Truth and Realism

Patrick Greenough (Editor),

Michael P. Lynch (Editor)

end p.i

Blank Page

end p.ii

Truth and Realism

CLARENDON PRESS · OXFORD
end p.iii

Great Clarendon Street, Oxford OX2 6DP

Oxford University Press is a department of the University of Oxford.

It furthers the University's objective of excellence in research, scholarship,
and education by publishing worldwide in

Oxford New York

Auckland Cape Town Dar es Salaam Hong Kong Karachi

Kuala Lumpur Madrid Melbourne Mexico City Nairobi
New Delhi Shanghai Taipei Toronto

With offices in

Argentina Austria Brazil Chile Czech Republic France Greece

Guatemala Hungary Italy Japan Poland Portugal Singapore
South Korea Switzerland Thailand Turkey Ukraine Vietnam

Oxford is a registered trademark of Oxford University Press

in the UK and in certain other countries

Published in the United States

by Oxford University Press Inc., New York

© the several contributors 2006

The moral rights of the author have been asserted

Database right Oxford University Press (maker)

First published 2006

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,
without the prior permission in writing of Oxford University Press,
or as expressly permitted by law, or under terms agreed with the appropriate
 reprographics rights organization. Enquiries concerning reproduction
outside the scope of the above should be sent to the Rights Department,
Oxford University Press, at the address above

You must not circulate this book in any other binding or cover
and you must impose this same condition on any acquirer

British Library Cataloguing in Publication Data

Data available

Library of Congress Cataloging in Publication Data

Data available

ISBN 0-19-928887-9 978-0-19-928887-8

ISBN 0-19-928888-7 (Pbk.) 978-0-19-928888-5 (Pbk.)

1 3 5 7 9 10 8 6 4 2

end p.iv

Contents
List of Contributors vii
Acknowledgments viii
Introduction 1
I Truth and Relativism
1. What is Relativism? 13
Paul Boghossian
2. Intuitionism, Realism, Relativism and
Rhubarb 38
Crispin Wright
3. Modelling the ‘Ordinary View’ 61
JC Beall
II Realism and Antirealism
4. Realism: What's Left? 77
Michael Williams
5. Scientific Realism 100
Michael Devitt
6. Scientific Realism as an Issue in
Semantics125
Christopher Gauker
7. Abundant Truth in an Austere World 137
Terry Horgan and Matja Potr
8. Context, Vagueness, and Ontology 162
Mark Richard
III Methodology and the Nature of the Debates
9. Must Do Better 177
Timothy Williamson
10. A World without Isms 188
Paul Horwich
11. Horwich's World 203
Marian David
end p.v
12. Intuitions and Truth 208
Ernest Sosa
13. Trusting Intuitions 227
Michael P. Lynch
14. Truth and Realism: Remarks at St
Andrews 239
Richard Rorty
Index 249
end p.vi

List of Contributors
JC Beall is Associate Professor of Philosophy at the University of Connecticut, and Arché Associate
Fellow, University of St Andrews.
Paul Boghossian is Professor of Philosophy at New York University.
Marian David is Professor of Philosophy at the University of Notre Dame.
Michael Devitt is Distinguished Professor of Philosophy at The Graduate Center, The City
University of New York.
Christopher Gauker is Professor of Philosophy at the University of Cincinnati.
Terry Horgan is Professor of Philosophy at the University of Arizona.
Paul Horwich is Professor of Philosophy at New York University.
Michael P. Lynch is Associate Professor of Philosophy at the University of Connecticut, and Arché
Associate Fellow, University of St Andrews.
Matja Potr is Full Professor and the Chair of Logic, Methodology and Philosophy of Cognition,
Department of Philosophy, The University of Ljubljana, Slovenia.
Mark Richard is Professor of Philosophy at Tufts University.
Richard Rorty is Professor of Comparative Literature and, by courtesy, of Philosophy at Stanford
University.
Ernest Sosa is Romeo Elton Professor of Natural Theology and Professor of Philosophy at Brown
University, and a Distinguished Visiting Professor at Rutgers University.
Michael Williams is Krieger-Eisenhower Professor in the Department of Philosophy, Johns Hopkins
University.
Timothy Williamson, FBA, FRSE, is Wykeham Professor of Logic at Oxford University.
Crispin Wright, FBA, is Professor of Logic and Metaphysics, Wardlaw Professor, and Director of
Arché, University of St Andrews, and Global Distinguished Professor at New York University.
end p.vii

Acknowledgements
The papers in this collection, with the exception of Paul Boghossian's paper ‘What is Relativism?’,
were all presented at the conference on Truth and Realism, held at the University of St Andrews,
Scotland, 17–20 June 2004.
For invaluable financial assistance with this conference we would like to thank: the Arts and
Humanities Research Council, the Arché Research Centre at the University of St Andrews, the
Department of Philosophy at the University of St Andrews, the British Academy, the Scots
Philosophical Club, the Mind Association, and the Analysis Trust.
We would like to give warm thanks to all the participants at the conference, and particularly,
the main speakers, Robert Brandom, Michael Devitt, Terry Horgan, Paul Horwich, John McDowell,
Ernest Sosa and Michael Williams, and Crispin Wright; the commentators, Dorothy Edgington,
Christopher Gauker, Mark Richard, Marian David, Jane Heal, Michael Lynch, John Hawthorne and JC
Beall; the conference commentators, Simon Blackburn, Paul Boghossian, Richard Rorty and
Timothy Williamson; the chairpersons, Tad Szubka, Sven Rosenkranz, Jennifer Hornsby, Brian
Weatherson, Kit Fine, Pascal Engel, Ralph Wedgwood and Alessandra Tanesini; and the invited
discussants, Bob Barnard, Paul Bloomfield, Andrew Cortens, Steven Hales, Henry Jackman, Max
Kölbel, Philip Percival, Barry Smith and James Woodbridge.
Thanks also to all those who spoke at the Postgraduate Truth and Realism Preconference: Michael
Devitt, Terry Horgan, Frank Jackson, James Ladyman, Michael Lynch, John McDowell, Panu
Raatikainen, Sven Rosenkranz and David Velleman.
Many people went out of their way to help in all sorts of ways in the organization of both
conferences. Special thanks go to Eline Busck and Nikolaj Pedersen for co-organizing the
Preconference. They were paragons of efficiency and propped both of us up throughout. The
following people were stalwart helpers: Elizabeth Barnes, Ross Cameron, Roy Cook, Tom
Cunningham, Philip Ebert, Sune Holm, Amy Hughes, Chris Kelp, Anne Manolakas, Mary McKay,
Paula Milne, Walter Pedriali, Simon Robertson, Marcus Rossberg, Markus Schlosser, Margarita
Stapleton, David Ward, Michael Weh and Robbie Williams.
Peter Momtchiloff of Oxford University Press proved to be patient, forgiving and immensely helpful.
Thanks, Peter.
end p.viii

Thanks also to Peter Clark, John Haldane and Crispin Wright for much invaluable advice and
support. Untold thanks to Terry Berthelot and Maxine Greenough who helped beyond measure.
Patrick Greenough
Michael P. Lynch
end p.ix

Blank Page

end p.x
 p p p p p y

Introduction

Is truth objective or relative? What exists independently of our minds? This book is about these two questions. The
essays in its pages variously defend and critique answers to each, grapple over the proper methodology for addressing
them, and wonder whether either question is worth pursuing. In so doing, they not only enter debates that have been
the focus of much recent philosopical and cultural interest, they carry on a long and esteemed tradition—for our two
questions have vexed almost every major philosopher, from Plato, to Kant to Wittgenstein.

The volume is organized into three parts: “Truth and Relativism”; “Realism and Antirealism”; and “Methodology and
the Nature of the Debates”. Given the interrelated nature of the topics, the distinction between the parts is of course
not precise, but the essays collected in the first two parts primarily address the book's two questions, while those in
the third part pull back and focus on the debates themselves, and our methods (or lack thereof ) for solving them.
Moreover, all three parts have a uniform structure. Each begins with a lead-off essay that frames the question under
discussion and provides a brief overview of the terrain. These essays are in no way merely “introductory”; rather,
they offer a sophisticated and opinionated take on the issues from a major figure in the field. Subsequent papers in
each part come in pairs, offering substantial positions and then critiques of those positions, as thus capturing the
dialectical nature of philosophical debate.

In what follows, we offer a brief, part by part overview of the volume's topics along with some remarks about the
papers in each part. Since the papers are mostly self-sustaining and require little insider knowledge, we don't offer
detailed elaborations or critiques of their individual arguments. Rather, our aim is to introduce the issues the papers
address.

Truth and Relativism

The question “what is truth?” is a classical philosophical conundrum, and the answer that truth is relative is both
familiar and ancient, going back to Protagoras'

end p.1

thought that man is the measure of all things, and finding a faint if confused echo in the first-year university student's
plaintive cry that “what is true for you isn't true for me.” The core idea behind alethic relativism is that propositions
are only true or false relative to some parameter, whether that parameter consists of the views of a particular person,
culture, form of life or conceptual scheme. Relativist views about truth continue to grip the contemporary imagination,
and are discussed not just in philosophy but throughout the academy, appearing, for example, at the center of the
debate over postmodernist literary theory.

Alethic relativism is opposed by alethic realism, according to which truth depends not on us and our beliefs, concepts,
and so on, but on the way the world actually is. As Aristotle put the thought: to say of what is, that it is, is true. Often
this remark is glossed as encapsulating the correspondence theory of truth. According to this theory a proposition is
true just when it corresponds to reality, where “correspondence” is understood variously in terms of structural
isomorphism between our thought and the world, or reduced to the referential relationships that obtain between the
elements of our thoughts and the objects those thoughts are about.

Jostling around with these views are other contenders for the title of the truth about truth. The coherence theory of
truth holds that beliefs are true just when they cohere with other (often a particular privileged set of ) beliefs.
Classical pragmatists, on the other hand, maintain that true beliefs are those that would be accepted at an ideal limit
of inquiry, or which would be practically useful in helping us meet our goals in the long run.

Finally, deflationary conceptions of truth reject all of the above options as misguided. According to deflationists, truth
has no nature to discover. We understand truth simply by grasping instances of the T-schema:

(TS): The proposition that p is true if and only if p.

Thus, the concept of truth doesn't pick out a robust relation between mind or world (relative or otherwise), but is
merely a logical device, useful because it allows us to make generalizations over potentially infinite numbers of
propositions (such as the generalization that every proposition is either true or false).

Paul Boghossian's lead-off essay delves into the nature of relativism. He begins by noting that most philosophers
believe that global relativism is straightforwardly incoherent. Global relativism, after all, is the idea that all truths are
relative, and that claim is itself only relatively true or it is absolutely true and therefore false. But as Boghossian says,
most philosophers also believe that one can coherently restrict relativism to a local domain. Thus various philosophers
have adopted relativism, for example, about morality or epistemic justification.

Boghossian's surprising argument is that this second assumption is unfounded. His argument rests on the claim that a
coherent relativism about a particular kind of facts K has these three ingredients: first, the relativist claims that there
are no absolute facts of kind K; second, she recommends that we stop asserting

end p 2
 p p p p p y

absolute propositions which purport to be about such facts and assert only relational or relative propositions; and
third, she offers a constraint on the values that the relativization parameter for those propositions and their truth-
conditions can take. Boghossian thinks there are coherent local relativisms about various kinds of facts (such as the
facts about motion) that have these three features, but he argues the kinds of facts that philosophers have been
interested in resist relativization in this way. Consequently, the prospects for a philosophically exciting relativism
about morality—to name just one example—are much dimmer than previously thought.

Crispin Wright's paper focuses on what he calls disputes of inclination, such as a dispute over whether something is
delicious or attractive or funny. What Wright sees as the “ordinary view” of such disputes is that they involve the
following three key elements: that the subjects can have conflicting incompatible attitudes, that no one is necessarily
mistaken, and yet that both sides can rationally maintain their position. Maintaining the ordinary view is what lies at
the root of relativism, Wright thinks. Moreover, disputes of this sort, and the ordinary view of them, arise in more
philosophically interesting cases, as well: over what is right or wrong, or what is justified or not, for example.
Therefore, he argues, understanding whether the ordinary view of these disputes is coherent, and what sort of theory
of truth it entails, will help us understand similar disputes in ethics or epistemology. After pointing out that the three
elements of the ordinary view are in manifest tension with one another, Wright canvasses and rejects several ways of
resolving this tension. He then suggests his own novel solution to the problem, one arising out of his position that
truth can itself come in different forms. If this is so, Wright argues, then one is free to understand truth in some
domains—such as the domains that give rise to disputes of inclination—as superassertibility, or sustained justification
without defeaters. According to Wright, suitably understood and applied, superassertibility can be invoked to explain a
defensible, albeit local relativism about truth.

Rather than critiquing Wright's view directly, JC Beall offers two other ways to understand or model the three
seemingly incompatible elements of the ordinary view of disputes of inclination. First, Beall argues, rather than
making “is true” relative, one can semantically descend and underwrite the ordinary view by adopting a relativism
about the non-semantic predicates involved in the dispute in question, such as “is delicious.” Beall wonders what
virtues Wright's relativism involving superassertibility could have that this view wouldn't. Leaving that issue open,
Beall then argues that Wright and other philosophers have overlooked the possibility that one could be a relativist and
a correspondence theorist of truth at once, by adopting what Beall calls the “polarity” view of facts or states of affairs.
The core thought is that a statement is true in virtue of the way the world is, but how the world is happens to be
relative to a person's attitudes. While stopping well short of endorsing the position, Beall suggests that its

end p 3

very possibility illustrates that the philosophical terrain is more complicated than previously thought.

