive use, and no terms with any meaning. Still the
examples do little to show us how commonsense theory is corrigible. For we
may have very little idea indeed how scientific progress, or any other, could
tempt us to abandon any of them, let alone sufficient of them to count as
abandoning the overall commonsense perspective (nobody is concentrating
on small adjustments in this debate).
The real danger in using the word 'theory' is supposing that it does this
work for us. Stich and Churchland frequently remind us of other 'folk theo-
ries' that might once have inspired similar certainties, but turned out wrong
root and branch: folk demonology, folk cosmology, folk dynamics. As Stich
roundly puts it: 'Nor is there any reason to think that ancient camel drivers
would have greater insight or better luck when the subject at hand was the
structure of their own minds rather than the structure of matter or of the
cosmos.'5 But the question will be whether the certainties—the platitudes
exampled above—of commonsense psychology could be relevantly similar to
the mistakes early scientific theory made about force, inertia, demons, or
motion. It is certainly overplaying anything gained by introducing the word
'theory' to suppose that it guarantees such a similarity. Since the role or
network view applies to any term, the right point of comparison may be not
early cosmology and dynamics, but those other things that camel drivers
doubtless got quite right as they skilfully coped with their world: such un-
pretentious certainties as that generally it takes more effort to lift a heavy
stone than a light one, that sunshine warms and dampness chills, that people
may safely drink water but also can drown in it, that a long journey takes
more time than a short one, and so on without end. The very generality of
holistic considerations and the network model shows that these may be much
nearer the right points of comparison than anything more overtly 'theoreti-
cal'. So we have nothing to support the vision of likely error.
The same trap of overplaying what is achieved by introducing the term
'theory' is fallen into by Paul Churchland when he charges that commonsense
3. Patricia Churchland, op. cit., p. 256. See also Paul Churchland, 'Eliminative Materialism
and Propositional Attitudes', in Journal of Philosophy, 1981, p. 71.
4. J. Fodor, Psychosemantics (Cambridge, Mass.: Bradford Books, 1987), chapters 1-3.
5. Stich, Folk Psychology, p. 229.
232 Mind and Matter

psychology is likely to be bad theory, since it is both incomplete in its domain

and stagnant. 6 Whether or not these are bad signs in science, or at least in
research science, they certainly need not be in the loosely knit body of plati-
tudes connecting hunger and attempts to get food, or anger and bad behav-
iour. Nothing in the network model shows that if concepts are suitable for
stating truths, the generalizations of the network must be complete by any
standard. Nor need they be under pressure to evolve. Perhaps they remain
unchanging through the generations because they are certain, like the plati-
tudes about sunshine, rain, and longer journeys taking more time: one man's
stagnation is another man's certainty. It is also relevant to point out that the
impression of stagnation may be largely an artefact of the philosophical ten-
dency to abstraction—to talking as if the only relevant concepts are belief and
desire. If one looks instead at the richer textures of psychological understand-
ing that inform our lives, change is much more readily visible (it is also
noticeable that change typically does not bring falsification: in a future culture
people may not think in terms of, for instance, 'falling in love', and may not
do so, but that would not refute the fact that we do both).
On the other hand, it will not help to press the other way—to say, for
example, that the sample generalizations are incorrigible because close to
tautologous, or, if we prefer, 'criterial' for the states of anger, hunger, inten-
tion. For this meets the familiar point that if the criteria for the existence of a
state, in terms, say, of what causes it, are a priori, and if there are similar a
priori criteria for its consequences, then to avoid a pretended a priori status
for what is all too visibly a contingent and a posteriori matter—the correlation
between the antecedents and the consequences—it must be contingent and a
posteriori that any such state exists. Better then not to award too grand a
status to arbitrary elements of the network. Our success in understanding one
another is at least a success in putting behaviour into familiar patterns, and it
is a contingent fact that those patterns are not disrupted, in one or many
instances, sufficiently to baffle the attempt.

II. The Sovereignty of Physics

We now engage the essential part of the position: the idea that there is
sufficient tension between the claims of folk psychology and the increasing
body of knowledge of the sciences for it to be on the cards that the sciences
undermine it. The general eliminativist strategy is to try to say something
specific about the kind of state that having a belief or desire must be taken to
be, and then to argue that there may be no possibility of evolving science
countenancing that state.
Within that general strategy there are two ends to work from. One might
have a precise and exclusive sense of the states that science countenances, and
do relatively little to show that psychological states fail. Or one might think
6. Paul Churchland, op. cit., p. 74.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 233

that there is something very specific about psychological states such that,
without too much precision about science, one can nevertheless see that sci-
ence cannot countenance them. Of course, one can work from both ends at
once. Because it is physical thinking that centrally troubles me, I consider the
first end the more important.
Let us accept the sovereignty of physics. I take this sovereignty to be an
ontological thesis best stated in terms of supervenience: we suppose that fixing
all the facts discernible to physics fixes everything. 7 This does not deny the
existence of things, states, and perhaps properties that supervene on the
physical. But it ought to deny that there is a dimension of freedom in the way
these things supervene: it should not be contingent whether, given the physi-
cal truth, there is also this or that further truth, of chemistry, or biology, or
psychology. David Lewis believes that supervenience will be contingent upon
the absence of 'alien' properties, but I hope that their absence will be a truth
of our physics. If the way the supervening facts overlaid the total physical
truth were contingent, God would have to fix something else than the physics:
he would have to fix the possible relations that the overlying truths bear to the
physics. Real sovereignty should mean that there is not this extra thing to do.
Only if real sovereignty exists is everything physically fixed. Otherwise the
ways in which extra states overlie the physical states would amount to further
contingencies, ineradicable 'nomological danglers'. We could well mark the
difference by saying that the sovereignty of physics cannot countenance emer-
gent properties, where emergence is thought of as a brute contingent extra,
but can countenance supervenient properties. Supervenience is physically
fixed emergence.
If there is supervenience, is there to be physical explicability? Clearly the
natural or best explanation of a physical thing having a physical property need
not belong to physics. In most contexts, the best explanation of this chair
being within three miles of Carfax may be that it belongs to me, and this is
where I live. The doctrine of the sovereignty of physics implies only that the
position of the chair is physically fixed, not that the explanation is the most
natural or quickest to give, or the one asked for in any particular context. But
there should be this connexion with explanation: another God, who knew
how the creating God has fixed the physics, could explain why the chair is
where it is.
The doctrine of the sovereignty of physics does not entail determinism
over time. It does not entail that if God has fixed the physics and the laws at t,
he has thereby fixed the states of the world at t'. He may or may not have.
Sovereignty is a synchronic doctrine, not a diachronic one. But it should not
be read as implying that fixing the physics at a snapshot time t fixes all the facts
at t. It might not fix the historical facts at t, such as how big the tree was three
years ago, and in principle this might matter to the evolution of physical states
after t. To allow for indeterminism and external facts of this sort, it is best to
7. This definition is also given by, e.g., John Haugeland, 'Weak Supervenience', in American
Philosophical Quarterly, 1982.
234 Mind and Matter

use two temporal quantifiers: read the doctrine as saying that anything that
fixes all the physical facts at every time fixes all the other facts that obtain at
any time.
We are looking for argument implying that the sovereignty of physics
threatens the existence of psychological states. So which states does a science
like physics countenace? The question is ambiguous. It may be asking which
states physics allows and which it disallows. Or it may be asking which states
physics describes—which it is concerned with. It is one thing for any science
to imply that some state cannot exist, and a quite different thing for it simply
to work in other terms. Physics may deny, for instance, that some kind of
causation exists ('ghostly mental causation', perhaps) because it would break
conservation laws. But it does not deny that chairs exist, just by working in
other terms. Chairs are physical things.
Matters apparently stay the same when we think of properties, or states of
affairs: chairhood, or being a chair or a chiliagon. Singular reference intro-
duces more freedom in our descriptions of fact: physics does not work with
the predicate 'being within three miles of Carfax'. Yet surely its sovereignty is
consistent with truths of the form 'this chair is within three miles of Carfax'.
There is a seductive answer to the question of which states physics counte-
nances, which goes like this: Since physics is sovereign, a true theory must
be reducible to physics. Folk psychology is a theory; hence, if it is true it
must reduce to physics. But reduction implies coextensive predicates to give
us an 'image' of the theory taken over in the larger theory. However, there is
no real prospect of there being physical predicates coextensive with those of
psychology. Hence, there is no real prospect of psychological theory being
true. 8 In this argument the sovereignty of physics is taken to imply that all
truths can be given an extension-preserving mapping into truths of that
science. The implication is that if 'thinking that the ornaments need dusting'
cannot be made to correspond to a kind of physical theory, it cannot be true
that people think that the ornaments need dusting. But why is this argument
more plausible than an equivalent proof that there are no chairs within three
miles of Carfax?
I shall come at this problem by considering the dialectic with which Patri-
cia Churchland supports one part of her position. Her concern is to deny that
functionalism defends psychology against the requirement of reduction. (Func-
tionalism is here not much more than a label for the role theory, talking of
what we do because of our psychological states and what typically gets us into
them.) Functionalism famously deflects the demand for reduction because in
talking of functional identities between things we remain indifferent to same-
ness or difference of underlying physical states and mechanisms. This is true
within a science: in physics, radios, resistors, thermometers, and batteries can
be made of many different kinds of thing. In biology, birds and bees both fly,
but the material constitution of their wings and muscles is different. This
8. This is my phrasing of the view I take to be implicit in Patricia Churchland's Neuro-
philosophy (Cambridge, Mass.: Bradford Books, 19S6), chapter 7.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 235

suggests the possibility of benign cross-classification between psychology and

physics. As it is often put, there are 'role states' and variable 'realizing states'.
Churchland opposes this rapprochement with two arguments.
Considering the equation of temperature with mean kinetic energy of
molecules, she rightly points out that the identity is inapplicable to the tem-
perature of solids, of plasmas, and of empty space. It is unattractive to deny
that there is a reduction of temperature to mean kinetic energy of molecules,
so the moral she draws is that reduction can be domain-specific: temperature
can be one thing in a gas, another in other things. 'Though this is called
"multiple instantiability" and is draped in black by the functionalist, it is seen
as part of normal business in the rest of science . . .'9 The moral is that
reduction may be relative to a domain: 'we may, in the fullness of time and
after much co-evolution in theories, have one reductive account of, say, goals
or pain in vertebrates, a different account for invertebrates, and so forth'. The
implication is that wherever the predicates cross-classify, there is still to be
found a physical state which 'is' the property (of believing that p or being in
pain or whatever) in one thing, and another which 'is' the same property in
another thing.
The dialectical position is a little confused here. If reduction is allowed to
go on in this way, then it will not require extension-preserving translations of
predicates of the old theory into those of the reducing theory. It simply will
not matter whether one physical kind (and term) or many exist that corre-
spond to the old term. If there are many, we just say that temperature (or
whatever) is one thing in the one case, and another in the others. This gives us
a very relaxed attitude to the relation between the old predicates and their
physical partners, and one would have thought that far from draping such a
thing in black, the functionalist is happily celebrating the freedom thus of-
fered. It becomes no argument either against functionalism or against the
reduction by these standards of any folk theory to physics, that no extension-
preserving mappings can be found. So how does Churchland see this claim as
difficult for the functionalist, and good for the eliminativist? Surely it is be-
cause she sees the underlying classifications as the only real ones, so that the
concept of temperature turns out to lack 'empirical integrity' once this many-
one relationship to underlying physics emerges. It is as if God, to know
everything, can now avert his eyes from old-style temperature, the original
role state, and look at only the differing energies or realizing states. He would
need to frame no claims involving temperature to know everything. Tempera-
ture has lost 'empirical integrity'. In this view, physics as it were absorbs the
role state into the realizing state, and thus eliminates it.
The same attitude to physics is visible in the second thing that Churchland
says. Following Eng, she makes the point that mean kinetic energy of mole-
cules in a gas may itself be 'variably realized' since identity of mean kinetic
energy is compatible with quite different distribution of molecules and direc-
tions of motion: a given volume of gas may stay the same temperature as its
9. Op. cit.,p. 357
236 Mind and Matter
molecules change position. 'To be consistent functionalists should again deny
reductive success to statistical mechanics since, as they would put it, tempera-
ture of a gas is differently realized in the two cases. If, on the other hand, they
want to concede reduction here but withhold its possibility from psychology,
they need to do more than merely predict hardware differences between
species or between individuals.' 10 Again, the implication is that with the ex-
panded vision that enables us to see the endless variation in molecular posi-
tion, the overlying role state becomes invisible: reduced, eliminated.

III. Physical Thinking

Many readers will be protesting that the last two words of the last section
should not go together. The identification of role state with realizer state, they
say, is certainly a reduction, but certainly not an elimination. Before comment-
ing on that, I shall say a little about whether we should accept Churchland's
description of the case in physics.
If a bright learner asks the teacher why we can talk of the temperature of a
solid, when temperature is mean kinetic energy of molecules, and those in the
solid are not moving, it is hardly satisfactory to reply that temperature is one
thing in gases and another thing elsewhere. The learner wants to know what
justifies us in talking of the same property at all, so that a solid and a liquid
each can be at 100° Celsius, for instance. And there had better be an answer.
Of course there is. Physics can explain how gases, liquids, solids, and
space can be at the same temperature. In equilibrium thermodynamics the
concept of temperature is justified in the first place by the zeroth law, that if
two bodies are in thermal equilibrium with each other, and one is in equilib-
rium with a third, then the other is also in equilibrium with the third. 11 This is
what allows for the definition of an equivalence class of systems in thermal
equilibrium. The definitions allow for gases, solids, and any other physical
system to be in thermal equilibrium with one another. They are in equilibrium
when no net energy transfer arises from their thermal contact (e.g., from
placing the body in the space or gas or liquid). This will be because of a
relation between the mean kinetic energy of molecules in the gas, and what-
ever makes up the thermal energy of the other system: the vibration energies
of molecules in some solids, of molecules and electrons in metals, and so
forth.
Should we identify temperature with these 'realizations'? Why should we
want to? We already have it down as the property two systems have in com-
mon when they are each capable of thermal equilibrium with a third. In other
words, the fundamental thing to be said about temperature, as about number
and most concepts of physics, is that what is needed is a definition of the
equivalence relation of having the same temperature as . . . ; this and the
10. Op. cit., p. 357
11. See, e.g., C. J. Adkins, Equilibrium Thermodynamics (New York: Cambridge University
Press, 1983).
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 237

further relations necessary to produce a scale are provided, and nothing fur-
ther need be said about what temperature is, although a great deal can be said
about the kind of energy involved and its source in the kinetic energy of
molecules or other property of the body, gas, space, and so on. We also see
why the consequence Churchland is drawing, that propositions about tempera-
ture become invisible to science, is undesirable, to put it mildly: why should
we reject the salient equivalences as unreal just because we have found a
mechanism for them?
So one must beware of rewriting the relation between thermodynamics
and kinetic theory as if a question of identity and its answer were the main
point. In physics, as opposed to philosophical glosses on it, there is simply no
use for a statement like 'the temperature in this system is . . .' that implies,
for instance, that over a phase change the temperature has to change (it was
one thing, e.g., the kinetic energy of moving molecules, and is now another,
e.g., the vibration energy of static molecules). On the contrary, the only use
physics has for the form 'the temperature is . . . " insists that the temperature
of a solid and a gas in equilibrium is the same: 0°C, for instance.
It may help to see this if we reflect that the equivalence between two
volumes of gas at the same temperature is an equivalence visible to the kinetic
theory. Sharing mean kinetic energy of molecules is a concept of the theory
and defines an equivalence class of volumes of gases. In fact, given the gas
laws it follows from the equation of pressure with transferred momentum that
temperature in a perfect gas must vary with mean kinetic energy—the concept
as it were forces itself on the subject. Enc is therefore wrong to say that mean
kinetic energy is a kind term of kinematics 'only by courtesy'.12 One might as
well say that energy is a concept of classical dynamics only by courtesy, or
(since things of different constitutions each do it) that being in a state of
uniform motion in a straight line is a concept of physics only by courtesy.
Physics is not at all confined to constitutional kinds (as if it were not a physical
truth that differently composed material things might share mass or charge or
velocity).
To use a different example, finding what is the same about a closed system
of objects at different heights and moving at different velocities in a gravita-
tional field was a great achievement of classical dynamics. It was only by doing
this that dynamics identified the equivalence of potential and kinetic energy.
The whole point of the concept is that it classifies across difference of 'state',
where that is conceived in terms of mere spatial array and velocity of constitu-
ent particles. For just this reason it would be a misdescription of two such
systems to say that energy is 'one thing' in one system and another in another,
merely because the positions and velocities of the particles are different.
I now present the first part of my paradox. It is essential to physical
understanding that its predicates unify. They classify across differences of
realization, for the following reason: The basic concept is that of a system in a
state. The state is characterized by intensive and extensive magnitudes: pres-

12. Berent Enc, 'In Defence of the Identity Theory', in Journal of Philosophy, 1980, p. 279.
238 Mind and Matter

sure, force, and density are intensive (local) magnitudes, and mass, volume,
and internal energy extensive ones. These magnitudes take possible values
(the number that are capable of independent fix defines the number of de-
grees of freedom of the system); the art of the physicist is to find the right
variables and the right laws connecting them to give a function for the evolu-
tion of one state into another. Now it is absolutely vital to doing this that
concepts emerge that cover systems with different forms of thermal energy, or
cover changes of 'realization'—that allow, for example, for something such as
temperature or entropy which applies indifferently across changes of phase
from solid or liquid to gas. If there is no unification, there is no scope for
conservation across changes, and the central concept of physical thinking
disappears.
I sum up this truth about physics in the thesis (T): physical thinking is
essentially a question of finding the one state that covers many realizations. It
is forever a question of finding a unifying feature, a pattern in the particular
evolutions of systems. If this is finding a role state that permits different
realizations, physics only ever deals in roles: it is role seeking 'all the way
down'. No harm in that: thesis (T) is bland enough, only perhaps serving to
avert a lay misunderstanding of physics as essentially concerned only with
microscopic facts, or concerned only with identity. Thesis (T) opposes the idea
that the 'states' that physics countenances are to be thought of as literally
spatial items, morphologically identified. Nothing could be less true to phys-
ics. It requires no controversial functionalism to see that is simply false that
accelerating at 25m per second per second 'is one thing in this car and another
in that cannonball or that being able to pass petrol at 25cc per sec is 'one thing
in this tube of circular cross-section and another in that one, which is square'.
Remember too that even causal favourites like solidity are only the possession
of some property that precludes occupancy of the same space by other things
with some properties.
The view of physics that (T) warns against could be called the Tractarian
view, although in calling it this I am conscious of pointing to a cluster of
thoughts, or images determining thoughts, rather than a clear unique doc-
trine. In its simplest form, it is the view that physics itself is incapable of
describing equivalences between systems that differ in some microscopic
configuration—in the spatial array of constituent particles. The idea is that
identity of a physical state consists in the spatial configuration of particles, and
evolutions of state are changes in this over time. Every state is then what it is
and not another thing—a unique combination of a value for each physical
parameter at each place. But for it to explain at all, physics must find the one
in the many. This means finding properties that do not answer only to configu-
rations of constituent particles (although of course it may also, at some level,
find properties that do so answer). The moral drawn from temperature and
energy is quite general, and must be if physics is to explain.
I hope now that the reader feels the first stirrings of unease. It is the
unique state, the realizing state, or array of magnitudes or tropes or instances
of properties at points, that causes. It is here that the 'making happen' hap-
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 239