Realism and Antirealism

The word “realism” is used in many different ways. As we've already seen, it can be used to characterize objective
views of truth. But it can also be used—and perhaps should primarily be used—to describe certain views about what
exists. In this, metaphysical, sense “realism” names a position which, roughly speaking, holds that most of what exists
in a certain domain does so mind-independently. Something exists mind-independently at some time when it would
continue to exist were there no minds. Following Michael Devitt's influential distinction, we can say that two of the
most common forms of realism are these:

Common-sense realism: most of the physical, “middle-sized dry goods” posited by common sense (trees,
mountains, cats, and cars, etc.) exist mind-independently.
Scientific realism: most of the unobservable entities posited by our best scientific theories (electrons, quarks,
etc.) exist mind-independently.

Of course, one can be a realist in other domains as well, and live debates occur over realism about moral properties,
mathematical objects like sets and numbers and more recherché objects like propositions and possible worlds. And one
can be a realist about some of these entities while being antirealist about others.

The mind-independence feature of realism means that any sort of realism is always opposed to idealism which holds,
classically, that everything that exists is either a mind or dependent on a mind. And realists within a given domain are
opposed to more local antirealisms, such as views which simply deny that entities of a given type exist, or claim that
they are, in an appropriately explicated sense, merely “fictional” in nature. Moreover, some philosophers—if by no
means all—also claim that realism is opposed to conceptual relativism, which holds (in relation to a particular domain)
that most of what exists only does so relative to the conceptual schemes humans employ to categorize the world.

Michael Williams begins his essay by asking two key questions: What is realism? And is there an issue here still worth
arguing about? Williams' preferred answer to the first question is minimal or deflationary in spirit: realism is simply
the view that most current common-sense and scientific objects exist and most claims about such objects are true.
Such a view, Williams argues, is realist enough: it allows one to avoid classic “antirealisms” (such as idealism), while
avoiding the metaphysical extravagancies of the correspondence theory of truth and “mind-independent” existence.
Perhaps not surprisingly, Williams then answers negatively to his second question. For him, what remains of the
realist issue is the question of naturalism; we should be concerned less with what exists

end p.4
 p p p p p y

mind-independently than with the question of whether (and in what sense) all that exists is natural or physical in
character.

Williams situates his discussion within what he sees as a division between two major camps of philosophers: those like
himself whom he calls the neo-pragmatists, and those like Michael Devitt or Philip Kitcher whom he calls the
neo-Cartesians. While both camps reject skepticism and classical robust foundationalism, the neo-Cartesian takes a
“metaphysics first” approach and takes realism about truth seriously, while the neo-pragmatist keeps Descartes'
epistemological orientation (start reflection by wondering what and how you know) and adopts a deflationary theory
of truth.

Michael Devitt is a realist if there ever was one, both about common sense and about the unobservable entities of
scientific theory. Here, Devitt aims to defend realism of the second type. Devitt characteristically sees scientific
realism as a metaphysical, rather than a semantic doctrine. It is the thesis that most of the unobservable entities
posited by our best scientific theories actually exist, and exist independent of our minds, conceptual schemes, and the
like. Devitt discusses and defends one basic type of argument for, and two arguments against, this sort of realism. The
basic type of argument that Devitt takes to support realism is a version of what is sometimes called the “success
argument.” The root idea is that the best explanation of a scientific theory's success is that the unobservables posited
by the theory exist and have roughly the properties specified by the theory. Such an argument, Devitt points out,
requires the use of abduction, but as long as the realist is granted that, he argues, she can make a strong case for her
view. Devitt then goes on to defend scientific realism against two prominent objections: the classic “undetermination
argument”—which he sees as a non-starter, and the “pessimistic induction” argument, which claims that on the basis
of past theoretic failures, we have no reason to think that the entities posited by our current scientific theories
actually exist. While admitting this is a more difficult argument to defeat, Devitt suggests its force is sufficiently
blunted by what he takes to be the plausible view that our ability to chart the unobservables has improved.

Christopher Gauker takes issue with Devitt's core assumption that the debate over scientific realism is primarily about
what exists as opposed to how we represent what exists. As Gauker's title suggests, he argues that a debate over
scientific realism is primarily a debate over deep semantic issues—such as the plausibility of the causal theory of
reference—a theory which Devitt accepts and Gauker rejects. Moreover, Gauker questions Devitt's core view that
unobservables posited by a theory exist just when the theory has succeeded in giving good explanations. Rather,
Gauker suggests, it often works the other way around: theories are said to give good explanations just when it is
found that their posits exist.

Terry Horgan and Matja Potr 's paper intentionally defies easy characterization along the usual realist and antirealist
lines. The position they defend straightforwardly denies common sense and scientific realism as defined above:

end p 5

it claims that most of the posits of common sense and science simply do not exist. Nonetheless they do not deny that
there is a mind-independent world—it simply contains much less than we normally think it does. Thus their name for
their position: austere metaphysical realism. Horgan and Potr , however, conjoin their metaphysical position with a
position on truth, one which affirms that despite the fact that common sense posits entities that don't exist, its claims
are nonetheless perfectly capable of being true. According to what they call contextual semantics, truth is correct
affirmability—and what such affirmability amounts to can vary depending on the context. For most of the statements
we make, their semantic correctness consists in what the authors call their indirect correspondence with the world. A
statement can indirectly correspond to the world (and hence be true) even if the ultimately correct ontology fails to
contain the objects that statement purportedly is about. The overall position is a novel one: combining an austere
ontology with a robust contextualist account of truth, which as Mark Richard points out, bears certain affinities with
contextualist accounts of knowledge.

Richard criticizes Horgan and Potr 's position along several fronts. A principal criticism is this. Horgan and Potr 's
theory of contextual semantics, Richard argues, requires that we are cognizant in some way of shifts in semantic
context, which allows us to subtly interpret utterances as true or false depending on our evaluation of which is the
appropriate context at the moment. But if so, then we should be sensitive to the possibility that some claims of the
form that there are Fs are true even when the ultimate “real” ontology would reject Fs as nonexistent. Richard takes
it to be implausible that we are sensitive to such a possibility. Richard ends his paper with a reflection on,
appropriately enough, a topic that brings us to the next section: the methodology of the realism debate.

Methodology and the Nature of the Debates

When debates get entrenched, the opposing sides dig in, and minefields are laid, it is always wise to pull back and
take a look at the issue from afar. Is the issue really worth pursuing? Is all the effort—intellectual or otherwise
—justified by a noble goal or wasted on a dead end? And if the dispute is a real one, then how might we better pursue
it? It is wise to take a hard look, in other words, at both the nature of the debate and one's methodologies for
prosecuting it.

This advice informs the topic of the third section of this volume. Not surprisingly, given the age of the debates over
truth and realism, meta-philosophical and methodological questions are an increasingly prominent aspect of the
current philosophical landscape with regard to these issues. It is also, again not surprisingly, where many of the
 p p p p p y

deepest disagreements emerge.

end p 6

As with so many philosophical debates, disputes over truth and realism and second-order discussions about these
disputes and the best methodology for pursuing them are constantly intertwined. The contents of this section reflect
this fact—the authors of these essays not only argue over methodology and so on, they have their own realist or
antirealist agendas. Reflecting this fact, as well as the importance of these debates, we've chosen to include in this
section both a “lead-off” and a “capstone” paper, each representing radically different perspectives not only on realism
and truth, but on the nature of the debates over these topics.

Timothy Williamson begins his lead-off paper by reminding us that while the fruitfulness of many philosophical
debates has been called into question, those debates have often resulted in significant intellectual advances. This,
Williamson argues, is, to some extent, the case with debates over truth and realism, issues about which we know
more now than we did forty years ago. The reason we know more, Williamson suggests, is that we have made
important strides in articulating our philosophical intuitions with extreme logical precision. Nonetheless, in
Williamson's eyes, we can and must do much better in this regard. In his view, we must adopt a methodology largely
inspired by mathematics, one that prizes clarity, rigor, and open-eyed reflection on the sorts of constraints—including
logical constraints—that should be held to fix the contours of the debate. Only by doing so, Williamson argues, can we
hope to make philosophical progress.

Paul Horwich's paper, in stark contrast, amounts to something of a frontal assault on the very idea of a significant
debate between realists and antirealists. For Horwich, philosophical progress on these matters is not possible because
there is no meaningful debate to be had. In Horwich's view, disputes over whether a fact “really” exists or not emerge
out of the desire to avoid allegedly “weird” facts—like those about morality, mathematics, or art. So-called weird facts
are so, Horwich claims, only because they seem so different from the facts with which we are more familiar in our
everyday life—facts about cats, cars, kettles, and cradles, and so on. As a result of this dissimilarity, we are tempted
into constructing theories which explain the “nature” of such facts, or somehow explain them away—even while
allowing that our talk about such facts is in some sense legitimate. But in Horwich's view, all such attempts fail
because they cannot explain a distinction they need: the distinction between a fact being real and really REAL. Thus
Horwich suggests we either abandon or re-cast these debates, and give up on the idea that it makes sense to worry
over the nature of reality as it is in itself.

In his response to Horwich, Marian David argues first that Horwich has failed to make the case that he isn't actually
taking sides in the very debate that he rejects as meaningful. In particular, David takes Horwich to be committed to

end p.7

a form of non-reductive realism about all the “weird” sorts of facts mentioned above. Second, David claims that
Horwich's arguments for his non-reductive realism are not compelling. In his view, Horwich moves too quickly from a
point about language and concepts to a point about the world and the facts it contains.

Those who doubt the seriousness of debates over truth and realism often point out that such disputes come down to
“conflicts of intuition.” A realist attempts to refute an antirealist position by showing that it conflicts with the intuition
that if there were no minds, there would still be chairs; while another maintains that the opposition unintuitively
denies the law of excluded middle, and so on. In his contribution, Ernest Sosa defends the appeal to intuition in
philosophical debate. Sosa articulates and defends an account of intuitions according to which, roughly, to intuit is to
be inclined, on the basis of a reliable ability, to believe a modally strong or self-presenting proposition without appeal
to the usual sorts of evidence. As such, intuitions can be gainfully used as evidence for and against philosophical
positions and philosophers should therefore not shy away from their use of intuitions as a methodological strategy.

Although agreeing with Sosa that intuitions are useful and perhaps indispensable philosophical tools for solving
debates over realism, truth, and the like, Michael Lynch questions Sosa's account along several lines. He argues, for
example, that Sosa's account of intuitions is too limited—that it can't account for the wide variety of propositions
(including contingent propositions) that philosophers frequently call “intuitions.” Moreover, he worries that Sosa's
account requires the existence of a distinct “faculty of intuition” over and above the more usual faculties, including
our conceptual faculties. Finally, Lynch suggests an alternative, “minimalist,” model of intuitions, one according to
which an intuition is merely something one finds believable without knowing why one finds it believable. On this view,
intuitions may indeed be the result of reliable belief-forming abilities, but those abilities may be nothing more
mysterious than a conjunction of the usual sources of our beliefs.

Truth and Realism ends, fittingly enough, with a paper by a philosopher who has been at the center of these debates
for over thirty years. In Richard Rorty's summary paper, he comments on several of the papers included in the
volume, dividing them between those like Williams, Horwich and himself who see debates over what's ultimately real
or true to be misguided, and those like Williamson, Devitt, Horgan and Sosa who see the traditional questions as
having important value and significant answers. In Rorty's view, the lesson of the last millennia of philosophy is not
that we must do better at answering the questions we have, as Williamson suggests, but rather that we must come up
with different questions.

So what is the state of the debates over realism and truth today? Are they vigorously moving forward, opening new
 p p p p p y

theoretical doors, as Wright, Boghossian,

end p 8

Williamson, Horgan and others suggest? Or are such debates more like doors to overly cluttered rooms that should be
shut (and perhaps boarded over) as Williams, Horwich and Rorty might be seen to suggest? These are issues for the
reader to decide for him or herself. We add only a modest plea for a theoretical modus vivendi and a simple reminder:
let us give new answers to old questions, and let us ask new questions in place of old; but let us not fool ourselves
into thinking that there is always a clear dividing line between them. A thousand flowers can bloom in a garden of old
soil.

Patrick Greenough

University of St Andrews

Michael P. Lynch

University of Connecticut

end p.9

end p.10

This page intentionally left blank.

 p p p p p y

Part I Truth and Relativism

end p.11

end p.12

1 What is Relativism? *
For comments on an earlier draft, I am grateful to Kit Fine, Paul Horwich, Stephen Schiffer, Crispin
Wright and the participants in the Summer School in Analytic Philosophy in Florence 2004.

Introduction
Few philosophers have been tempted to be relativists about absolutely everything—although, we are told, there have
been some notable exceptions (Protagoras).

Many philosophers, however, have been tempted to be relativists about specific domains of discourse, especially about
those domains that have a normative character. Gilbert Harman, for example, has defended a relativistic view of
morality, Richard Rorty a relativistic view of epistemic justification, and Crispin Wright a relativistic view of judgments
1
of taste.
1
See Gilbert Harman and J. J. Thomson, Moral Relativism and Moral Objectivity (Oxford: Blackwell, 1996); Richard Rorty, Philosophy and the Mirror of

Nature (Princeton: Princeton University Press, 1979); Crispin Wright, ‘Intuitionism, Realism, Relativism and Rhubarb,’ this volume, Ch. 2.