pens: how, then, can we identify the cause by citing the relational, disposi-
tional or role-given properties with which physical thinking leaves us? At least
in the seventeenth century the condensation seemed to yield something solid,
and (people wrongly thought) something that leads us to intelligible point-by-
point causation. But we cannot have that thought any more. We are left only
with point-by-point magnitudes for things that are essentially dispositional—
electrical charge, for instance—but we are not left with a 'quality' that under-
lies the 'power', or with a 'ground' that underlies the disposition, or with a
state that finally realizes the role. If the property is not identified as a power,
or equivalently, if the 'state' of it being realized is not identified as a role state,
then the categorical, pure presence of an instance at a point will be unknow-
able to physics.13
An especially interesting example of the lure of the microscopic, Tractar-
ian physical state is given by John Haugeland's excellent paper, 'Weak Super-
venience'. Haugeland is arguing as I am against the need for identity theory.
But his example of the problem of identifying the 'robust' events of common
description with the 'mathematical events' of physical theory depends eventu-
ally on a Tractarian view of what these events are. His example is the crossing
of a pair of waves generated from different ends of a tank. Imagine that as
they cross each crest hits a cork. Haugeland asks which event describable in
the language of ultimate physics is identical to the wave-hits, and finds none:
'when we turn up our microscope, however, and look at the positions and
velocities of the water molecules, there isn't a trace of either wave-hit to be
found anywhere. . . ,'14 But why should one expect to find the events and
states of interest to physics by 'turning up our microscope'? If fluid dynamics
is treated in kinetic theory, the magnitudes characterizing a liquid at a point
and time (velocity, viscosity, density, pressure, temperature) are not associ-
ated with the specific detail of microscopic states. They are associated with
averages of properties of many microscopic entities. The 'mathematical
events' of physics are therefore quite large-scale enough to map onto the
events of waves travelling and crossing. It is the individual events and magni-
tudes that are not.
Does a Tractarian vision infect eliminativism? Consider Stephen Stich's
discussion of the relation between mind and brain.15 Trouble arises because,
via the notion of a state, commonsense psychology becomes landed with
rather specific commitments to the nature of neurophysiological workings.
So Stich can urge that the notion of a belief state is undermined if it turns out
that one part of the brain is causally responsible for my saying that there is a
cat under the bed, and another part for my reaching down to lift it out. The
idea is that, according to folk psychology, one belief state (that the cat is

13. I am here grateful to Robert Kraut and George Pappas. The difficulty over solidity is
heralded in Locke's ambivalent attitude to the possibility of knowing the qualities, as distinct
from just the powers, of things (including microscopial things): compare, for instance, Essay,
Book II, chapter IV, sections 1 and 6. See also essay 14.
14. Haugeland, op. cit., p. 100.
15. Stich, op. cit., p. 237 ff.
240 Mind and Matter

under the bed) is responsible for both the saying and the doing; neuroscience
finds two different states separately responsible for the saying and the doing;
hence folk psychology was wrong. Notice, however, that typically people
tend to say only that there is a cat under the bed when they are disposed to
act in whichever way is appropriate to that information, so presumably there
typically exists some connection between any different areas of the brain that
are involved, ensuring that one area goes into a cat-behaving state only if the
other area goes into a cat-reporting state. Hence there is room to find a
single fact or feature such as that of having the action and the speech area
coordinated. But this will be simply a fact about the organization of the brain
(it is not a fact that is 'located' anywhere). I suspect that Stitch failed to see
this counter because of implicit Tractarianism: a count of located neuro-
physiological states would be blind to that kind of state. In short, unless we
are in the grip of the Tractarian vision, the argument is out of court from the
beginning. The single belief state of folk psychology involves no commitment
to one or many genuinely internal mechanisms, any more than the single
kinematic state of a motor car—being in uniform motion in a straight line,
say—requires that anything, let alone any given number of things, is happen-
ing inside it.
If predicates are taken in the resolutely functionalist way I am commend-
ing, then we are not finding that F-ing is one thing in one object and another
in a second, because they do it by different mechanisms. It might be a differ-
ent thing, but only if it is a different thing that they are doing. From the
functionalist's point of view, this would be because it is impossible to find a
functionally equivalent state across the divide. But that would be a different
matter, although English is unfortunately cavalier at the crucial point.16 For
the English question 'what is flying?' permits either a (quick and easy) role-
state answer, or a (long, and variable) realizing-state answer. It may be right
too that in many contexts the latter is the more interesting. What is not true is
that this undercuts role-state explanation, nor that it threatens the causal
efficacy of such states.
Notice, however, that impatience with traditional ('analytical') philosophy
of mind might lead a forward-looking eliminativist to resurface here. Perhaps
traditionally there were two questions, one scientific and explanatory, and the
other more analytic. But why respect the second? Perhaps in a modern cli-
mate the question 'what is thought (belief, desire)?' admits only of a scientific
construction: either we are asking for underlying mechanisms and need to sit
at the feet of neurophysiologists, or we back to playing with concepts that lack
empirical integrity. Alas, there is no escaping metaphysics without doing more
of it than the mere hymning of science involves. The eliminativists believe that
the sovereignty of physics and the rise of neurophysiology pose a distinct
threat to psychological categories. But they cannot make this claim look
plausible without giving or presupposing some account of how science threat-
ens role states. They retain responsibility for showing how the undermining

16. I owe David Lewis thanks for pointing out the need for caution here.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 241

works: in round terms, what belief and desire explanation was, that it should
be shown to have had its day.

IV. Role States and Causation

Thesis (T) says that physical thinking is essentially a matter of finding the one
feature that covers many realizations. One might call this the one state, but it
is dangerous. We envisage states, very naturally, as things with positions and
boundaries, and this spatial imagining is surely responsible for part of the
problem. With this caveat in place, I now turn to the idea that finding the
causes—the real causes—of events cannot be a matter of identifying a role
state or feature covering variable realizations. I shall illustrate the difficulty—
although it is of far greater generality—by thinking of the trouble we have
fitting psychology together with causality.
Folk psychology has little overt to say about states. If it talks of them at all,
it is to talk of us who get into psychological states, not of any states that get
into us. If I get into a state of believing that I am Jesus, no doubt things will
have gone on in my brain. But there is no first-order project of finding
whether the state I have got into is some very state that has got into this part
of me. Commonsense psychology gives no reason to think of the states we get
into as spatial arrays. None of the platitudes defining folk psychology talks in
these terms: as with physics, it is philosophical gloss, not first-order theory,
that insists. So why impose talk of states and then reify and locate them? The
ruling answer goes like this: Folk psychology may confine itself to using
predicates such as '. . . believes (desires, etc.) that p'. But we need a view of
the conditions making such ascriptions true. Functionalism is a start. But we
want our psychological states to have causal powers, enabling them to affect
the real world, and we want them to do so in ways we can see as conforming to
the sovereignty of physics. The best way to do this is to identify the states
making the ascriptions true, first with neurophysiological and then, hopefully,
with physical states. If those states will not go into that shape, then so much
the worse for them.
This is the heart of the issue, set forth with his usual clarity and force
by Lewis, who wrote that his argument for identity theory (the identifica-
tion of role with realizer states) parallels the following, 'which we will find
uncontroversial':
Consider cylindrical combination locks for bicycle chains. The definitive
characteristic of their state of being unlocked is the causal role of that state,
the syndrome of its most typical causes and effects: namely, that setting the
combination typically causes the lock to be unlocked and that being un-
locked typically causes the lock to open when gently pulled. That is all we
need know in order to ascribe to the lock the state of being or not being
unlocked. But we may learn that, as a matter of fact, the lock contains a row
of slotted discs; setting the combination typically causes the slots to be
aligned; and alignment of the slots typically causes the lock to open when
242 Mind and Matter

gently pulled. So alignment of the slots occupies precisely the causal role
that we ascribed to being unlocked by analytic necessity, as the definitive
characteristic of being unlocked (for these locks). Therefore alignment of
slots is identical with being unlocked (for these locks). They are one and the
same state.17
Lewis claims three advantages for so thinking. First, the states become recog-
nized as 'real and efficacious'. Second, 'unrestricted natural interdefinition of
the state and others of its sort becomes permissible'. Third, it becomes intelli-
gible that the state may sometimes occur despite prevention of its definitive
manifestations. The alternative he considers preserves the separation of role
state and realizer state. But, according to Lewis, this denies efficacy to the
former, for a pure disposition is an inefficacious entity.
The last two advantages are not central to my problem, but I shall briefly
register a query about each of them. First then, mutual definition seems to go
on just as well if we stay with the role predicates of commonsense theory: if
there is a definition of belief or desire to be had, it will be had whether we
think of it as saying what an underlying mechanism is, or instead as defining
the truth that someone is believing or desiring. A definition of flying can be
had (roughly: staying aloft in a gravitational field without external support)
without adverting to any particular mechanism whereby some particular thing
does it.
Second, the identification of role and realizer state may seem to solve
problems over externally prevented manifestations of state. But this is not
clear, for these problems arise as much with the identification in place as
without it. This is easy to overlook, since once we think of the state as an inner
spatial presence, it seems evident that it can be there regardless of whether its
effects are its typical effects. But what will not be evident is that the internal,
realizing state is on this occasion the original role state, for this is now contin-
gent. Consider the lock again. Another lock might share a whole physical part
with the original: five disks coming into line, for instance. But it may have a
device that prevents this from being enough to open it (you also have to
whisper the magic word that triggers the minute relay). The five disks are in
line, but the lock is not unlocked. If we talk the identity way, we have to say
that its being unlocked is a composite: its five disks getting into line and its
relay tripping. For this lock, it is not true that its being unlocked is its five
disks getting into line. Calling the state of being unlocked UL, and that of
having five disks in line S, we learn that the following is not a valid argument:

Lock 1 is unlocked iff it is in state S,

so State UL is state S. Lock 2 is in state S,
so Lock 2 is in state UL,
so Lock 2 is unlocked
17. David Lewis, 'An Argument for the Identity Theory', Philosophical Papers, Volume 1
(New York: Oxford University Press, 1983), p. 100.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 243

because Lock 2 is not unlocked. Such an argument trades on forgetting that

the identity 'state UL is state 5' must be relativized to locks for which it is true
that when the five disks are in line, they are unlocked. How to tell which locks
these are? Easy, provided we know when something is unlocked. But any
difficulties in telling when a lock with five disks in line but unopenable is
unlocked remain. For example, if there is a disruption of the typical workings
of the first lock, preventing the chain opening, we may say that it is unlocked
but jammed. But we may prefer to say that it is still locked. Or there may be
an accidental disruption to the first kind of lock preventing it from getting into
the state S, although it now opens when lightly tugged (it is a raped lock: the
fifth disk is damaged, and the other four give it its function). Here we are
hardly likely to say that it cannot get into the state of being unlocked. The role
state theorist who avoids the identity has exactly this freedom, or (if we want
principles for settling the description) this problem. Of the first lock he can
say that the thing is unlocked but jammed. He can say this because he can say
that if the occasion were right for manifestation of its disposition (e.g., if the
rust were out of the way), it would open. Or, he can say that it can't be
unlocked, because this is the right occasion for display of the disposition, and
the thing will not open. Similarly, if it opens although the state S has not come
about, the role state theorist can deny that it is unlocked (forced, rather) if
unlocking implies responding by opening to the right kind of input (a key, not
a drill).
In other words, indeterminacies over what to think when we would expect
a disposition or other functional trait to be manifested, but it is not (or not to
be manifested, when it is), need handling in the same way whether or not the
identity is flourished. If there is a disruption between the physical state that
normally explains the disposition and its exercise on an occasion, the identity
theorist simply confronts the open question of whether to say that the state 'is'
on this occasion the original state. His penchant for identity does not tell him
the answer. And the role state theorist faces the question of whether this is the
'right' occasion for exercise of the role. The same is true of the more urgent
case in which paralysis prevents beliefs and desires from becoming manifest.
There are some facts about the paralysed agent that would be expected to lead
to his acting some way, but others that prevent it. Is he disposed so to act? He
would be if the occasion were right (i.e., if the disability disappeared), but the
occasion will not be right; he will not talk or act as things are. If this leaves us
not knowing what to say, try bringing on the identity theory. Does he possess a
state that (in him) is the belief or desire? This is just as hard to tell. There are
difficulties in knowing what to say of such cases: the contingent or kind-
relative identification with underlying state makes it no easier.18
18. Lewis reminds me (private correspondence) that his target in the paper was not this kind
of role-state theory, but a behaviourism that thought it did know what to say: it had to say that if
there was no display of the disposition, the psychological state was absent. I quite accept that this
position is unhappy, but my point remains that it is not necessary (nor, as Lewis of course knows,
sufficient, since you have to find a principle for selecting locks for which the identity holds) to
insist on the identity by way of refuting it.
244 Mind and Matter

Of course, these were two subsidiary arguments for making the identity-
arguments parallel to the famous Davidsonian argument for event-identity
between the mental and the physical. In each development the most impor-
tant issue is the first that Lewis cites: that of causation, and the intuition that
role states do not cause, whereas realizer states do. Alongside the thesis (T)
we now have the antithesis (A):

Thesis (T): Physical thinking is essentially a question of finding the one

state that covers many realizations.
Antithesis (A): If one state covers many realizations, it is they and not it
that are causally relevant (powerful, active).

These are not incompatible, of course, but together they do give us the
conclusion (C):

Conclusion (C): Physical thinking is not a question of finding states that

are causally relevant.

And this is sufficiently strange to count as paradoxical. If physics is not the

discovery of causally relevant states, what is?
So far I have merely introduced (A) and done nothing to defend it. First,
we need to be clear what it says and does not say. It does not say that role
states cannot be cited in causal explanations. They may be if causal explana-
tion 'points toward' causally active states, without itself citing them. Thus, to
take Jackson and Pettit's example, the fact that someone coughed may be
mentioned in explaining why the conductor got angry.19 But it features in such
an explanation, according to them, by pointing us toward the particular state
that actually made the conductor angry—Fred's coughing, or perhaps some
feature of Fred's coughing on this occasion. We mention role states, on such
an account, in order to shrink the domain from which causally relevant particu-
lar states must come. In the same way we might cite a power or disposition,
not because in mentioning these things we are actually giving the relevant
feature, but because it confines the domain within which such a feature exists.
Thus we can mention the dormitivity of the opium and provide an explanation
of someone sleeping, but what we are saying is that there is some other
property the opium has, that causes sleep.20
For a similar reason (A) is not denying that role states may be nomologi-
cally connected with effects. A causally relevant property will bring about an
effect. It will also bring it about that other higher-order, relational, and dispo-
sitional properties are instanced, and they may then be in a lawlike correlation
with the effect. Thus in a gas the increase in kinetic energy of these molecules
in this configuration will transfer energy to the thermometer; it also will give
19. F. Jackson and P. Pettit, 'Functionalism and Broad Content', in Mind, 1988, p. 394.
20. Ned Block, 'Can the Mind Change the World', in Meaning and Method: Essays in
Honour of Hilary Putnam, ed. G. Boolos (Cambridge: Cambridge University Press, 1990).
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 245

the gas new powers. The increase in kinetic energy brings it about that the gas
has some property responsible for extra available thermal energy, and it
thereby also brings it about that the gas bears a different energy-exchange
relationship to many other things. These changes in role state are, however,
epiphenomenal: the molecules do the pushing, and the other properties, super-
vening upon the first, come along for the ride.
Except, of course, remembering the discussion of physics, it will not really
be the molecules doing the causing, but (in the first instance) distributions of
other parameters of energy and force within them. And if thesis (T) is right,
then for all we shall ever know there will be other magnitudes and realizing
states beyond those again, all the way down. If our physics stops at some
level—say with electrical charge—then there will still be a quality or qualities
realizing that charge. But, as with Cartesian egos, it will be forever hidden
how many there are, or what it is, or whether it is the same on odd dates and
even dates.
There are other arguments in favour of (A). Writing of the second-order
property of being provocative to a bull (i.e., having some property that causes
the bull to react) Block says: 'supposing that provocativeness provokes the
bull would be supposing a strange and gratuitous sort of overdetermination of
the bull's anger. . . . [T]o suppose that it always happens would be to suppose
a bizarre systematic overdetermination' .21 And our dislike of this can be forti-
fied by remembering how easy it is to generate supervening states. To use an
example of Lewis's: if the steam in the cylinder is at 190 pounds per square
inch, then it is also in the state of being either at 190 psi, or full of suspended
particles of gold dust. We think the former causes the piston to move, but its
being in the latter state does not. Disjunctive states are bad candidates for
causal relevance; quantificational states (someone coughing) and higher-
order states (having some property that causes sleep) seem just as doubtful.
If thesis (T) characterizes physical thinking, and antithesis (A) is not short
of defence, can we live with conclusion (C)? Perhaps we can. After all, lawlike
explanatory relevance is left to us. There is nothing wrong with saying that the
thermometer went up because the temperature is rising, even if ('strictly
speaking') what made it go up was a succession of microscopic impingings, or
rather, the instancing of the properties that ultimately realize all the powers
we can ever know about. We might even take a Humean pleasure in reflecting
that the ultimate causes of things are forever obscured from us, so that all we
can do is mark the patterns they reliably create as events unfold. And it draws
some of the sting from the epiphenomenalism of the mental if its causal
inefficacy is on all fours with that of changes of temperature or energy levels.
Nevertheless, it would be good to put into place some other route to a solu-
tion. We ought to reflect that in spite of the shift in our understanding of
causation that Newton and Hume brought about, it is not compulsory to
describe it by saying that we never identify real causes. It is hard to believe
that 'being a feature in virtue of which things happen' is itself a transcendental
21. Block, op. cit.
246 Mind and Matter

property, forever defying identification. There ought to be a humanly accessi-

ble notion of what it is that makes things happen—one we can rely upon in
making recipes and controlling events.