But what exactly is it to be a relativist about a given domain of discourse?

The term “relativism” has, of course, been used in a bewildering variety of senses and it is not my aim to discuss each
and every one of those senses here. Rather, what interests me is the notion that is characterized by the following core
idea: the relativist about a given domain, D, purports to have discovered that the truths of D involve an unexpected
relation to a parameter.

This idea lies at the heart of the most important and successful relativistic theses ever proposed. Thus, Galileo
discovered that the truths about motion are unexpectedly relational in that the motion of an object is always relative
to a variable frame of reference. And Einstein discovered that the truths about mass are unexpectedly relational in
that the mass of an object is always relative to a variable frame of reference.

In this paper, I will develop a model for how such discoveries should be understood. And I will then consider to what
extent that model gives us a purchase on the sorts of relativistic theses—about morality, for example, or epistemic
justification—which have most interested philosophers.

end p.13

Relational Meanings
Galileo, we may suppose, discovered that truths about motion are unexpectedly relative to a frame of reference. What
does that mean?

A natural first thought is that the unexpected relationality is to be found in the propositions expressed by ordinary
motion sentences. On this view, whereas we might have been tempted to construe a sentence like:

(1) “The Earth moves”

as expressing a proposition involving the monadic concept of moving, after Galileo we know that it really expresses a
proposition involving the relational concept of moving relative to a frame of reference, with the variable frame of
reference being contextually supplied.

On this view, then, what Galileo discovered is that a form of contextualism is true for motion sentences: the
proposition expressed by a given motion sentence varies as a function of the context in which it is used.

Let us consider a case for which such a contextualist proposal is tailor-made, the case of the spatial relation expressed
by the phrase ‘. . . to the left of . . .’ How does this work?

Here is a first stab. Whenever someone utters a sentence of the form:

(2) “A is to the left of B”

he doesn't express the proposition:

(3) A is to the left of B

but rather the proposition:

(4) A is to the left of B relative to reference point F,

where F is some contextually supplied frame of reference.

Delicacy is called for in stating this view. We can't just say that whenever someone uses the sentence:
 p p p p p y

“S”

he ends up expressing not the proposition:

but rather the proposition:

S relative to F

for, then, any utterance of “S” will end up expressing the hopelessly infinitary proposition:

end p.14

. . . (((S relative to F) relative to F) relative to F) relative to F) . . .

The correct way to formulate the view we are after is to say, rather, that anyone uttering the sentence “A is to the
left of B” intends that utterance to be elliptical for the sentence:

(5) “A is to the left of B relative to reference point F.”

With that qualification, we then get the contextualism we were after.

Now, such a view is fine for the predicate ‘. . . to the left of . . . ’ because it is plausible that any competent user of
that predicate knows that it is in fact a three-place and not a two-place predicate. It is, therefore, not implausible to
say that any competent user of (2) would intend it to be elliptical for (5).

But it is not similarly plausible to claim that any competent user of the predicate “moves” knows that it expresses the
concept of a relation rather than the concept of a monadic property. As Harman correctly notes, it would be wrong to
claim that when people said:

“The Earth moves”

they intended their remark to be elliptical for:

“The Earth moves relative to a frame of reference F.”

Some perfectly competent possessors of the concept of motion were unaware that the only truths there are about
motion are relational ones and so had no reason to mean only the relational judgment.

And if that doesn't seem obvious in the case of motion, it should certainly seem obvious in the cases of mass and
simultaneity. It is simply not psychologically plausible to claim that, before Einstein, ordinary users of such sentences
as:

(6) “x has mass M”

and:

(7) “e1 is simultaneous with e2”

intended their utterances to be elliptical for, respectively:

(8) “x has mass M relative to reference frame F”

and

(9) “e1 is simultaneous with e2 relative to reference frame F”.

No one before Einstein knew that there are no absolute facts about mass and simultaneity but only relational facts
and so no one would have intended their utterances to be elliptical for the corresponding relational statements.

Intuitively, then, it would be wrong to construe the discovery of relativism about a given domain as the discovery that
the characteristic sentences of that

end p.15

domain express unexpectedly relational propositions. That construal wouldn't fit the classic cases drawn from physics.

Nor, intuitively, would it fit the cases that most interest philosophers, namely, the ones concerning morality or
epistemic justification. The relativist about morality, too, claims to have discovered some sort of unsuspected
relationism in the moral domain. But it would be utterly implausible to construe that as the claim that when ordinary
speakers assert such sentences as:

(10) “It is morally wrong of Paul to steal Mark's car.”

they intend their remarks to be elliptical for some relational sentence like:

(11) “In relation to moral code M, it is morally wrong of Paul to steal Mark's car.”

Relativism about morality, if true, is not vouchsafed by the relativistic intentions of the makers of moral judgments.
 p p p p p y

Harman: Relational Truth Conditions

If the discovery of relativism about a given domain is not a discovery about the kinds of propositions that are
expressed by the characteristic sentences of that domain, what sort of discovery is it, then?

Gilbert Harman has proposed an alternative answer: his idea is that we should take Einstein to have discovered
something not about the propositions expressed by the characteristic sentences of that domain but about their truth-
conditions. He says:

Einstein's relativistic conception of mass involves the following claim about the truth conditions of judgments of
mass:

For the purposes of assigning truth conditions, a judgment of the form, the mass of X is M has to be
understood as elliptical for a judgment of the form, in relation to spatio-temporal framework F the mass
of X is M.

Switching back to the example of motion, Harman's idea seems to be, then, that to be a relativist about motion is to
claim that, although the proposition expressed by (1) is just the disquoted non-relational proposition:

(12) The Earth moves

the truth-condition of that sentence is not just the disquoted truth-condition:

(13) The Earth moves

which attributes to the Earth the monadic property of moving, but is rather the relational truth-condition,

end p.16

(14) The Earth moves relative to the frame of reference F,

which attributes to the Earth the relational property of moving relative to a frame of reference F.

Put in somewhat other terms, Harman's view seems to be that although our concept of motion may just be the
concept of a non-relational property, the property denoted by that concept is the relational property of moving
relative to a reference frame.

On the way I am reading Harman, then, to discover that relativism is true about a given domain is to discover not that
the sentences of that domain express propositions that contain a reference to a hidden parameter, but to discover that
their truth-conditions do.

But can we really discover that although the sentence:

“p”

expresses:

p,

its truth-condition is not just the disquoted condition:

p,

but, rather, the relational condition:

p relative to F?

Do we really understand what it would be to abandon the platitude that:

The proposition that p is true if and only if p?

And whether or not we are able ultimately to make sense of this, is that really a plausible description of what Einstein
discovered about mass or Galileo about motion?

Harman's main argument for his view is that while it would be implausible to attribute to our ancestors a relational
meaning, it would be “mean-spirited” not to attribute to them the relational truth-condition. For if we didn't attribute
that relational truth-condition to them, we would end up accusing them of massive and systematic untruth—since
nothing has the monadic property of moving that they keep attributing to things—and that, apparently, would not be
very nice.

While I don't much like being called mean-spirited, I've never been sure how much weight to assign its avoidance in
philosophical theorizing. I mean: if I am willing to be mean-spirited, does that mean that I get to reject a relativism
about motion?

At any rate, I don't think the charge is fair: I think we can be perfectly nice to our ancestors even while we accuse
them of systematic error in certain domains.

end p.17

Notice, to begin with, that even Harman's account will have to attribute a serious error to our ancestors, for he is
 p p p p p y

going to have to say that they didn't know what the truth-conditions of their own thoughts were, that when they
stated those truth-conditions simply by disquoting, they said something false.

So there's no avoiding the imputation of some error and the only question is: which is the more plausible imputation?

To think that it must be better to impute to them false beliefs about the truth-conditions of their thoughts, rather than
just false beliefs about the world, is to endorse a peculiar version of the Principle of Charity as a constitutive
constraint on the attribution of truth-conditions, though not on the attribution of meaning.

Now, I have never been a big fan of the Principle of Charity; but certainly not when it is understood—as it has to be
here—to apply only to the truth-conditions and not to the meaning.

Even in its original version, as a constraint on the attribution of meaning, I never saw any justification for Charity
over the Principle of Humanity, according to which we are allowed to impute error in our interpretations of other
people provided those errors are rationally explicable. And, surely, Humanity is all that is needed for the purposes of
being nice, for it's consistent with our ancestors' having false views that those views were justified and even
ingenious.

But Humanity would certainly not give one any grounds for reconstruing the truth-conditions in the motion case: the
error involved—of not realizing the need for frames of reference—is certainly rationally explicable.

And, in any event, I certainly don't see the justification for applying Charity selectively, only to the truth-conditions
but not to the meaning. Think of what a peculiar result that would yield in a host of other cases. For example, our
ancestors also spoke of the soul departing the body. What they meant is that there is a non-physical substance which
leaves the body at the moment of bodily death. But, of course, we could, if we wanted, assign those remarks such
truth-conditions—involving the loss of consciousness and so forth—as would make them come out true. But that
would, of course, be absurd. Why should matters stand differently with motion or mass?

Factual Relativism
If the preceding considerations are correct, it is implausible to construe relativism about a given domain either as the
claim that the propositions of that domain are unexpectedly relational in character or as the claim that, while its
propositions aren't, its truth-conditions are.

In a sense, a difficulty with such construals of relativism should have been evident from the start, prior to a detailed
investigation of their prospects. The point is that it is hard to see how an adequate formulation of relativism about a
given domain could, in the first instance, be a claim about the contents of the

end p.18

sentences of that domain. Any such formulation, it seems to me, would leave open a possibility that any real
relativism should foreclose upon. I shall illustrate this point using the propositional construal, but similar remarks
apply to Harman's truth-conditional suggestion.

On the propositional construal, to say that moral relativism is true is to say that typical moral sentences like (10) do
not express such absolute propositions as:

(15) It is morally wrong of Paul to steal Mark's car

but, rather, such unexpectedly relational propositions as:

(16) In relation to moral code M, it is morally wrong of Paul to steal Mark's car.

Now, the trouble is that, on such a view, moral relativism is merely a view about what typical moral sentences mean.
It is merely a claim about the nature of the discourse as we have come to develop it. And that claim would appear to
leave it wide open that—out there—there are perfectly objective absolute facts about what ought and ought not to be
done, facts that our discourse, as we have come to develop it, fails to talk about, but which some other possible
discourse, that we have not yet developed, could talk about. In other words, this propositionalist construal of moral
relativism seems consistent with something that one would have expected any real relativism to foreclose upon,
namely, that there are objective moral facts out there waiting to be represented by our language and which we have
2
up to now somehow managed to overlook.
2
I heard Kit Fine make a similar point in an oral presentation on John MacFarlane's rather different formulation of relativism. There will be a problem, of

course, about how we are to express those missing facts, given the relativist's thesis about the meanings of ordinary moral terms, but there are obvious

strategies for getting around this difficulty.

A correct construal of relativism about a given domain, D, cannot locate the unexpected relationality in the contents
of D's sentences. It must locate it, rather, in the facts. Relativism cannot properly be seen as correcting our view of
what our sentences mean; it must rather be seen as correcting our view of what the facts are.

In other words, the relativist's project must be seen to be a reforming project, designed to convince us that we should
abandon the absolutist discourse we currently have in favor of a discourse which accommodates his conviction that
the only facts in the vicinity of that discourse are certain kinds of relational fact. Thus, in the case of motion, the
relativist must be seen as urging us to abandon talk of something's merely moving in favor of talk about its moving
 p p p p p y

relative to a variable frame of reference. And in the case of mass, he must be seen as urging us to abandon talk of
something's having mass in favor of talk about its having mass relative to a frame of reference. And so forth.

If we collect our various observations together, we get the following picture of a relativistic view of motion:

end p.19

(a) The sentence “The Earth moves” expresses the proposition The Earth moves which is true if and only if the Earth
has the monadic property of moving.
(b) Because nothing has—or can have—such a property, all such utterances are strictly speaking untrue.
(c) The closest truths in the vicinity are relational truths of the form:

x moves relative to frame of reference F.

Therefore,
(d) If our motion utterances are to have any prospect of being true, we should not make judgments of the form:

x moves

but only judgments of the form:

x moves relative to F.

Finally,
(e) No one of these frames is more correct for the purposes of determining the facts about motion than any of the
others.

This last clause, emphasizing that there is nothing that privileges one of these frames over any of the others, as far as
determining the facts about motion is concerned, is important because without it, it would be possible to satisfy
clauses (a) through (d) by supposing that the relativist is insisting on relativizing facts about motion to some
particular privileged frame of reference, say, the center of Earth.

So understood, then, a relativism about motion consists of three central ingredients: a metaphysical insight—that
there are no absolute facts of a certain kind but only certain kinds of related relational fact; a recommendation—that
we stop asserting the absolute propositions that report on those absolute facts but assert only the appropriate
relational propositions; and a constraint—on the values that the relativization parameter is allowed to assume (in the
case of the motion, there are no constraints).

Generalizing this picture, we can say that a relativism about a monadic property P is the view that:
(A) “x is P” expresses the proposition x is P which is true if and only if x has the monadic property expressed by “P.”
(B) Because nothing has (or can have) the property P, all such utterances are condemned to untruth.
(C) The closest truths in the vicinity are related relational truths of the form:

end p 20

x is P relative to F

where “F” names some appropriate parameter.

(D) If our P-utterances are to have any prospect of being true, we should not make judgments of the form:

x is P

but only those of the form:

x is P relative to F.

(E) There are the following constraints on the values that F may assume: . . .
The fewer constraints there are on F, the more extreme the relativism.

In light of its reforming nature, we may dub a relativism based on such a template “Replacement Relativism.”

Moral Relativism
Let me now turn to examining what sorts of theses Replacement Relativism leads to when it is applied to the sorts of
domains—morality for example, or epistemic justification—which have most interested philosophers. I will concentrate
on the moral case, but everything I say could easily be adapted to the epistemic case.

Applying the template just developed, we get the following Replacement view of moral relativism.
i. An ordinary assertion of “It is wrong of Paul to steal Mark's car” expresses the proposition that It would be wrong
of Paul to steal Mark's car, a proposition which is true if and only if Paul's stealing Mark's car has the monadic
property of being wrong.
ii. Because nothing has or can have the monadic property of being wrong, all such assertions are condemned to
untruth.
iii. The closest truths in the vicinity are related relational truths of the form:

It is wrong of Paul to steal Mark's car relative to F,

 p p p p p y

where “F” names some appropriate parameter.

(iv) If our moral assertions are to have any prospect of being true, we should not make judgments of the form:

It would be wrong of Paul to steal Mark's car.

but only those of the form:

It would be wrong of Paul to steal Mark's car relative to F.

end p 21

(v) There are the following constraints on the values that F may assume: . . .
In outline, anyway, this seems to capture fairly well what most moral relativists have intuitively wanted to hold. An
important question is: what should we take F to be? To what does the moral relativist propose to relativize ordinary
moral judgments?

The almost universal answer, given by friend and foe alike, is that the moral relativist proposes to relativize moral
judgments to moral codes or frameworks. As Harman puts it:

For the purposes of assigning truth conditions, a judgment of the form, it would be morally wrong of P to D, has
to be understood as elliptical for a judgment of the form, in relation to moral framework M, it would be morally
3
wrong of P to D. Similarly for other moral judgments.
3
Harman and Thomson, p. 4.

Adapted to my terminology, the idea is that we are to see the moral relativist as recommending that we give up as
untrue all propositions of the form:

It would be wrong of P to D

and replace them instead with propositions of the form:

In relation to moral framework M, it would be wrong of P to D

4
where “M” names some appropriately salient moral framework, typically the speaker's own.
4
An analogous view, with epistemic systems in place of moral frameworks—systems that specify how different kinds of information bear on the epistemic

justification of different kinds of belief—captures what many have wanted to call a relativism about justification.

Why is it so natural to construe the relativist as urging us to relativize moral judgments to moral codes or
frameworks?

The answer is that it is a natural elaboration of the central thought behind moral relativism—namely, that different
people bring different moral standards to bear on the morality of a given act and that there is no objectively choosing
between these different standards. A ‘mora code’ is just what codifies these standards. Hat is why it is the
relativization parameter of choice.

Moral Codes as Propositions

So far, so good. Now, however, we face the question: what could moral codes be such that they could perform the
function that the moral relativist expects of them? Let us split this question into two:
A. What is a moral code or framework?
B. What is it for P's doing D, to be “prohibited relative to a particular moral code M?”

end p 22

Starting with the first question, there are two possible conceptions of moral codes: as sets of general propositions or
as sets of imperatives. I will begin by discussing the propositional construal deferring until later in the paper an
examination of the imperatival construal.

When most people think of moral codes, they think of them as sets of general propositions which encode particular
conceptions of right and wrong, of moral prohibition and permission, with different societies accepting different sets of
such propositions and so adhering to different moral codes.

Thus, Harman, in describing the moral diversity which he finds in the world at large, says:

Members of different cultures often have very different beliefs about right and wrong and often act quite
differently on their beliefs . . . . Some societies allow slavery, some have caste systems, which they take to be
5
morally satisfactory, others reject both slavery and caste systems as grossly unjust.
5 Harman and Thomson, p. 8.

Or consider the following characterization, picked more or less at random from the web, due to an organization called
“ReligiousTolerance.org”:

The purpose of this essay is to show the wide diversity of moral codes that exist today and in the past. . . . The
result of this diversity is that one group of people may consider an action moral, while another group will
 p p p p p y

regard it as morally neutral, and a third group may decide that it is profoundly immoral. Each will be following
their own moral code.

Of course, it may take a certain amount of reflection for a person to be able to formulate the beliefs which constitute
his moral code. In that sense, moral codes might exist more as tacit beliefs than as explicit ones. But that is no reason
not to take them to be at bottom propositional attitudes.

Not only is it in general very natural to take moral codes to be sets of propositions; there are special reasons for the
relativist to insist on this construal and this brings us to the second question.

Recall that the relativist urges us not to assert that:

It would be wrong of Paul to steal Mark's car

but only that:

In relation to moral code M, it would be wrong of Paul to steal Mark's car.

But what could “in relation to moral code M” mean if not “is entailed by moral code M?” How else are we to think of a
moral code ruling on the correctness of a potential infinity of particular moral judgments except either by entailing
them or by failing to entail them?

If there is, indeed, no alternative to this picture, then the propositionalist picture of moral codes is forced, for that is
the only way in which we can make sense of moral codes as standing in entailment relations to particular moral
judgments.

end p 23

The answer to our two questions, then, is this: moral codes are sets of general propositions specifying alternative
conceptions of moral right and wrong. These codes entail particular moral judgments about specific acts. According to
moral relativism, then, we should speak not of what is and is not morally prohibited simpliciter, but only of what is and
is not prohibited by particular codes.

I will call any relativistic view that is characterized by this pair of features—the relativization parameter consists of a
set of general propositions and these propositions stand in entailment relations to the target proposition—a Fictionalist
brand of Replacement Relativism. (Compare truths about fictional characters: there are no truths of the form Sherlock
Holmes lived on Baker Street, but only ones of the form According to the stories of Arthur Conan Doyle, Sherlock
Holmes lived on Baker Street.)

Problems for Fictionalist Moral Relativism

I believe that this Fictionalist view of morality captures well what many have wanted to mean by the phrase
“relativism about morality.” However, no sooner is it stated than it begins to come apart.

One problem seems immediate. The judgment:

It would be wrong of Paul to steal Mark's car

seems appropriately normative; but the judgment:

In relation to moral code M, it would be wrong of Paul to steal Mark's car

seems just to be a logical remark about the relation between two sets of propositions. It seems to have no normative
impact whatsoever: even someone who was in no way motivated to avoid stealing Mark's car could agree with the
claim that, in relation to a given moral code, it would be wrong of Paul to steal Mark's car. So, right away, there is a
difficulty seeing how we could adopt the relativist's recommendation in this case without this being tantamount to our
6
giving up on moral judgments altogether, rather than merely relativizing them.
6 J. J. Thomson makes a similar point in her response to Harman.

It might be thought that the relativist has an easy fix to this objection. After all, it's part of his story that although
there are many possible moral codes, a particular person will accept a particular one of these while rejecting all the
others. It also seems plausible that when someone makes one of the relativist's relational judgments involving a code
that he accepts, that judgment will have a special normative pull on him. Thus, the relativist would appear
well-advised to revise his view to bring this acceptance of particular moral codes into the picture. He should
recommend that the replacing proposition be not:

In relation to moral code M, it would be wrong of Paul to steal Mark's car

but, rather,

end p 24

(17) In relation to moral code M, which I, the speaker, accept, it would be wrong of Paul to steal Mark's car.

How effective is this modification in dealing with the bruited objection? It's not clear. Whereas:
 p p p p p y

It would be wrong of Paul to steal Mark's car

is clearly normative,

I believe that it would be wrong of Paul to steal Mark's car

which is, in effect, what (17) states, looks to be just a description of one's own mental states.

But whatever one thinks of the effectiveness of this patch, the deeper question is whether we are able to make sense
of a thinker's endorsing a particular moral code, once he has accepted the guiding thought behind moral relativism. To
see the problem here, let us turn our attention to the particular moral judgment:

It would be wrong of Paul to steal Mark's car

that we are supposed to relativize to moral codes. There are no absolute facts about moral prohibition or permission,
we have said, so all statements of this form are condemned to untruth. If we are to hang on to moral discourse, we
must reform it so that we talk not about these nonexistent absolute facts but only about the code-related relational
facts.

Now, there are, of course, two ways for a statement to be untrue. On the one hand, a statement may be untrue
because it is false; and, on the other, it may be untrue because the proposition it expresses is somehow incomplete,
so that it doesn't specify a fully evaluable truth-condition. Let us call the first option an Error Theory about the target
utterance and the second an Incompleteness Claim about it.

Now, it should be fairly clear that our Fictionalist relativism about morality will work best on an Error Theory of moral
utterances and not so well with an Incompleteness Claim about them. For if we are recommending that we stop
judging “x is morally prohibited” but only make judgments of the form:

M entails that x is morally prohibited

we cannot very well be thinking of:

“x is morally prohibited”

as expressing an incomplete, non-truth-evaluable proposition: it is not possible, I take it, for an incomplete
proposition like:

end p 25

Tom is taller than . . .

to be entailed by a set of propositions.

In and of itself, though, I do not regard the commitment to an Error Theory to be a problem. As I have already
indicated, I don't regard Error Theories as intrinsically implausible. And, given a choice between an Error Theory and a
Non-Factualism about a particular sentence, there is, I believe, strong reason to prefer the Error Theory, all other
things being equal. The view that speakers who have been trying to make absolutist claims about motion, mass or
morality, and who have developed an internally disciplined discourse about these subject matters, have nevertheless
not succeeded in so much as making a complete claim is not particularly plausible. It would be far more plausible to
say that they had succeeded in making such claims, but that those claims have turned out to be false in certain
systematic ways.

Unfortunately, though, an Error Theory of moral discourse turns out to be a problematic commitment for Fictionalist
relativism in a variety of different ways.

To begin with, if we say that the target judgments, like:

It would be wrong of Paul to steal Mark's car

are false, we will have to say that the propositions which make up the code to which this judgment is to be relativized
are also false, for these are very general propositions of much the same type. Whereas an ordinary moral judgment
will speak about the permissibility of stealing some particular person's particular item of personal property, a code will
speak much more generally about the permissibility of damaging anyone's possessions. But if there are no absolute
facts with which to confirm particular moral judgments, there won't be any general moral facts with which to confirm
these more general moral judgments.

Now, however, a serious puzzle emerges for the moral relativist. Our only idea, you may recall, about how to preserve
the normative character of moral judgments, on a relativist view of them, involves giving a central place to our
acceptance of moral codes. But how could we be expected both to accept the relativist's claim that the target
judgments are false, which clearly implies that the propositions constituting the code are also false, and to continue
accepting those codes in the way required to make sense of the relativist's picture?

Moral codes, we have said, are composed of general moral propositions. To accept such propositions is, presumably, to
believe them and act on their basis. However, once we have come to agree with the relativist that there are no
absolute moral facts, we seem committed to concluding that the propositions that make up the codes are false as well.
How, at that point, are we supposed to continue believing them? How does one continue to believe a proposition that
 p p p p p y

one believes to be false?

Some may be tempted at this point to invoke a distinction advocated by constructive empiricists, between acceptance
and belief. To believe a proposition,

end p 26

they may say, is to hold it to be true. But to accept it need not require that. One can accept a proposition in the sense
7
of being willing to use it as a premiss in one's reasoning without taking it to be true.
7
See Bastian C. van Fraassen, The Scientific Image (Oxford: Clarendon Press, 1980).

Perhaps the relativist can help himself to this distinction, insisting that we can go on accepting codes even while we
disbelieve what they say.

I, myself, am quite suspicious of the cogency of this distinction. But whatever its general merits, it seems to me that a
puzzle would remain even if we were allowed full use of it. The puzzle would be to explain how any particular code
could continue to have any special normative authority over us, once we have come to think of all the codes as
uniformly false. How could we be motivated to follow one set of precepts as opposed to another, when we have come
to think of all of them as uniformly false?

Of course, one remedy would be to allow that some propositions in the codes can be true. But that would not help
rescue moral relativism because it would involve acknowledging the existence of some absolute moral truths and that
is precisely what the relativist needs to deny.

But even if we were to put this powerful point to one side, the relativist faces a further difficulty avoiding commitment
to the existence of some absolute moral truths, on a Fictionalist picture.

Recall that we are operating with the assumption that ordinary moral judgments express complete truth-evaluable
propositions that are false, and that, as a result, so too are the propositions which make up the moral codes.

But how could all codes be equally false? Suppose we have one code, M1, which says:

M1: Slavery is prohibited.

And another moral code, M2, which says:

M2: Slavery is permitted.

If we take the view that M1 is false, doesn't that require us to say that M2 is true?

At least the way the matter is usually taught, it is an analytic truth about moral prohibitions that if it is false to say
that x is morally prohibited then it must be true to say that not-x is morally permitted. But the relativist about
morality no more wishes to say that there are absolute facts about moral permission than he wishes to say that there
are absolute facts about moral prohibitions. And so, once again, a commitment to the falsity of one absolute moral
proposition seems to commit us to the truth of some other absolute moral proposition.

Now, I suppose that there have been philosophers who have denied that prohibition and permission are duals of each
other in this sense. But it would be odd to think that whether a relativistic view of morality is so much as available
turns on whether this widely believed conceptual claim is rejected. It would be better if

end p 27

we could find a conception of relativism which did not depend on something so tendentious.