V. Higher-Order Causal Relevance

Things with the same temperature form a certain equivalence class, although
the thermal energy can be created by many different means. Sameness of
temperature is therefore a higher-order property: the having of some energy
source ensuring that no heat is transferred when in thermal contact with any of
a class of other things. Now imagine a chain of events as follows. A thermome-
ter is affected by its thermal contact with something, and records a tempera-
ture. Suppose it is connected to a further system—a current flows in a wire if
the reading is below a certain level. The current in turn triggers a switch, and
lights the boiler, raising the temperature. Then there is a good sense in which
the system 'looks through' which realization of thermal energy caused the
reading. It is sensitive equally to the different creations of thermal energy (if
the thermometer is flexible enough, by moving molecules, vibration energies,
electromagnetic energies and so on). That is, the current flows regardless of
any particular realization of the temperature, and in the same way the boiler
lights regardless of any particular configuration of the electrons in the wire
carrying the current. The boiler raises the temperature today as it did yester-
day. We ought to have a humanly applicable conception of causal relevance,
enabling us to say that the system works the same on each occasion: the same
features cause the same effect.22
Consider now a lock. Think of a burglar as an instrument. The lock being
unlocked can be responsible for the burglar getting in: it is the feature to cite,
and to use if we wish to control the movements of burglars. Its being unlocked
is the upshot of its having four disks in line, on this occasion. But that is not
the feature the burglar detected, or is perhaps even capable of detecting. Nor
is this only a problem about burglars. A much simpler instrument can be
sensitive to whether there is no force on the chain, without being sensitive to
whether this is because there are four disks in line, or five, or anything else.
Again, the upshot 'looks through' the preceding realization.

22. An excellent discussion of the issues that trouble me, in the Davidsonian context of event
identity, is the interchange between Ted Honderich and Peter Smith: Ted Honderich, 'The Argu-
ment for Anomalous Monism', in Analysis, 1982, p. 59; 'Anomalous Monism: Reply to Smith', in
Analysis, 1983, p. 147; and 'Smith and the Lover of Mauve', in Analysis, 1984, p. 86. Peter Smith:
'Bad News for Anomalous Monism?' in Analysis, 1982, p. 220; and 'Anomalous Monism and
Epiphenomenalism: A Reply to Honderich', in Analysis, 1984, p. 83. Honderich's original in-
stinct is also found in F. Stoutland, 'The Causation of Behaviour', in Essays in Honour of G. H.
von Wright, ed. J. Hintikka (Dordrecht: Reidel, 1976). Another good discussion is Davidson,
'Thinking Causes', in Mind and Brain: Perspectives in Theoretical Psychology and the Philosophy
of Mind (Bielefeld: Zentrum fur Interdisziplinare Forschung, 1991), with the accompanying
sceptical comments by Kim and Sosa. My position is most similar to that of Peter Menzies,
'Against Causal Reductionism', in Mind, 1988, p. 551.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 247

A vivid case of this 'looking through' is indifference of a physical result to

the question of which instance of a quantification is realized. The burglar
alarm goes off because someone is in the room. It was Bill: Bill (or more:
some particular result of Bill being in some particular place at some particular
time) caused it to go off, but it is still true that it went off because someone
was in the room. That is what makes it go off; that is what is supposed to make
it go off. It is not as if it malfunctioned. You cannot take it back saying that it
is supposed to go off because someone is in the room, whereas it actually was
made to go off by Bill being in the left hand corner of the room standing on
one foot. Is this giving in to a 'bizarre causal overdetermination', since on a
given occasion the instrument will have responded both to some highly particu-
lar flux of energy and to someone being in the room? Surely there is nothing
bizarre about it. We simply do not have to choose between these properties as
ones whose instancing makes things happen, ones with causal power and
relevance.23
But does citing anything less than the particular, as we do when we find a
common feature in causal transitions, leave us with just a signpost to some-
thing else, which alone has 'real' causal relevance? I say not. In finding causal
relevance we are finding features, like rising temperature, or carrying current,
that are essentially repeatable, but whose presence can be registered, and
brought about, by mechanisms that function in the same way on each of
numerous occasions. The one we find in the many can also be a one whose
being there makes things happen.24
So far then I am insisting on having my cake and eating it: separating the
role and realizer states, but keeping genuine causal efficacy for the role state.
It might occur to some to wonder whether this challenges the sovereignty of
physics. Again, I say not. The sovereignty of physics requires only that the
connexion between the role or disposition and the underlying state be explica-
ble. Its sovereignty would be challenged only if there were something it could
not explain in (say) the fact of the five disks coming into line enabling the lock
to be opened. But there need not be. In this lock, when the five disks are in
line, there is no force opposing the tension of a pull on the chain. So the chain
moves. In other locks, when the five disks are in line, there might still be a
force opposing the tension of such a pull. In them, the chain will not move.
The property or fact of being locked or unlocked is thus explained, and
physical explicability does not require that we work in terms of states and
their identity at all. We do not have to 'identify' the velocity of the car with the
23. Lewis rightly reminds me that his original example was a bicycle lock, and that cylindri-
cal locks are not commonly used on houses. But 'folk burglar prevention' sounded better to me
than 'folk bicycle theft prevention', so I have let it stand.
24. It is only the particular instance, not the feature, that causes things to happen. Instead of
'the instancing of a feature', we get the event in its full (transcendental) particularity. Pondering
this intuition, I find myself reminded of Bradley's hostility to abstraction: the view that only a
picture of reality that involves no abstraction (i.e., reality itself) is (really) true strikes me as very
like the view that only the total Tractarian particular causes. But I cannot explore the relationship
here. See Stewart Candlish, 'The Truth about F. H. Bradley', in Mind, 1989, for the 'identity'
theory of truth.
248 Mind and Matter

application of the energy to the wheels in order for the velocity to be explica-
ble by the energy.
The intuition I am opposing is probably this: If there is anything other than
identity, then there are two things (two states) to be connected—on the one
hand, the underlying physical configurations, and on the other hand, the fact
that the lock opens when gently pulled. But unless we can squeeze these
together, as it were, there is then bound to be a gap for physics to cross, and it
will never explain the exact landing point: the very way the categorical state
matches up to the disposition. And my rejoinder is that any intuition that
there is an explanatory gap here simply transposes into the same intuition that
there is an explanatory gap on the identity story: a gap in understanding why
or when state S is state UL. If we know why the configuration gives rise to the
disposition, we know this, but otherwise we do not. Nothing is made more
intelligible by the identity.
My claim then is that being unlocked is not an epiphenomenon, rendering
literally false folk burglar prevention—the theory in which we say 'because it
was unlocked, the bicycle rolled downhill'; 'because it was unlocked, the
burglar could push it open'. The fact that no force opposed the tension on the
chain is in perfectly good standing as a fact that matters. It was because of that
fact (which in turn may be because the five disks were in line) that gravity or
the burglar met no resistance. Would it be somehow better not to cite the fact
of it being unlocked, and only to cite the disks being in line? Not at all: gravity
and the burglar 'look through' the reason why it was unlocked.
If (A) was at fault, what of the degenerate cases, such as disjunctive states,
or dormitivity? Here I can only sketch an approach. Arbitrary disjunctions?
The disjunctive state of 'having the steam at 190 psi or being full of suspended
particles of gold dust' is not a state that any natural causal system 'looks
through' to. A cylinder would be in that state by being full of suspended
particles of gold dust. But there is (I take it) no normal effect common to that
state and to the state of being full of steam under pressure. Only if there were
would the state begin to stand in the same relation to having the steam as
temperature does to a particular molecular distribution, and thence to look
like a candidate for causal efficacy. There is no genuine nomological connec-
tion between satisfying this predicate and doing anything (perhaps I should
say: anything whose unity is visible to a non-Goodmanian, well-behaved cate-
gorization of the world). Quantification? A conductor, like a burglar alarm,
may be sensitive to someone coughing. Making someone cough gives us con-
trol over his temper, a recipe for making him seethe, and a causally relevant
state of an audience. We have here a system that looks through its realiza-
tions, and I see no objection to causal efficacy. Higher-order properties, like
the provocativeness of the cape? A bull is certainly a good instrument—the
best going—for detecting whether a stimulus is one that provokes bulls. If a
change in temperature is causally relevant to the thermometer, why isn't a
change in provocativeness causally relevant to changing demeanour in the
bull? My own answer to this would be to reverse Block's. What is wrong with
citing this feature is not that it is causally irrelevant but that it is explanatorily
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 249

such a dud: it provides no basis for prediction and no useful recipe for doing
things that control bulls. Being told to wave something provocative at the bull
is indeed merely being given a signpost: it points me toward the need to
discover the properties that cause the reaction. Whereas there is virtue in
citing temperature: I can control the rise of the thermometer by controlling
temperature, without needing to discover anything about the realization of
temperature at all.
It may be said that in all these examples, the causal efficacy of the role
state, like the role state itself, supervenes upon the Tractarian particulars and
their evolutions through time. I do not want to deny this, provided we are
comfortable with such a property—provided it is not a subtle invitation to
denying that the causal power, the relevance and efficacy of the feature, is real
enough.