Completion Relativism
All of these problems for a Fictionalist relativism about morality trace back to the assumption that our target
utterance:

“It would be wrong of Paul to steal Mark's car”

expresses a complete truth-evaluable proposition. Once that assumption is in place, there is no alternative but to
embrace an Error Theory of that utterance, and, given a propositional construal of moral codes, no alternative but to
take a relativistic conception of it as involving its entailment by a set of similar, though more general, false
propositions.

The question, therefore, quickly suggests itself: is that assumption optional?

To get away from absolute facts about morality, all we really need is for judgments of the form:

It would be wrong of Paul to steal Mark's car

to be untrue; we don't have to think of them as false. An obvious alternative, then, is to think of the relativist as
having discovered that the target judgments suffer not from Error but from Incompleteness, and so as calling for their
completion by reference to moral codes. Call this Completion Relativism.

Couldn't we use Completion Relativism to generate a more plausible model for moral relativism?

In contrast with the view I earlier attributed to him, Judith Jarvis Thomson reads Harman as endorsing precisely such
 p p p p p y

a Completion view. She describes him as claiming that:

Moral sentences (such as “It would be wrong of Paul to steal Mark's car”) are in a certain way incomplete;
indeed, it is because [such] sentences are incomplete in that way that they lack truth values. . . . Non-moral
sentences such as [“In relation to moral code M, it would be wrong to steal Mark's car”] are completions of
8
moral sentences.
8
Harman and Thomson, p. 190.

So perhaps we simply got off on the wrong track, using the case of motion to motivate a Fictionalist Relativism about
morality rather than a Completion Relativism about it.

Well, suppose we say that a proposition of the form:

It would be wrong of Paul to steal Mark's car

is an incomplete proposition, in much the way that:

Tom is taller than . . .

end p 28

is clearly incomplete. And suppose we try to complete it with:

In relation to moral code M, it would be wrong of Paul to steal Mark's car.

Once again, though, it is hard to see how to make sense of this.

First, we are still operating under the assumption that moral codes are sets of propositions which encode a particular
conception of moral prohibition and requirement. That means that a moral code must be seen as consisting in
propositions of the form:

x is morally wrong.

But since we have just finished saying that particular propositions of the form:

x is morally wrong

are semantically incomplete, we would have to say the same thing about the more general propositions which are
supposed to constitute the moral code, for they are basically propositions of the same general type.

But just as it was hard to see how anyone could believe a set of propositions that they knew to be false, so it is hard
to see how anyone could believe a set of propositions they knew to be incomplete.

Second, if the propositions that constitute the code are incomplete, it is very hard to see how they could constitute a
conception of anything, let alone a conception of right and wrong. Before they could be said to amount to a conception
of anything, they would have to be completed. But our only idea about how to complete them is by reference to moral
codes! And now we would seem to have embarked on a vicious regress in which we never succeed in specifying the
conception of permission and prohibition which is supposed to constitute a particular community's moral code.

Third, how are we to understand the phrase “relative to moral code M”? Since we have said both that the propositions
which constitute a moral code as well as the target propositions are incomplete, that relation cannot be the relation of
logical entailment. “Relative to moral code M,” then, must be understood as expressing some non-logical relation that
obtains between x's being morally prohibited and some moral code. But what could such a non-logical relation possibly
be?

For all of these reasons, then, it looks as though Completion Relativism is not a cogent option either.

What Has Gone Wrong?

The preceding argument establishes, I believe, that if moral relativism consists in the claim that ordinary moral
judgments are to be relativized to moral codes, where moral codes consist in general propositions of much the same
ilk as the judgments that are to be relativized to them, then there is nothing very coherent

end p 29

for moral relativism to be. Yet this is, of course, a very familiar conception of moral relativism. What has gone wrong?
Why would moral relativism turn out to be not even coherently assertible when the relativisms on which it is
based—those concerning mass, motion and simultaneity—are not only coherent but true?

The first thing to say in response to this question is that the sense in which a Fictionalist Relativism about morality is
similar to the successful relativisms drawn from physics is very superficial. The two cases look alike but in fact have
very different logical properties. Thus, in the motion case, we say that we have discovered that:

“x moves”

is untrue, and needs to be replaced by:

 p p p p p y

“x moves relative to F.”

And in the moral case we say that we have discovered that:

“x is wrong”

is untrue and needs to be replaced by:

“x is wrong relative to moral code M.”

However, this similarity in surface grammar masks a deep difference in the underlying logical forms.

In the moral case, we are moving from a judgment of the form:

x is P

to a judgment of the form:

(x is P) bears R to S.

In other words, in the moral case, the replacing proposition is built up out of the old proposition in a quite literal
sense: the replacing proposition consists in the claim that the old proposition stands in some sort of contentual
relation to a set of propositions that constitute a code. This feature is crucial to the thought that lies at the core of
standard conceptions of moral relativism that although we can no longer assert moral judgments in some unqualified
way, we can truthfully talk about which of those judgments are permitted by different moral codes.

In the physics cases, by contrast, while the proposition expressed by “The Earth moves” is of the form:

x is P

the proposition expressed by the replacing sentence “The Earth moves relative to F” is of the form:

x R y;

end p 30

the concept of a monadic property moves has been replaced by the concept of a relational property moves-relative-to,
a concept which no more contains moves as a proper part than the concept alternative contains the concept native as
a proper part. There is no way to understand the replacing proposition in this case as built up out of the old
proposition in the manner that is presupposed by a Fictionalist view of morality.

This difference at the level of logical form leads to the problems that attend Fictionalist Relativism. Because, in such a
view, the old proposition is retained intact in the new proposition, it carries with it the charge of untruth that launches
a relativistic conception. Because the new proposition places the old absolute proposition into a contentual relation to
a set of propositions of a very similar nature, that charge carries over to the parameter to which the truth of moral
propositions is being relativized. The result becomes something barely intelligible.

The moral relativist's best shot is to say that the original absolute propositions are complete but false. On this option,
however, he faces the difficulty of explaining how we are able to accept a set of propositions that we know to be false;
and the difficulty of explaining how he avoids commitment to there being some absolute moral facts somewhere after
all, because the falsity of moral prohibitions seems to entail the existence of moral permissions.

If he now attempts to retreat to the view that the target propositions are incomplete, his problems get even worse: for
now he faces the difficulty of saying how alternative moral codes could be alternative conceptions of the moral facts
and how proposition fragments could complete, in the required way, the incomplete character of the target
propositions.

None of these problems arise in the physics cases because in those cases the old proposition is discarded and is simply
replaced with a new relational proposition, which, it turns out, is the only truth that there is in the vicinity of the old
judgments.

Codes as Imperatives
Well, why isn't the lesson of these considerations that we should hew to the physics model far more closely? Why
couldn't we formulate a satisfactory moral relativism by saying that in the moral case,

It would be wrong of Paul to steal Mark's car

is discovered to be untrue and is to be replaced not by:

[It would be wrong of Paul to steal Mark's car] is entailed by moral code M

but rather by:

Paul stealing Mark's car is wrong-relative-to M,

end p 31
 p p p p p y

where the hyphenation makes clear that we are now talking about a replacement proposition of the form “xRy”?

We will turn to looking at what this might mean in a moment, but before we do so, we should pause to remark on a
small puzzle that attends the cases that are drawn from physics.

The puzzle can be put as follows. If my account of the physics cases is correct, classical motion judgments are untrue.
If anything in their vicinity is true, it is not those original motion judgments but some other judgments, with a
different logical form and involving distinct concepts. But how could the truth of these other judgments, which do not
themselves involve the concept of motion, amount to the discovery that a relativism about motion is true? If I replace
talk of phlogiston with talk of oxygen, that is not a way of discovering the atomic nature of phlogiston. Why do we
take ourselves to have discovered that motion is relative?

In the usual account of things, there is supposed to be a distinction between eliminativism and relativism. In the case
of the former, we respond to the discovery that a domain of discourse is systematically untrue by rejecting that
discourse and, possibly, substituting some other discourse in its place. In the case of the latter, we respond by
declaring that we have discovered the truths in that domain to be relative in nature. But how are we to make out this
distinction between eliminativism and relativism if, even in the classic cases of alleged relativism, what we get is the
wholesale replacement of one set of judgments by another? Why aren't all cases, including the famous ones from
physics, just cases of eliminativism?

This puzzle would be solved if, say in the case of motion, there were a more general concept, MOTION, itself neither
absolutist nor relativist, such that both the absolutist and the relativistic notions could be seen to be subspecies of it. I
think it is likely that there is such a concept, but I won't attempt to define one now.

It will suffice for present purposes to point out that, given that a relativist view of a given domain always involves the
replacement of the original absolute judgments by certain relational judgments, we need to be shown that these two
sets of judgments are sufficiently intimately related to each other, in the sense just gestured at, to justify our saying
that what we have on our hands is relativism and not eliminativism. Let us call this the requirement of intimacy.

Let us turn, now, to the question of what moral relativism would look like if it assumed a non-Fictionalist form of
Replacement Relativism.

The two central questions we face are: what relation could “is-wrong-relative-to” be? And what parameter could “M”
designate such that specific acts of stealing might or might not bear that relation to it?

It is already clear from the foregoing considerations, that whatever a moral code is going to be, it cannot consist of a
set of propositions.

A natural alternative conception, as I've already noted, is to think of it as a set of imperatives, rather than
propositions. So let us see how far we can take that idea, which, in any event, might be thought to have independent
appeal.

end p 32

If moral code M consists of a set of imperatives of the form:

Don't do x!

how should we think of the relation “is-wrong-relative-to”? A natural suggestion would be that it means “is one of the
acts prohibited by.”

If we put all this together, and we remind ourselves of the importance of including a reference to the speaker's
endorsement of a particular code, we get the following account:

“Paul's stealing Mark's car is wrong-relative-to M” means: Paul's stealing Mark's car is one of the acts prohibited
by the set of imperatives, M, which I, the speaker, accept.

How well does this proposal fare?

Problems for Imperatival Moral Relativism

One immediate question concerns the requirement of intimacy. How are we going to show that the replacing and
replaced propositions are sufficiently intimately related to justify our saying that we have shown moral truths to be
relative? How are we to ward off the suggestion that what we have here instead is just moral nihilism, with moral
discourse discarded in favor of some replacement where we talk not of what is good and bad but only of what is and is
not permitted according to imperatives that we accept and choose to live our lives by?

This story would be easier to fill in, if we could explain what makes a given set of imperatives moral imperatives.

The relativist's idea is that moral judgments need to be relativized to moral codes. On the propositional construal, it
was quite clear what made the relevant codes moral: they were constituted by propositions which encoded particular
conceptions of moral right and wrong. That conception of moral codes, however, proved unworkable for the purposes
of relativism. On the imperatival construal, however, we simply have a set of imperatives without any distinctive
moral content. What makes these imperatives moral imperatives as opposed to some other sort of injunction?

Clearly, we would have to look not to the imperatives themselves, but to their acceptance by agents to see what
 p p p p p y

might distinguish them from imperatives of another variety—prudential imperatives, for example, or aesthetic ones.
There had better be something distinctive about the acceptance of moral imperatives for otherwise we would not have
pinned down that we are relativizing to a specifically moral code. But it is far from clear that there is such a distinctive
9
state of mind.
9 For a particularly sophisticated account see Allan Gibbard, Thinking How to Live (Cambridge: Harvard University Press, 2003).

end p 33

A further problem that attends the imperatival construal concerns its ability to capture norms of permission. Let me
explain.

As we have already had occasion to note, moral norms might be either norms of permission or norms of requirement.
We may be required to do something, A, under particular conditions, C, or simply permitted to do it.

Imperatives, however, are, by definition, of the form:

If C, do A.

It is this logical form that marks them out as non-propositional—incapable of assessment as true or false.

However, it is not clear how something of imperatival form is going to capture a norm of permission, a norm that
merely allows doing A if C but does not require it. If I issue the imperative:

“If that car doesn't belong to you, don't scratch it”

I am requiring you not to scratch it, not merely permitting it.

For that reason, it is very hard to see how imperatives alone could constitute a fully satisfactory construal of what
moral codes are.

A further problem for the imperatival construal will be familiar from our discussion of the propositional construal. In
that discussion, we saw that the relativist had difficulty accommodating the normative character of moral judgments,
even after the thinker's endorsement of a particular moral code is brought into the picture.

(17) In relation to (propositional) moral code M, which I, the speaker, accept, it would be wrong of Paul to steal
Mark's car.

As I pointed out, this still seems to be a descriptive remark about which beliefs one has, rather than a genuinely
normative remark about what ought and ought not to be done. This problem, it seems to me, only intensifies on the
imperatival construal whose counterpart of (17) is:

In relation to the set of imperatives, M, which, I, the speaker, accept, it is prohibited for Paul to steal Mark's car.

The final and probably most important objection to the imperatival construal derives from an observation we had
occasion to make earlier—namely, that moral codes are normally expressed through the use of indicative sentences
that are themselves just more general versions of the sentences by which particular moral judgments are expressed.
Thus, it would be natural to say that our moral code, which contains a prohibition against stealing, would be expressed
by the sentence:

(17) “Stealing is wrong.”

And this sentence is itself just a general version of the various particular sentences by which we express ordinary
moral judgments such as:

end p 34

(18) “Paul's stealing Mark's car is wrong.”

Now, though, if (17) is taken to express an imperatival content roughly given by:

Don't Steal!

it's very hard to see how (18) could have any other sort of content because they are both sentences of exactly the
same type, the only difference being that (17) is a generalization of (18). In other words, if the imperatival construal
of moral codes is correct, we would appear to have no choice but to take (18) to express the content:

(19) Don't steal Mark's car, Paul!

(18), though, is just the sort of ordinary, unqualified, absolutist sentence that the relativist is going to seek to replace
by relational sentences of the form:

(20) “Paul stealing Mark's car is wrong-relative-to M.”

on the grounds that there are no absolute facts about morality with which to confirm the absolutist sentences.

However, if it is really true that (18) expresses an imperatival content of the kind specified by (19), then it is very
hard to see what motive there could be to replace it, or how the discovery that there are no absolute moral facts could
 p p p p p y

be relevant to it in any way. Since it just expresses the imperatival content:

Don't Steal Mark's car, Paul!

it has not made any claim; and since it has not made any claim it can hardly be true that it is condemned to having
made a false claim; and since it is not condemned to having made a false claim, it can hardly be true that it needs to
be replaced on the grounds that it is so condemned.

The imperatival construal of moral codes, then, far from showing us how coherently to implement a relativistic view of
morality seems rather to make relativism irrelevant and inapplicable.

Conclusion
In this paper, I have attempted to explain how a relativistic conception of a given domain should be understood. And I
have then tried to outline the obstacles that stand in the way of extending such a conception to the domains that have
10
most interested philosophers—those of morality and epistemic justification.
10
In this paper, I have confined myself to talking about the moral case; the epistemic case is treated explicitly in my book Fear of Knowledge (Oxford:

Oxford University Press, 2006), chs 5–7.

end p.35

It is natural to wonder whether there is an alternative conception of relativism that would serve some of these
traditional philosophical preoccupations better. The main alternative to the account I have been exploring is provided
by the idea that a relativistic conception of a given domain consists in the claim that there can be faultless
disagreement in that domain. Here is Crispin Wright:

Imagine that Tim Williamson thinks that stewed rhubarb is delicious and that I beg to differ, finding its dry
acidity highly disagreeable. There is, on the face of it, no reason to deny that this is a genuine
disagreement—each holding to a view that the other rejects. But it is a disagreement about which, at least at
first pass, the Latin proverb—de gustibus non est disputandum—seems apt. It is, we feel—or is likely to be—a
disagreement which there is no point in trying to settle, because it concerns no real matter of fact but is merely
an expression of different, permissibly idiosyncratic tastes. Nobody's wrong. Tim and I should just agree to
disagree.
Call such a disagreement a dispute of inclination. The view of such disputes just gestured at . . . combines
three elements:

That they involve genuinely incompatible attitudes (Contradiction);

That nobody need be mistaken or otherwise at fault (Faultlessness), and
That the antagonists may, perfectly rationally, stick to their respective views even after the
disagreement comes to light and impresses as intractable (Sustainability).

According to this characterization, then, relativism about a given domain is the view that there may be faultless
disagreement in that domain. One person may assert p and the other not-p, yet it needn't be the case that either of
them is at fault, not merely in the sense that both might be equally rational, but in the far more demanding sense
that both might have said something true.

But how could any domain pull off such a trick? How could it turn out that a proposition and its negation might both
come out true? Remarkably enough, a number of writers have recently proposed answers to this question.

Wright proposes to make sense of the combination of Contradiction and Faultlessness by invoking his view of truth as
super-assertibility. Kit Fine has suggested that we could make sense of it by regarding the opposed judgments as
targeting “different realities” (though Fine proposes his solution only in connection with issues about the passage of
time and not necessarily in connection with other subject matters). And John MacFarlane has explored the idea that
we can make sense of it by regarding the truth of a proposition as settled not just by a world and a time but also by a
11
“context of assessment.”
11 Crispin Wright, ‘Intuitionism, Realism, Relativism and Rhubarb,’ this volume (Ch. 2); Kit Fine, ‘Tense and Reality,’ in his Papers on Tense and Reality

(Oxford: Oxford University Press, 2006); John MacFarlane, ‘Making Sense of Relative Truth,’ Proceedings of the Aristotelian Society, 2003.

I am doubtful that we can ultimately make sense of the notion of a proposition that can sustain faultless
disagreement. I don't see how any such proposition

end p 36

could serve as the plausible object of belief, the very thing for which the notion of a proposition is needed. I also
believe that many of the arguments developed in this paper, about the difficulty of coming up with a satisfactory
characterization of the “standards” to which moral or epistemic judgments are to be relativized, will carry over to the
conceptions of relativism characterized by faultless disagreement.

However, many important questions remain unexplored and there is much interesting work that remains to be done.

end p 37
 p p p p p y

2 Intuitionism, Realism, Relativism and Rhubarb

1 The Ordinary View of Disputes of Inclination

Imagine that Tim Williamson thinks that stewed rhubarb is delicious and that I beg to differ, finding its dry acidity
highly disagreeable. There is, on the face of it, no reason to deny that this is a genuine disagreement—each holding to
a view that the other rejects. But it is a disagreement about which, at least at first pass, the Latin proverb—de
gustibus non est disputandum—seems apt. It is, we feel—or is likely to be—a disagreement which there is no point in
trying to settle, because it concerns no real matter of fact but is merely an expression of different, permissibly
idiosyncratic tastes. Nobody's wrong. Tim and I should just agree to disagree.

Call such a disagreement a dispute of inclination. The view of such disputes just gestured at—I'll call it the Ordinary
View—combines three elements:
1. that they involve genuinely incompatible attitudes (Contradiction);
2. that nobody need be mistaken or otherwise at fault (Faultlessness); and
3. that the antagonists may, perfectly rationally, stick to their respective views even after the disagreement comes to
light and impresses as intractable (Sustainability).
Assuming that there are indeed disputes as so characterized, it is of course an important and controversial issue how
far they extend—whether, for example, certain differences of opinion about ethics, or aesthetics, or justification, or
even theoretical science, come within range. But my question here is more basic: it is whether the three noted
elements can be combined coherently—whether there are any disputes of inclination, as characterized by the Ordinary
View, at all.

The question is given urgency by the fact that the four most salient alternatives to the Ordinary View all seem
rebarbative or misconceived. There is, first, the rampant realist proposal—an analogue of the epistemic conception of
vagueness. Rampant realism holds that there have to be facts of the matter which

end p 38

either Tim or I are missing. Rhubarb just has to be either delicious or not, so one of us has to be mistaken, even if
there is no way of knowing who. Such a view is vulnerable to a charge of semantical and metaphysical superstition. It
also arguably precludes Sustainability—the possibility of persisting in the dispute with rational integrity—since neither
Tim nor I have the slightest reason to think that our own tastes reflect the putative real facts about deliciousness,
once rampant—realistically conceived.

Realism need not be rampant. A more moderate realism might try to domesticate the relevant facts by attempting to
construe them as, in one way or another, response dependent—proposing, for instance, that what is delicious is what
(a majority of) well-qualified judges find to be so. But this seems a misdirection too: for one thing, I don't think we
really believe in ‘well-qualifiedness’ in basic matters of taste—that's the point of the Latin proverb. For another, the
proposal promises no better than its rampant counterpart in accommodating Faultlessness and Sustainability.

Recoiling from these views, one may be tempted by the thought that perhaps no genuine dispute is involved after all.
Perhaps the impression to the contrary is somehow an artefact of language. One—expressivist—version of that idea
has it that we are misled by the indicative surface of the dispute: maybe Tim's avowal that rhubarb is delicious serves
merely to give expression to the pleasure he takes in the stuff and is thus something with no properly negatable
content; maybe my avowal to the contrary serves merely to give expression to my corresponding distaste for it. Such
a proposal will face all the familiar difficulties in the philosophy of language—difficulties for example in accounting for
routine conditional, disjunctive, tensed and attitudinal constructions embedding such apparent indicative contents
—which are faced by strict expressivist proposals in other areas, and to which many believe they have no satisfactory
response.

An alternative strategy for denying that there is any genuine disagreement is to take the indicative appearances at
face value, but hold that the contents in question are not really in conflict—for instance, that they are elliptical and
that when the ellipsis is unpacked, the impression of incompatibility vanishes. It may be suggested, for instance, that
Tim's view is properly characterized as being that rhubarb is delicious by his standards, and that I am saying that
rhubarb is not delicious by mine. So we are talking past each other and may both well be right.

This suggestion is open to the charge that it distorts the meaning of what we intend to say when we give voice to
judgements of taste. There is, for example, a challenge involved in the question: ‘If, as you say, rhubarb is delicious,
how come nobody but you here likes it?’, which goes missing if the proper construal of it mentions an explicit
standard-relativity in the antecedent. So it looks as though a larger package will be called for, involving not just
hidden constituents but an error-theory concerning our ordinary understanding of the relevant kinds

end p 39

of claim. A related consideration points out that, on our ordinary understanding, the explicitly standard-relativized
kind of formulation represents a fall-back claim if the original, unqualified claim gets into difficulty—a puzzling
phenomenon if they coincide in content.
 p p p p p y

There are other forms of semantic contextualism, of course, besides those which postulate ellipsis or hidden
constituents. But the awkwardnesses just noted will remain on any such view. If Tim's and my differing tastes are
sufficient, one way or another, to ensure that we express different concepts of the delicious in our respective
assessments of rhubarb, and hence that there is no obstacle to our both being right, then why will we each be inclined
to withdraw if suitably many others don't concur? Why doesn't the contextualist explanation of why my judgement is
not in conflict with Tim's survive as a means to explain why I can be right no matter how idiosyncratic my view? And
why fall back on an explicitly standard-relativized claim if the content of my original claim was already implicitly
relativized?

Each of the four views canvassed—that there is a real but undetectable fact of the matter about whether rhubarb is
delicious, that there is a real but response-dependent fact of the matter, that there is no real matter in dispute
because no truth-evaluable content is involved, and that there is no real dispute because the contents involved are
elliptical, or otherwise contextually distinct—each of these four views not merely involves compromise of one or more
of the three components of the Ordinary View but seems open to additional objection. If we want to avoid
metaphysical hypostasis, snobbery in matters of taste, unplayable philosophy of language, or misrepresentation of
linguistic practice, then we should want the Ordinary View. So it comes as an unpleasant surprise that it seems, under
quite modest pressure, to collapse.

2 The Simple Deduction

1
The collapsing argument is what in earlier work I dubbed the Simple Deduction.
1
Wright (2001, 2002).

It is disarmingly straightforward. The idea that there is genuine disagreement involved in the dispute goes with the
idea that there is a genuinely indicative content, capable of featuring in attitudes and standing in relations of
incompatibility to other such contents. Any such genuine content can also be supposed. So: suppose that rhubarb is
delicious. Then I'm mistaken. But the Ordinary View has it that no one is mistaken (Faultlessness). So rhubarb isn't
delicious. But then Tim is mistaken. So someone has to be mistaken after all. Contradiction precludes Faultlessness.

end p.40

More explicitly:
1 (1) A accepts P —— Assumption
2 (2) B accepts Not-P —— Assumption
3 (3) A's and B's disagreement involves no mistake —— Assumption
4 (4) P —— Assumption
2, 4 (5) B is making a mistake —— 2, 4
2, 3 (6) Not-P —— 4, 5, 3 Reductio
1, 2, 3 (7) A is making a mistake —— 1, 6
1,2 (8) Not-[3] —— 3, 3, 7 Reductio
The occurrence of genuine disagreement seems to demand, by elementary and uncontroversial logical moves, the
2
existence of mistakes.
2
Note in particular that there is no appeal to the Law of Excluded Middle.

Further, once that's recognized, it becomes impossible to see how Tim and I can persist in our disagreement with
rational integrity. Apparently one of us has to be mistaken. But if one of us is mistaken, how can we tell who? Isn't it
just a conceit to think it has to be the other? So Sustainability is compromised too. Thus the three components in the
Ordinary View fall apart.

Faced with this difficulty, the natural temptation for a proponent of the Ordinary View is to try to refine the second
component—to qualify Faultlessness. Maybe it's too much to demand that there need be no mistake involved in the
dispute. Maybe the most that can be asked is that there be no epistemically blameworthy mistake. Perhaps
Faultlessness should be replaced by something like the idea that neither Tim nor I need have done anything which
would have opened our opinions to proper suspicion when considered in isolation, by someone with no view on the
matter in hand but otherwise as knowledgeable as you like. Or something like that.

But this suggestion doesn't really help. For one thing, part of the attraction of Faultlessness is that, while we want to
acknowledge that there may be no settling a dispute of inclination, we precisely don't want that acknowledgement to
commit us to the idea of potentially unknowable facts of the matter—that's why the rampant realist proposal strikes
us as so bizarre. The rhetoric of ‘no fact of the matter’ expresses the natural, folk-philosophical view: such disputes
are potentially irresolvable, we think, not because the facts in question can transcend our impressions but because the
impressions themselves are in some way basic and constitutive; so when they conflict, there need be no further court
of appeal. If that thought can be reconciled with the idea of truth at all, then truth—at least in matters of taste—had
better be per se knowable. But then the Simple Deduction is easily emended to argue not just that Tim's and my
disagreement must involve a mistake but that it must involve a cognitive shortcoming in the stronger sense proposed,
3
since one of us fails to know something that can be known.
3
For elaboration, see the ‘EC-Deduction’ at p. 60 of Wright (2001).