VI. Physical States

If we have a good idea of how the thermometer can be made to act by the
temperature, can we immediately see how we can be made to act by beliefs
and desires? Obviously there is a huge gap: sameness of temperature is at
least visible to physics; sameness of belief arguably not.25 Still working, as it
were, from the physical end, the question will be one of defining a more
realistic concept of a physical state or property. Presumably physical states
include the possession of a physical property by a physical thing; the bearing
of physical relations by «-tuples of physical things; the being in complexes of
physical states or bearing of amalgams of physical properties by conglomera-
tions of physical things. There is nothing in the sovereignty of physics that
forbids (physically explicable) complexity, amalgamation, or conglomeration.
It is sometimes thought that there should be, because of the following
argument. Physics explains by citing laws; there may be a law covering the
transition from A to B, and one covering the transition from C to D, but this
would not entail that there is one covering the transition from (A & C) to (B
& D). This may be true about laws. But physical explicability is agglomerative
and transitive, even if law is not. Thus when a system composed of A and B
evolves to a subsequent state C and D because of a law L determining C from
A, and another L' determining D from B, then even if there is no single law
covering this evolution, there is still nothing physically unexplained about it.
For an example, consider a body of known weight, position, and temperature
<p,T> in a gravitational field of known strength and direction. Suppose a
kinematic law K will give a position p' for it after t seconds. Suppose the body
consists in a battery passing current through a circuit, and a law of electromag-
netics L determines a temperature T' for it after the same interval. At t it will
be in a certain position and at a certain temperature, and nothing in this

25. For recent scepticism, see Hilary Putnam, Representation and Reality (Cambridge,
Mass.: Bradford Books, 1988), p. 84 ff.
250 Mind and Matter

evolution is physically unexplained. The transition from <p,T> to <p',T'>

is both determined and explained by physics—even if physics uses two differ-
ent laws to do it. It is a transition in which nothing is physically inexplicable.
God did not have to do anything extra to secure this evolution of state.
Churchland's vision of the relation between special sciences and physics
demands coextensions of terms. Will the sovereignty of physics require any-
thing so strong? In the example given, it does not matter if physics has no
interest in the particular evolution, and no kind term to pick out the complex
states involved. Further examples suggest a two-stage analysis of this phe-
nomenon. The first part allows for 'token-token' identity, but the second is
less familiar. Consider (to use Fodor's example) the physical process summed
up in the geological 'law' that meandering rivers erode their outside banks. A
physical process, and one in principle explicable by fluid hydrodynamics. Yet
neither river nor meander nor bank is a term of physics. All that the 'reduc-
tion' of the law to one of physics, in this relevant sense of the word, requires,
is that rivers are (each of them) some, but not necessarily the same, kind of
flowing liquid; that meanders are (each of them) some, but not necessarily the
same, kind of curved configuration; and that banks are (each of them) some,
but not necessarily the same, kind of material liable to lose its cohesion under
transverse forces. Each token of the type is an event describable in terms of
hydrodynamics. If we think these kinds are invisible to physics, then the fact
that a kind is thus invisible is no bar to reductive physical explanation. A
fortiori it is no motive to 'eliminativism' or rejection of the truths expressed in
terms of the kind, even given the sovereignty of physics. The coming of
hydrodynamics does not prove that there are no folk rivers.
The example takes us only halfway. We are already prepared to believe
that bank erosion is a physical process, and we do not mind saying that rivers
and meanders are, each of them, something physical. But we might still worry
how physical explicability transfers from any given token of the type to the
overall explanandum. Would it not require seeing a physical 'shape' in what is
perhaps invisible to physics: the reason the disjunction of tokens all count as
examples of the same type—river, meander, or bank? It is here that the
demand for an 'image' of the predicate in the explanatory science (physics)
reemerges. It is as if it may be all very well explaining each token of the type,
but unless there is a physical explanation of the tokens constituting and ex-
hausting examples of the type, the original explanandum is not yet touched.
Thus it is that more impressive cross-classification is typically introduced via
examples like chairs, ornaments, money, and so on. These kinds correspond
even less well to physical boundaries. Whether or not they think it undermin-
ing, writers generally agree that there is a distinction between classifications
that carve nature at the joints and those that do not, and these kinds, reflect-
ing as they do the particular interests of human beings, surely do not.
This thought needs a little attention. Of course it is true that we discrimi-
nate things by other than physical properties, we 'look through' things like
mass, velocity, shape, conductivity, constitution, and so on. And a bullheaded
attempt to find disjunctive physical kinds underlying most terms of classifica-
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 251

tion, including those that enter into some laws, is fruitless. Again using Fo-
dor's example, it is a law that bad money drives out good, but an attempt to
'define' a term like 'money' by thinking of the enormous (limitless) potential
disjunction of physical kinds of stuff that might serve as money, and then
framing 'laws' covering that disjunction, is a hopeless way to try to make the
law 'physically explicable'. Even if such a disjunction were somehow created,
it is doubtful whether we could think of it as making anything explicable. It
would not be the 'physical' properties of the members of the disjunction that
are responsible for them being there (round bits of metal, printed pieces of
paper, cowrie shells . . .): there is nothing visible to physics tying the bundle
together. Since the disjunction threatens to be quite shapeless, we might insist
that it still plays no role in helping us to understand why things of that kind
drive out other things of that kind.
But this is the wrong way to go. Instead we must enlarge our gaze to think
of the wider system: us-using-money, or us-discriminating-chairs. It does not
follow from the cross-classification of chairs with respect to properties like
mass, material constitution, shape (up to a point), and so on that there is
anything invisible to physics in the wider system. It would be possible, for
instance, to construct a physical system including various receptors that se-
lects objects in its environment by properties other than weight, mass, consti-
tution. Think of a robot capable of 'sitting'—taking up some position, and
sensitive to feedback telling it whether some of its parts are under mechanical
stress in that position. Then think of it discriminating likely objects on which it
can 'sit', capable of selecting further among them on the basis of feedback
from its own states, and eventually able to mimic our own chair-selecting
practices (if it is complained that we discriminate chairs by features other than
whether we can sit on them without strain, imagine the robot endowed with
other sensitivities—e.g., to the history of production or whatever else mat-
ters). Of course it will be by physical means that its sensors and feedback
mechanisms work. But given the sovereignty of physics, the same is true of us.
The whole system of chair discrimination by robot would be physically explica-
ble. Hence there is no a priori reason why the same is not so of us. In this
development the physicalist explanation does not try to complete an inventory
of the woods, plastics, metals, and other things that chairs are and might be
made of. Nevertheless the explanation proceeds by finding what is physically
common to occasions on which a classification is effected. Notice too that the
physical explicability of the classification in turn does not imply that the
robots are built identically. As already emphasized, physics is capable of being
quite relaxed about what makes two occasions physically similar. A circuit
diagram gives a perfectly good physical description of a device, although there
are endless different things that can serve as wires, resistors, or capacitors.
Similarly with money: nobody of course has the least idea how it would go
in detail, because social facts are one stage further from physics than psycho-
logical ones. But if there is physical explicability of people desiring things and
exchanging things, then there will be physical explicability of Gresham's law,
and again without providing the shapeless list of things that can serve as
252 Mind and Matter

money. The upshot is that it should not be barely asserted that 'chair' or
'money', let alone 'river' or 'bank' is a kind invisible to physics, even though it
cross-classifies with respect to many physical properties of its members. The
supervenience requirement is satisfied in the following sense: by creating the
entire physical system, God had done enough to create the use for the kind
term. He has also done enough to create a physical relation, with chairs the
things in its domain—the things disposed to elicit and sustain the chair verdict
from us.

VII. Psychological States

With this much understanding of physical states behind us, the task of this
paper is nearly finished. But I shall end by saying a little about the other end
of the relationship.
The subordinate clauses in prepositional attitude psychology mention con-
tent; the object of the demonstrative is a thought or proposition, not a sen-
tence, nor words nor utterances.26 Folk psychology is quite silent over imple-
mentation. It makes no claim that there is some 'level' of explanation of our
thinking that fruitfully relates us to sentences or words. It is therefore not
affected by discoveries, such as seem likely to come in the wake of the suc-
cesses of parallel distributed processing architecture for recognitional and
other systems, that there is no fruitful level of explanation that sees such
things as involved with rules, or words, or syntactically complex instructions
and subcognitive processing. Just as a parallel distributed processing network
can 'recognize' a sound or a pattern without anything that ought to be called
computation, so, perhaps, we can think that our friend is coming down the
street equally without literal computation. There is no a priori reason to
suppose that because the skill is great, the mechanisms explaining it involve
real, but quick and unconscious, cognitive processes. On the other hand,
there is no a priori reason against this: the fruitfulness of such a level of
analysis is a matter of scientific, not philosophical, speculation.
Will psychology, granted this much independence, ever make a cognitive
science? Some see the hostility to treating psychological states as natural
objects as implicit in Frege's antipsychologism. It is certainly present in Col-
lingwood, in whom the capacity of a thinker for self-critical change of his own
beliefs and desires blocked the pretensions of psychology, or anything else, to
be a 'science' of thought. 27 Similarly Putnam argues that the business of
26. Of course, there is a huge literature trying to deny this. Some arguments are canvassed in
my 'The Identity of Propositions', in Blackburn, ed., Meaning, Reference and Necessity (Cam-
bridge: Cambridge University Press, 1975). Paul Churchland has been active in allowing that
intentional predicates can enter causal relations although the relata are abstract, by drawing the
parallel with 'numerical attitudes' (e.g., Matter and Consciousness, op. cit, 2nd ed. [Cambridge,
Mass.: Bradford Books, 1988], chapters 3, 4). Contrast Stich, ibid., p. 243.
27. R. G. Collingwood, An Introduction to Metaphysics (Oxford: Oxford University Press,
1940), chapters XI and XII. I discuss this further in Theory, Observation and Drama' in Mind
and Language, October 1992.
 Losing Your Mind: Physics, Identity, and Folk Burglar Prevention 253