Note in particular that there is no appeal to the Law of Excluded Middle.

 p p p p p y

And indeed, even if the Ordinary View can somehow avoid commitment to evidential constraint, the situation is still
not stable. For the conclusion of the

end p.41

Simple Deduction, that there is a mistake—false belief—involved in any such dispute, still stands unchallenged, even if
no cognitive blame need attach to either disputant. And now, since for all I can tell it may as well be me who has a
false belief as Tim, and since Tim is in an analogous position, it still seems impossible to understand how it can be
rationally acceptable for us to agree to differ and persist in our respective views. The threat to Sustainability is
already posed by the concession that Faultlessness in the weak sense is precluded by Contradiction.

So far I've not said anything about relativism. It may be thought that the Ordinary View—the suggestion of the
possibility of genuine but fault-free disagreements in which the protagonists are fully rationally entitled to persist in
their conflicting opinions—is tantamount to relativism—specifically, to the idea that truth in the region of discourse in
which the dispute is articulated should be viewed as relative to the differences in standard, or context, or whatever,
which generate the disagreement in the first place. But this is not correct. Relativism, I want to suggest, is best
viewed as a theoretical attempt to underwrite and reconcile the elements in the Ordinary View. It is a response to the
problem, rather than merely a label for the amalgam of ideas which gives rise to it. Whether it is an adequate, or
theoretically attractive, response remains to be seen.

3 An Intuitionistic Response
First I want to table a different response. The Simple Deduction—exploiting, be it noted, only the most elementary
logic and placing no reliance on any distinctively classical moves—elicits a contradiction from the three assumptions,
that Williamson believes that rhubarb is delicious, that Wright believes that rhubarb is not delicious, and that nobody
is mistaken. The conclusion seems to be forced, accordingly, that somebody has to be mistaken in any genuine such
dispute. But it's not forced. There is a distinctively classical move involved in the interpretation of the reductio as
indicative that mistake always has to be involved. Specifically, take the third assumption as that:

It is not the case that Williamson is mistaken and it is not the case the Wright is mistaken.

Then to interpret the reductio as showing that someone must have made a mistake is to take it that the negation of
that conjunction licenses us in concluding:

Either Williamson is mistaken or Wright is mistaken.

That's to make an inferential transition of the form:

—a pattern whose classical validation demands elimination of double negations, and which is not in general
intuitionistically valid.

end p.42

Very well. But so what? How might sticking at the intuitionistically valid conclusion—the negated conjunction—put us
in a position to accommodate the components in the Ordinary View, and to reconcile them with each other? And even
if resisting the transition to the disjunction would help, how might intuitionistic restrictions sufficient to block the
relevant de Morgan Law be motivated in the type of context at hand?

Let's consider the second question first. The key issue, as always, concerns the status of the principle of Bivalence for
statements of the relevant kind. For since:

Not (Not A & Not B)

is, by uncontroversial steps, equivalent to:

4
Not Not (A V B),
4 Assume Not (Not A & Not B) and Not (A V B). Use the latter, vel-intro and Reductio to derive each of Not A and Not B. Conjoin them to derive

a contradiction with Not (Not A & Not B), and discharge Not (A V B) by a further Reductio.

Assume Not Not (A V B) and Not A & Not B. Assume A V B and reason by vel-elim and Reductio to Not [Not A & Not B]. A further step of

Reductio yields Not [A V B] on Not A & Not B as assumption. One more step of Reductio then gives Not [Not A & Not B] on Not Not (A V B) as

assumption.

the move at which it is being suggested we may balk is tantamount to double negation elimination for disjunctions. If
this class of cases of double negation elimination is accepted, Excluded Middle will hold quite generally, since its own
double negation may likewise be established by wholly uncontroversial steps. Thus assuming—as we may in this
5
dialectical context
5 To explain: the present dialectical context is one in which we are assuming that disjunction is distributive—that the truth of a disjunction requires the

truth of at least one of its disjuncts in particular. Otherwise, the conclusion that either Tim is mistaken about rhubarb or I am carries no implication of the

actual existence of a mistake. But where disjunction is distributive, the validity of Excluded Middle rests on Bivalence.

—that Excluded Middle rests upon Bivalence, the defensibility of the transition from the thesis that Tim and I cannot
 p p p p p y

both be right to the uncomfortable claim that someone in particular—either Tim or me—is mistaken about rhubarb,
rests on the defensibility of Bivalence for claims like: ‘rhubarb is delicious’.

In intuitionistic mathematics, the challenge to Bivalence is best seen as flowing from a combination of two claims: first
an insistence on a form of evidential constraint—that truth in mathematics may not defensively be supposed to outrun
decidability in principle by a certain loosely characterized class of constructively acceptable methods; second that, for
any theory at least as rich as number theory, we possess no guarantee that any given statement is indeed decidable
by such methods. Simply put: if Bivalence holds for Goldbach's conjecture—if either the conjecture or its negation is
true—then, by evidential constraint, one or the other will be verifiable by intuitionistically acceptable methods. So
since we do not, in our present state of information, know that either can be so verified, we do not, in our present
state of information, have any right to claim that

end p.43

Bivalence holds for Goldbach's conjecture, nor therefore throughout number-theoretic statements as a class.

The intuitionistic reservation about Bivalence is thus one of agnosticism. But it is not an agnosticism based on the
spectre of third possibilities—additional truth-values, or truth-value gaps. Rather it is based on our inability to
guarantee the possibility of knowledge, along with the thesis—held for independent reasons—that truth requires that
possibility for the type of statement for which the validity of Bivalence is under review.

Either of these claims may of course be contested for a given class of statements. But both may seem attractive for
each of two non-mathematical kinds of example, for which, accordingly, the validity of Bivalence may consequently
come into question. One comprises those vague statements typified by predications of adjectives like ‘red’ and ‘bald’.
The other is precisely our present focus: judgements of taste and other matters of inclination. In both these cases we
are antipathetic to the idea that truth has no implication of ascertainability; but in both cases we are likewise uneasy
with the suggestion that claims have to be decidable, one way or the other. In the terminology I have used in earlier
work, borderline cases of vague predications, and predications of concepts of taste, are, no less than mathematical
statements like Goldbach's conjecture, liable to present quandaries: examples where we may be uncertain not merely
what it may be correct to think but even whether there is any metaphysical space for knowledge, or all-things-
considered best opinion, properly so-termed. These two pressures—evidential constraint and the potentiality for
quandary—squeeze out an unqualified acceptance of Bivalence over the two classes of statements in question; but
they put no pressure on a continued adherence to the law of non-contradiction. So we should not, in reasoning among
these statements, rely on a logic which forces us to be insensitive to the distinction between them which, it appears,
had better be made.

This comparison—between statements like Goldbach's conjecture, borderline predications of vague concepts and
6
judgements of taste—has been misunderstood by at least one commentator
6
Kölbel (unpublished).

so further clarification may help. Undeniably, there is the following difference. While no one knows whether
knowledge either of Goldbach's conjecture or its negation is metaphysically possible and—it is tempting to add—no
one is really entitled to an opinion (contrast: a hunch) about the matter, borderline cases of vague predicates may
quite unobjectionably give rise to weak, qualified opinions. And matters of taste, for their part, may give rise to strong
ones. So what is the intended analogy between the three kinds of statements? What similarity is the notion of
quandary meant to mark? The answer is: a similarity which is manifested by each of the three kinds of statement as a
class. Sure, any particular statement of each of the three kinds in question is such that we cannot rule out the
possibility of a competent determinate—positive or

end p.44

negative—view of it (though with statements about borderline cases of vague concepts we can, admittedly, rule out
the possibility of a competent but strong view). But nor, in each of the three kinds of case, do we have any grounds
for thinking that knowledge, or in all things considered best opinion, has to be possible for every example. In
particular, while I may indeed have many opinions on matters of taste, and consider them competent, or even
superior, I have to acknowledge that I know nothing which ensures that a determinate knowledgeable or best opinion
is possible about every matter of taste or inclination generally. That would be a guarantee that all disputes of
inclination have a winner. We have no such guarantee.

There, then—in the combination of quandary and evidential constraint—is one kind of motivation for broadly
intuitionistic reservations about classical logic in general, and about the (in my formulation above, implicit) final step
in the Simple Deduction. If accepted, it allows us to stop short of letting the Simple Deduction conclude that someone
has to be mistaken in any dispute of inclination—indeed in any dispute about a genuinely indicative content.

Maybe the foregoing train of thought is of most interest in a context in which the primary question is whether the
intuitionists' ideas about the logic appropriate to mathematics can be generalized to other regions of discourse.
Anything properly viewed as an extension of their ideas will have to involve some kind of play with evidential
constraint, since that is the role, in the mathematical case, of their very constructivism. However we should not
overlook another, simpler, and perhaps yet more compelling line of reservation about Bivalence in the cases that
concern us, which puts evidential constraint to one side. Reflect that the opinion that Bivalence holds, of necessity,
 p p p p p y

throughout vague discourse is a commitment to holding that each vague predicate is associated with a property of
absolutely sharply bounded extension as its semantic value. But for a very wide class of such expressions—including
especially predicates of Lockean secondary qualities—we have no clear idea what kinds of properties these may be.
Nor, in general, do we have any clear idea how the required semantic associations might have been established. A
commitment to Bivalence holding of necessity in all such cases is a commitment to postulating a kind of arcane
natural history of semantic relationships for which we have absolutely no evidence. And it's just the same with
predications of taste. There is just the same semantic mystery, just the same puzzlement, in a wide class of cases,
about the nature of the properties that would be fit to discharge the demanded role. What is deliciousness if it is to be
possible for normally competent speakers, like Tim and me, to go so completely astray about it in a perfectly ordinary
case? The idea that there is a mandate for unrestricted Bivalence is, one way or another, a commitment to
philosophical obligations—perhaps rampant realist, perhaps response-dependence realist—which we simply do not
know how to meet. Surely the mere idea that Tim and I hold contradictory opinions about rhubarb ought to impose no
such obligations. The reductio carried out in the Simple Deduction properly

end p.45

takes us no further than to the conclusion that our opinions cannot both be true. It is classical logic that is responsible
for muddying the distinction between that and the idea that one in particular of us is missing the real fact.

4 Can the Intuitionistic Response Rescue Faultlessness and Sustainability?

As remarked, however, it is one question whether there is a well-motivated intuitionistic distinction to draw, in the
service of stabilizing the Ordinary View, between the claim that Tim and I cannot both be right about rhubarb and the
claim that one of us in particular must be wrong. Even if so, it is a further question whether we thereby secure the
means to say something effective in stabilizing the Ordinary View of disputes of inclination. The challenge was to
harmonize the three ingredients—Contradiction, Faultlessness and Sustainability. And the point hasn't gone away that
if it is insisted that a dispute can be regarded as fault-free only if it's open to us to suppose that each antagonist has a
correct view, then a mere acceptance that the dispute is genuine—so involves contradictory opinions—precludes
regarding it as fault-free. Punkt.

The question, of course, is what, in regarding such a dispute as potentially fault-free, we really intend to maintain.
Well, each will have to examine their own preconceptions. But my own impression is that the principal point is to
contrast the case with situations where, should attempts at resolution fail, the mere existence of a contrary opinion,
no worse supported than one's own, is sufficient to put one at fault in persisting in one's view. That will be a
characteristic of the rhubarb dispute once the Simple Deduction is allowed to establish the disjunctive conclusion:
either Tim is mistaken or I am. As soon as it is accepted that one of us has to be mistaken, the fact that neither of us
is able to make his opinion prevail ought to encourage the worry that the mistaken party could as well be him as his
antagonist. And once one recognizes that, then it should seem at best pig-headed not to withdraw from one's initial
opinion. If this is right, then the really important thing about the idea of fault-free disagreement in such cases is
actually its implication of Sustainability—its implication of the idea that the opinions in a dispute of inclination may
justifiably be persisted in, even when it is clear that it is a stalemate.

This comes close to but is not quite the same thing as suggesting that the essence of the Ordinary View can be
captured just by the first and third components—Contradiction and Sustainability. But that conclusion is not right.
There are readily conceivable cases where Contradiction and Sustainability are satisfied but where there is—or may
be, depending on one's view—no proper comparison with disputes of inclination. Consider for instance two rival
scientific theories which match in their empirical, explanatory and other virtues, which are unsurpassed by any other
extant theory, and for which we've yet to

end p.46

devise a crucial experiment. It is debatable whether it should be regarded as irrational for a supporter of either theory
to persist in holding to it even after he becomes aware of the credentials of the other. After all, there is, by
hypothesis, no sufficient reason to adopt the opposing view—there is, by hypothesis, parity of virtue. And merely to
abandon either theory without putting anything in its place would mean restoring all the disadvantages, whatever
they may be, of having no theory of the subject matter in question at all. In such a case, then, regarding the dispute
as genuine and factual is quite consistent with Sustainability. If so, then even if the intuitionistic response can indeed
save Sustainability—I will address that in a minute—the scientific example shows us is that we need something extra,
something to play the role of Faultlessness, if we are to explain the difference between the two kinds of case. And we
are still no wiser about what that extra might be, consistently with the Simple Deduction, nor about whether the
intuitionistic setting can provide it.

The difference between the two kinds of case—rhubarb and the scientific theoretical disagreement—consists in the
way in which Sustainability is supported. In the scientific example, there is reason to accept (at least if one is scientific
realist) the disjunctive claim: one theory or the other—and perhaps both—will be false to the facts. One in particular
—perhaps both—of the rival theorists will be proposing a misrepresentation of Nature. And the point is then that,
notwithstanding that consideration, there are nevertheless overriding pragmatic reasons, grounded in the desirability
of having a theory in the first place, for each to persist in their respective views—so that we have Sustainability
 p p p p p y

anyway. In the rhubarb dispute, by contrast, there is—according to intuitionistic proposal—no impartial reason to
suppose that one disputant in particular—Tim or me—is making a mistake; and it is because there is no reason so to
suppose that we have Sustainability.