attributing a mental state, with a certain content, to another always involves

issues of judgement seamlessly bringing to bear any of a formless class of
cultural, historical, or hermeneutic pressures, and the prospects for the out-
come of such a process being a scientifically respectable predicate seem very
poor.28 Allied to this are the normativity or rationality constraints governing
psychological attribution and the resulting suspicion that we have an essen-
tially dramatic, rather than descriptive, activity. But the Catholicism that I
have been urging enables physical respectability to gobble up this too: in
principle, a norm-wielding system is not more invisible to physics than a chair-
detecting one or a money-using one.
If we make the role-realizer identity, we can say that redescription of states
under predicates that reveal their lawlike connections makes good any preced-
ing causal explanation: it is the fund that backs up what is otherwise only a
promise. It is essential to this idea of 'making good' that the thing described in
the initial, lawless way is the thing that can also be described in the later,
lawlike way: this is the model common to Davidson and Lewis. Sometimes
this is plausible. The stone balances the scale, not because there is a law that
stones do, but because the stone is the thing of mass x at distance d from the
fulcrum. But with unnecessary identities out of the way, there is no analogy to
this in the case of psychological explanation. We can be agnostic about
whether psychological explanations can aspire to the condition of laws (even if
they do not for us, it is hard to see why they could not for a more simple
animal). Whether there is psychological law or not, no redescription of mental
states that succeeds in introducing a lawlike connection missing from psychol-
ogy would make good the psychological explanation. In other words, all we
would get would be something citing a different fact about me or my neural
states. We might find that 'as I believe' or 'when I believe' or 'upon coming to
believe' that p, such-and-such changes happen, and they are lawlikely corre-
lated with others. But we will not thereby get a new lawlike correlation
between 'a state that is the mental state' of coming to believe that p, and the
other changes. Such a filling out changes the subject. It substitutes mention of
something different.
It will be said again that I want it both ways: I want both to stress the
separateness of role states (role features) and to award them the causal acco-
lade that (it will be said) properly belongs only to realizer states. I reply that it
is the opposition that wants it both ways: both to preserve the causal power
belonging to the presence of features wide and small, and to succumb to the
Tractarian vision, by insisting that the real engine of change is the ir-
redeemably particular configuration of things. And this, I have argued, ren-
ders it invisible to physics, and indeed to all human thought.
Folk psychology is a wonderful thing. It is no wonder that we are baffled
by the prodigious knowledge and abstraction, the sheer instrumentation, that
enables us to look through the particular features of ourselves and others that
might otherwise get in the way, and type us all by our beliefs and desires. The

28. Putnam, op. cit., passim.

254 Mind and Matter

difficulty lies in seizing and sustaining the idea that in so doing we are identify-
ing real features whose instantiation is causally effective in producing events
in the world. In this essay I have not addressed all the obstacles to doing this,
but I hope I have removed some.
 14
Filling in Space

Why do people think that dispositions must have categorical grounds underly-
ing them? Well, the clock tells the time because there is such-and-such an
arrangement of little bits inside it; Sandy barks because her vocal chords
vibrate; the light glows because electrons whizz around in its filament. The
explanations are excellent, but do they illustrate the doctrine?
They do if they bring us to categorical grounds that indeed underlie the
dispositions. But they do not if they bring us to something else. For instance,
they might only bring us to a point where the possession of some disposition
by something is explained by the possession of other dispositions by the same
or different things. Or they might bring us only to the instancing of a power
(disposition) at some region of space explained by the instancing of some
other power at some related region of space.1
When we think of categorical grounds, we are apt to think of a spatial
configuration of things—hard, massy, shaped things resisting penetration and
displacement by others of their kind. But the categorical credentials of any
item in this list are poor. Resistance is par excellence dispositional; extension
is of use, as Leibniz insisted, only if there is some other property whose
instancing defines the boundaries: hardness goes with resistance, and mass is
knowable only by its dynamical effects. Turn up the magnification and we find
things like an electrical charge at a point, or rather varying over a region, but
the magnitude of a field at a region is known only through its effect on other
things in spatial relations to that region. A region with charge is very different
from a region without: perhaps different enough to explain all we could ever
know about nature. It differs precisely in its dispositions or powers. But
science finds only dispositional properties, all the way down.
Evans talked of a prejudice against bare ungrounded dispositionality,
'equally offended by the idea of two places alike in what occupies them
between visits, yet of which one is such that if one goes to it, one will have
certain experiences, and the other is not'.2 He wanted as well a 'relatively
1. To avoid clutter I should say that from now on I shall talk indifferently of powers,
dispositions, and 'counterfactuals' to describe the features on the noncategorical side of the fence;
it is their contrast with the other side that matters, not possible differences between them.
2. Gareth Evans, 'Things Without the Mind', in Philosophical Subjects, ed. Zak van
Straaten (Oxford: Clarendon Press, 1980), p. 102.

255
256 Mind and Matter

abiding property' that fills out the intervals in the exercise of dispositions.
Physics can give him this only in a backhanded way that will not satisfy the
demand. An electrical field can abide, certainly, but that means just that there
is a period of time over which various counterfactuals are true. It does not
give us reason to think of a different property, quietly persisting, as it were,
even when the disposition is not exercised.3
Perhaps I have been unduly verificationist. True, you might say, things like
fields and masses are known by their effects on other things. But this should
not prevent us thinking of them as in themselves categorical. There will be a
categorical ground, G, for the (multitrack) disposition D whereby we know of
mass or charge. It will be in virtue of the instancing of such a G that an object
has the mass that it does, or a region of space the charge. When we think this,
I believe we surreptitiously imagine an improvement in science that would
enable us to identify G: a new theory about something true of charged regions
of space, for example, rather like the molecular theory of gases. But this road
leads only to the same place. Just as the molecular theory gives us only things
with dispositions, so any conceivable improvement in science will give us only
a better pattern of dispositions and powers. That's the way physics works.
Is it the way it has to work? I believe so. A quick route to this conclusion is
to see the theoretical terms of a science as defined functionally, in terms of
their place in a network of laws.4 A slower route is to reflect on what is needed
from physical thinking. What is needed is the use of concepts—energy, tem-
perature, entropy—that cover changes of state, permitting the formulation of
conservation laws. Such concepts in effect tell us what is the same about a
changing system, in terms precisely of its powers and dispositions.5
G will remain, therefore, entirely beyond our ken, a something-we-know-
not-what identified only by the powers and dispositions it supports. And then
the possibilities multiply. Perhaps there is a possible world just like ours, not
only in surface appearance, but in all that physics could ever discover, in
which the dispositions have a different categorical ground, G'. Perhaps in our
own world G' supports dispositions on Mondays and Wednesdays, while G
supports them on the other days. But the real problem arises with the nature
of the underlying and support that G provides for D. Is it logically necessary
that G support D? Presumably not—what could logic have to say about it?
And since we know nothing of G, it can hardly be a priori in any other way
that it does what it does. Presumably then there is a law whereby G supports
D, and this law imputes a power to G. So it ought to need a separate categori-
cal ground, G*, it being in virtue of G* that G gives rise to D in the worlds
that obey this law. But then the power of G*, to bring it about that G gives rise
3. Cf. Strawson's reply to Evans, van Straaten, op. cit.,, p. 280. 'If it seems true of the
sensory properties in general that they all dissolve together, under reflective pressure, into
dispositions, this seems even more certainly true of the "physical" properties which are held to
constitute their categorical base'.
4. David Lewis, 'How to Define Theoretical Terms', in his Philosophical Papers, volume I
(New York: Oxford University Press, 1983).
5. I tell this story at greater length in essay 13.
 Filling in Space 257

to D, will itself need a ground, and so forever. To stop the regress we need a
brute or bare power without a categorical ground: better in that case not to
insist on grounds in the first place.
Are we in danger of proving too much? It seems as though we have excised
categorical properties from nature altogether, leaving only features that, as
Russell said, are each other's washing.6 The problem is very clear if we use a
possible-worlds analysis of counterfactuals. To conceive of all the truths about
a world as dispositional is to suppose that a world is entirely described by what
is true at neighbouring worlds. And since our argument was a priori, these
truths in turn vanish into truths about yet other neighbouring worlds, and the
result is that there is no truth anywhere.7 Here is a good analogy. Early
philosophers such as Whitehead and Collingwood, reflecting on modern phys-
ics, saw that it resolved substance into function and drew the consequence
that there is no such thing as a state of nature at a literal timeslice: processes
take time. 'There is no nature at an instant'.8 The present problem is that
processes take possibilities as well as actuality, so there is no nature at an
actual world.
In Lewis's exploration of these issues, categoricity comes in the 'Humean
mosaic' or pattern of 'perfectly natural intrinsic properties which need nothing
bigger than a point at which to be instantiated'.9 It seems as though we need
them, but it now also seems as though we cannot have them—our best physi-
cal understanding of the world gives us no conception of what they might be.
But the mention of Hume is suggestive. We can think of the Humean mosaic
in experimental terms: a colour here, a tactile sensation there, a sound some-
where else. Categoricity in fact comes with the subjective view: there is noth-
ing dispositional, to the subject, in the onset of a pain or a flash in the visual
field. Such events come displayed to us as bare, monadic changes in particular
elements of experience. In this perspective a change in perceived colour is as
categorical as a change in shape or a twinge of toothache, even if, from the
objective standpoint, 'all that goes on' when such changes occur is that a
change of functional (dispositional) state arises, the subject being disposed to
act and think differently as a consequence of changes in the dispositions of
surrounding things.
The trouble now is that such events, conceived of as categorical, play no
role in a scientific understanding of the world; they certainly do not serve to
6. The Analysis of Matter (London: Kegan Paul, 1927), p. 325. 'There are many possible
ways of turning some things hitherto regarded as "real" into mere laws concerning the other
things. Obviously there must be a limit to this process, or else all the things in the world will
merely be each other's washing'.
7. I have subsequently found this point made in very similar terms by Howard Robinson,
Matter and Sense (Cambridge: Cambridge University Press, 1982), chapter 9. Robinson uses it to
attack any conception of matter, whereas I leave the issue in Hume's hands rather than Berkeley's.
8. See for example, R. G. Collingwood, The Idea of Nature (Oxford: Oxford University
Press, 1945), Introduction, sections 4 and 5. The saying is quoted from Whitehead, Nature and
Life (Cambridge: Cambridge University Press, 1934), p. 48.
9. David Lewis, Philosophical Papers, volume II (New York: Oxford University Press,
1986), Introduction, p. x.
258 Mind and Matter
ground anything. The question remains whether we can live with a concept of
objectivity that leaves us only powers without end, and escape from the
discomfort this causes by somehow retreating to a subjective point of view.
Strawson counsels that we can, self-consciously shifting perspectives to avoid
the contradiction of thinking of the same event as both categorical and disposi-
tional.10 But the problem remains that this gives us no help in understanding
what, except counterfactuals, is true of the objective order of nature, unless,
heroically, we see that order as a kind of construct from the categorical point-
instances of properties available to the subjective view—a kind of neutral
monism. It almost seems that carelessness and inattention alone afford a
remedy—the remedy of course of allowing ourselves to have any idea at all of
what could fill in space.

10. Strawson, in van Straaten, op. cit., pp. 280-81.

 Index

act adequacy premise, 165 colours, 57,178, 207

Alexander, S., 145 conditionalizing, 81, 83-91
analysis, 5, 54 conditionals, 17,105
analyticity, 136 consequentialism, 164
Anscombe, E., 100 n. consistency and desire, 189-97
Armstrong, D., 42, 59, 93, 98,179 n. convention, 61, 73, 200, 202
as-if, 55, 155,216 Copeland, B., 183 n.
Austin, J. L., 44 n. Cornell realism, 147, 180, 203
Axelrod, R., 168 n. correspondence, 76, 77,119,167
Ayer, A. I, 54, 59 Craig, E. J, 60, 66-70, 74, 94 n., 99, 102
criteria, 117, 232
Baldwin,!"., 126
Bare Humean, 106 Davidson, D.
beauty, 161 and epiphenomenalism, 246 n.
Bennett, J., 27 and mind-brain identity, 206, 244, 253
bent rules, 218 and supervenience, 139,140,146
Berkeley, G., 101-2, 104 n. determinism, 88
Bernoulli, J., 93 Dewey, J., 5
bicycle locks, 241 dispositions, 9, 255
bivalence, 23-28 and rules, 219
Block, N., 229 n., 244-48 Dretske,F.,38
Boghossian, P., 57 n. Dummett, M.,5,226
Bostock, D., 143 n. and bivalence, 23-27
Bradley, F. H., 247 n. and realism, 43-48,181
Branch, G., 146 and semantics, 183 n.
and truth, 111-12
Candlish, S.,247n. Dutch Book, 82
Carnap, R., 62 n., 78 Dworkin, R.,26, 208
Cassam, Q.,173n.
Castaneda, H., 189 n. eliminativism, 229 n., 240
causation, 94-107 Eliot, G., 153
cautious man, 61, 64, 66 empiricism, 106
chance, 78, 80-93 EnS, B., 235, 237
Churchland, Patricia, 230-34 epiphenomenalism, 245
Churchland, Paul, 231, 252 n. error theory, 4, 56, 149-50, 208, 214 n.,
Cohen, L. J, 88 215
Collingwood, R. G., 252, 257 Euthyphro dilemma, 54, 175

259
260 Index
Evans, G., 157 n., 255 induction, 84-87
event identity, 206 institutional facts, 120
explanation, 30, 198 internal realism, 31
Isaiah, 229
facts, 168, 216
Field, H., 183 n. Jackson, F., 205, 244
Fine, A., 166 n. Jeffreys, R., 91
focus imaginarius, 23 John the Baptist, 229
Fodor, J., 231 n., 250 justice, 198
Foot, P., 208
Forbes, G., 55, 66 Kanger, S.,183n., 193
frameworks, 142 Kant, I., 18, 33 n., 43, 71, 166, 185, 193
Frege, G., 125, 152,252 Kemp-Smith, N., 103 n.
Flicker, E., 143 n. Keynes, J. M., 88
Kim, J., 134, 145, 246 n.
Geach,R, 19, 152, 185, 188 kingdom of ends, 193
Gettier, E.,37 Klagge, I, 146
Gibbard, A.,165n., 195 n. knowledge, 35, 89
God, 171 Kraut, R., 239 n.
Goldman, A., 39 Kripke, S., 11, 64 n., 73, 200 n., 213-28
Goodman's paradox, 221 Kyburg,H.,78
Grice, H. P.,5n.
gridlock, 93 Leibniz, G., 64, 255
grounds of dispositions, 256 Lewis, D.
and identity theory, 229-54, 256-57
Hacking, I., 83 and modal realism, 7, 57, 73
Hale, R., 73, 189 n., 195 n. and probability, 82, 91-93
Hardin, C.,207n. and values, 201,206
Hare, R. M., 152, 167, 202, 208 Lewy, C., 114, 130
Haugeland, I, 233 n., 239 limit of inquiry, 21, 89
Hintikka, J.,193 Locke, I, 57, 170 n.
Hobbes,T., 163 logic, 183-97
holism, 28, 41, 150 Logue, J., 91
Honderich, T., 246 n. Lovibond, S., 170 n.
Hume, D.
and cause 95-107, 226 Mackie, J., 20, 58, 149-57, 176, 208
and projection 5, 55-56, 75, 150,163, Martians, 174
179 Matilda, 38
and scepticism, 176 McDowell, I, 43 n., 154 n., 157-58,
and standards 20, 29, 78-79 162, 169, 174
and the sceptical solution, 214-17 McFetridge, I., 73
Hurley, S., 146-47 McGinn, C., 47 n., 160 n.
hypothetical imperatives, 177 Mcintyre, A., 208
Menzies, P., 246 n.
ideals, 194-96 Millican, P., 100 n.
identity theory, 239-53 Millikan,R., 166 n.
images, 17 Milton, I, 199 n.
imagination, 66 Mirv/Pirv principle, 38-41
indeterminacy, 22, 29, 33 mixed worlds, 134
indirect contexts, 124 modus ponens, 182
 Index 261

mood, 186-88 Ramsey, F. P., 7,16,18, 27, 54, 75, 80,

Moore, G. E., 116,121,130,180 91
motive consequentialism, 165 rationality, 82
Rawls, J.,199, 205
Nagel, T.,58,176 realism, moral, 111-12,166, 171, 180-
natural kinds, 203, 250 83, 203
naturalism, 35, 70, 71, 166, 208 reduction of theories, 234
naturalized epistemology, 62 redundancy theory of truth, 16, 217, 227
necessity, 52-74, 106, 179 relativism, 177
negation, 183 reliabilism, 37, 39, 51
Neurath's boat, 79 representation, 105,106
Newton, I., 101 response-dependence, 9
Nietzsche, E, 22, 185-87 Robinson, H., 257
nomological danglers, 145, 233 Rosenberg, J., 229 n.
Nozick, R., 38, 204 Rowlands, M., 229 n.
rule following, 213-28
rule utilitarianism, 165
open question argument, 186
Russell, B., 152,172, 257

Pappas, G.,279n. Satan, 199, 201

parallel distributed processing, 252 sceptical realism, 94-107
Peacocke, C.,224, 225 scepticism, 36,129, 213
Peirce, C.S.,5, 21 Schiffer, S., 145
Pendlebury, M., 187 Schueler, G., 183-87
perception, 161 Scruton, R., 123 n.
Pettit, P., 92, 205, 229 n., 244 Searle, J. R.,120,125
physics, 229-58 secondary qualities, 159
Platts, M., 77, 170 n. self-respect, 154,155
plenitude, principle of, 135 Sellars, W.,21
Popper, K., 85 Skyrms, B., 91
positivism 94-98, 105 Smith, M., 92,143 n., 172, 191 n.
possible worlds, 54, 73, 130,131 Smith, P., 246 n.
Prawitz, D., 183 n. Smokier, J., 78
Price, A., 170 n. Socrates, 199
Price, H., 10 Sosa, E.,35, 246 n.
Principal Principle, 91-93 Spinoza, B., 55, 64
Prior, A., 196 Stich, S., 230 n., 231 n., 239 n., 252
prisoners' dilemma, 201 Stoutland, F.,246n.
private language, 225-28 straitjacket, 98-107
programme explanations, 205 Strawson, G., 94 n., 96, 98, 100
projectivism, 32, 55-60, 75,144, 152, Sturgeon, N., 198-209
167, 176 supervenience, 31, 114-48, 159
propositional reflection, 125, 185 Suppes,R, 82
Putnam, H., 5, 18-21, 48,158, 166 n.,
200, 249, 252 Talbot, M.,229n.
Tarski, A.,77
quasi-realism, 3-9, 15-34, 48, 53, 55, 77, Teller, P., 83
167, 206 temperature, 235-38
quietism, 76, 152, 163 thermodynamics, 236
Quine, W. V., 16 n., 61-67, 142 THI CK T ER MS , 11 , 201
262 Index

Thrasymachus, 199 vertigo, 98, 100

Toulmin, S., 128 von Mises, R., 88
tracking, 39
tractarian view of physics, 238 Weyl, H.,183n.
truth functors, 191 Whitehead, A.,257
twin earth, 203 Wiggins, D., 156-57, 169, 171-72,
176
Urmson, J. O., 124 Williams, B., 165, 176, 208
utilitarianism, 121; see also rule Winkler, K.,47n.
utilitarianism Wittgenstein, L., 6n., 11, 24, 51, 58, 60,
112, 174, 175
variable realization, 131, 234-38, 246 Wright, C., 6, 22 n., 60, 66, 70, 214 n.
Velleman, J.D.,57n. Wright, J., 94 n