So the suggestion at which we arrive is this: disputes of inclination may indeed be stably characterized by ascribing to
them versions of all three features proposed by the Ordinary View: they are genuine disputes in which conflicting
opinions are held; they may be fault-free; and they may be rationally sustained even after it becomes clear that they
are stand-offs. The refinements we need to add are, first, that in disputes of inclination Sustainability is properly seen
as grounded in Faultlessness; in disputes of fact, by contrast, Sustainability, where it occurs, is grounded
otherwise—in the scientific theoretical example, for instance, it is grounded pragmatically. Second, Faultlessness
needs to be interpreted not as something flatly inconsistent with genuine conflict—with Contradiction—but rather as
something that resides in the unavailability of any impartial reason to make (the relevant analogue of ) the
disjunctive claim: to insist that there is fault somewhere. What counts against rationally sustaining a dispute, once
debate is exhausted without producing a winner, is the thought, roughly, that someone is mistaken here and, for all
that has emerged, ‘it could as well be me’. Once it is granted that someone has to be mistaken, that thought locates a
concern that rationally ought to occur to each of the antagonists. The concern may still not

end p.47

mandate withdrawal if, as in the scientific theory case, there are overriding reasons that license retaining a view.
7
But—the crucial point—it does not get off the ground without independent
7 Independent, that is, of one's view of the matter in hand.

reason for the disjunctive claim. It is by refusing the disjunctive claim that the intuitionistic proposal rescues
Sustainability, and grounds it on Faultlessness, with the latter now understood precisely as located in the shortfall
between the negated conjunction—which, it is conceded on all hands, the Simple Deduction establishes—and the
stronger disjunctive claim which is what it takes to implicate error on one side or the other.

This proposed way of developing the Ordinary View and staving off the threat posed by the Simple Deduction seems to
me to be stable this far. The question is whether there is any serious additional cause for dissatisfaction with it.

5 A Problem for the Intuitionistic Rescue

The intuitionistic rescue reconciles Contradiction with Faultlessness by insisting that it is insufficient for a dispute to
involve Fault, merely that it be a genuine dispute—genuinely involving contrary or contradictory opinions. Conflict of
opinion—it is contended—suffices for the presence of mistake only when Bivalence is guaranteed to hold for the
discourse of the dispute; and that, it is argued, is something for which there is—in the cases which concern us—no
guarantee.

Someone who is sympathetic to intuitionistic ideas is not likely to find this a particularly controversial application of
them. And indeed I would suggest that this also makes for an argument in the opposite direction. Absent a better kind
of proposal, the need to make sense of the Ordinary View, and the apparent impossibility of doing so in a classical
framework, provides a powerful argument for sympathy with intuitionistic distinctions and for further work on them.

There is, however, a problem with the approach which, if we are convinced that coherent provision must be made for
the Ordinary View, threatens to force us to look further afield. Simply stated, it is this: that since the Ordinary View is
inconsistent with rampant realism, no justice can have been done to it by an account that is consistent with the
possibility that rampant realism is correct. But the intuitionistic proposal merely leaves us in a position of agnosticism
about that. The response to the Simple Deduction was to argue that there is no justification for the relevant transition
of the form:

Even granting the proposed interpretation of Faultlessness, that is merely to say that there is no extant justification
for regarding either Tim or me as having a

end p.48

mistaken opinion. But to say that there's no justification for regarding the dispute as involving a mistake is not to say
that it's not the case that the dispute involves a mistake. Yet surely, the objection says, Faultlessness should involve
the latter. Yet the latter—the negation of the disjunction—does entail, even intuitionistically, the negations of both
disjuncts. And those, conjoined, are then inconsistent with the intermediate conclusion—the negated conjunction
—which, everyone agrees, the Simple Deduction does establish (and indeed inconsistent in their own right).

So, a critic may contend, the intuitionistic rescue has not really saved Faultlessness in any intuitively sufficient sense.
The most that has been saved is justification for our reluctance to attribute fault in relevant cases, consistently with
acknowledging the Simple Deduction. This leaves it epistemically open that there is indeed a determinate fact of the
matter in the rhubarb dispute, and indeed in such disputes in general, and hence that there is indeed a determinate
fault on one side or the other. And that is exactly what we—most of us—are reluctant to believe. It is good if the
intuitionistic proposal can save us from being forced to think it true just by elementary logic. But we would like to be
 p p p p p y

in a position to think it false.

A supporter of the intuitionistic rescue may rejoin that it is no serious shortcoming in the proposal that it leaves us at
most unsympathetically agnostic towards the rampant realist view of the dispute. After all, that, as it may seem, just
is the extent of the justified position. The rampant realist view calls for the association of the predicate ‘delicious’,
understood as by both Tim and me, with a property that determinately applies or fails to apply to stewed rhubarb. We
may not believe there is, as a matter of metaphysics, any suitable such property, much less that our linguistic
practices somehow enthrone such a property as the Bedeutung of ‘delicious’. But come on: we do not know that these
things are not so—not if knowing requires being in a position to prove it. The honest objection to rampant realism is
not that we know that its presuppositions are not met but that there is not the slightest reason to regard it as true. If
the preconceptions that underwrite the Ordinary View slur that distinction, they are not to be respected to the letter.
We should stick to what we can justify.

This reply, though, only partially addresses the objection. Maybe we do not, strictly, know that rampant realism is
false. But at the level of analysis displayed by the Simple Deduction, even with intuitionistic distinctions
superimposed, the point remains that no space is left for a coherent belief that neither Tim nor I is mistaken in the
original dispute. In particular, no way whatever has been offered of recovering a content for the idea that there is no
‘fact of the matter’ to be mistaken about. Even if we don't know that the rampant realist's insistence that there is
indeed a fact of the matter is itself mistaken, it may yet be felt as a very serious limitation of the intuitionistic
treatment if it does not, so far, allow us so much as to attach content to the idea of that mistake. The worst mistake of
which we have been empowered to make sense is an epistemic mistake: one of lack of

end p.49

warrant—the unwarranted insistence that the world and the relevant concepts are bound to conspire to render true
one of the disputed opinions or the other. But nothing has been said to explain how, or in what respect, rampant
realism might be incorrect, rather than merely unjustified.

The intuitionistic rescue provides theoretically respectable houseroom for our reluctance to be press-ganged into
realism by the Simple Deduction. But it does not offer—and it seems has no resources to offer—any account of what it
would be for (rampant) realism to be, not merely not imposed, but false: a misrepresentation in its own right. Surely,
it may be felt, a satisfactory account of disputes of inclination should explain how it is possible that this might be so,
even if we are forced to grant that, in the end, we are not in position to show, once and for all, that it is so.

6 The Intuitionistic Rescue Rescued?

The objection may seem convincing. But in fact it runs together two distinct complaints and arguably derives some of
its force from the conflation.

One complaint is that the intuitionistic rescue treats the transition from the conjunction:

It is not the case that neither Williamson nor Wright is mistaken

to the disjunctive conclusion:

Williamson is mistaken or Wright is mistaken

merely as a non-sequitur, whereas someone who takes the Ordinary View will want to reserve space for the
belief—even if conceded not to be a strictly knowledgeable belief—that the disjunctive conclusion is incorrect: that
nobody need be mistaken. Since there is no provision within an intuitionistic framework for a coherent denial of the
disjunction, it appears that the intuitionistic rescue cannot do justice to the Ordinary View. However a second, distinct
complaint is that the intuitionistic rescue cannot so much as provide for a coherent belief that rampant realism is
false—even if it were granted that such disbelief would involve a degree of presumption. Since the Ordinary View is
indeed inconsistent with rampant realism, the two complaints converge on the thought that the intuitionistic rescue
cannot do justice to the Ordinary View. Nevertheless the complaints are not the same—for the straightforward reason
that denial of the disjunction is not required by the denial of rampant realism.

What are their relations? Well, rampant realism is—presumably—committed to the disjunction; conversely, an
acceptance that the disjunction follows just from the premiss that Tim and I have contradictory views is, arguably, a
commitment to rampant realism. But that is not to say that only a framework that provides for a coherent denial of
the disjunction can provide for a coherent disbelief in the metaphysical and semantic postulations of rampant realism.
That

end p 50

would be true only if disbelief in rampant realism were a commitment to denying the disjunction. But that cannot be
correct: after all, both Tim and I accept the disjunction, presumably (since each thinks the other is mistaken)—but at
most one of us is a rampant realist about matters of basic taste in desserts!

It is not—the point is—an acceptance of the disjunction qua propositional content that commits to rampant realism; it
is an acceptance that its truth is ensured simply by the fact of Tim's and my respective views being contradictories. In
fact, anyone with a determinate—positive or negative—view on whether stewed rhubarb is delicious should accept the
 p p p p p y

disjunction; no philosophical commitment is entrained. A philosophical commitment is entered into only when one
regards the disjunction as imposed by the nature of the subject matter and the kind of content carried by claims of the
kind in dispute. One may therefore reject a rampant realist—indeed, any form of realist—view about those matters
without commitment to any particular attitude to the disjunction.

The second of the two complaints is accordingly misconceived. A supporter of the intuitionistic rescue is quite at
liberty to deny rampant realism. It is not true that he can go no further than agnosticism about the point. He thereby
denies that the truth of the disjunction is guaranteed in the way rampant realism supposes. The dialectical situation
is, in fact, exactly analogous to that in the philosophy of mathematics, where the intuitionist may quite coherently—if
he wishes—deny the Platonist metaphysics of a crystalline world of determinate mathematical structures, potentially
conferring truth and falsity upon our mathematical statements in ways transcending all possibility of proof. That denial
commits him to denying that Excluded Middle holds of necessity for reasons connected with that metaphysics. But it
does not commit him to denying Excluded Middle itself, still less any instance of it. Rather, in the absence of
justification for the principle of any other kind, he merely regards it as unacceptable.

The first complaint still stands, though: the thought that the intuitionistic rescue leaves no space for a coherent belief
that neither Tim nor I is mistaken in the original dispute. The closest the intuitionistic rescue gets to this is in
establishing a position from which it can be allowed that the presence of error is dictated neither by elementary logic
and the contradictoriness of the attitudes involved nor—I have just argued—by the semantics and metaphysics of
discourse of taste. So we save a negative modal claim: there doesn't have to be error for those reasons at least. But
we don't, it seems, give sense to the idea that there doesn't have to be error tout court, nor therefore provide any
possibility for someone coherently to believe that there isn't any error in the case in point. But wasn't that suggestion
just the force of the Latin proverb? Recall that we initially glossed the Ordinary View with the words, ‘Nobody's wrong.
Tim and I should just agree to disagree.’

It is easy, of course, to dismiss the idea that there is any such coherent belief, stronger than any of the beliefs that
the intuitionistic rescue can accommodate and still remaining to be made sense of. After all, Tim and I do disagree. So
Tim

end p 51

must think, presumably, that my view is mistaken. And I must think that his is mistaken. So someone who thinks that
nobody actually is mistaken is committed to disagreeing with us both—and so to regarding everybody as mistaken:
Tim, me, and indeed themselves! If there is a way further forward, it must consist in finding the means to deny that
Tim and I must, in fact, each regard the other's view as mistaken—this despite the fact that our views are genuinely
contradictory. So in a certain sense, their contradictoriness notwithstanding, we have to agree that our views are not
in conflict—that we do not disagree.

This extra step is inaccessible on the intuitionistic treatment, and it is unquestionably of interest to consider what kind
of position could possibly accommodate an insistence on it while avoiding aporia. One may well think that, for all we
have so far seen, the intuitionistic treatment delivers enough to rank as a satisfactory explication of the Ordinary
View. But if there is a stable account which manages the extra step—which can somehow allow that while Tim and I
have genuinely contradictory attitudes, neither of us need regard the other as mistaken—it may well be felt to offer
progress.

7 True Relativism
That is the prospectus that what I will call the True Relativist exegesis of the Ordinary View aims to fill. According to
true relativism, it can be the case that Tim and I are both right even though we understand the claim that rhubarb is
delicious in the same way, and even though we are making incompatible judgements about it. And the reason is
because there are no absolute facts about taste—what it is true to say about taste depends upon a stance, or a set of
standards, or a set of affective dispositions. The very same claim can be true for Tim and false for me—and that it is
so can be something that is available to us both.

Familiarly, the idea that truth is globally relative—that some form of relativity is of the nature of truth—has often
been held to implicate dialectical incoherence, or worse. Whatever the fact about that, our questions are more
specific: whether relative truth is even locally coherent; whether, if so, it can accommodate each of Contradiction,
Faultlessness and Sustainability at all; and whether it can do so without undue metaphysical cost, and in particular in
a way which allows for more robust understanding of Faultlessness than could be secured by the intuitionistic
proposal—a way which allows for a consistent profession that Tim's and my views can both be correct.

Obviously, in order to accomplish the last of these things, true relativism has to have the means to block the Simple
Deduction before it reaches the problematical line:

It's not the case that (it's not the case that Williamson is
mistaken and it's not the case that Wright is mistaken).

end p 52

It is clear how the attempt should be made. The true relativist must insist that, for statements of the kind that
 p p p p p y

concern us, we may no longer validly infer from the supposition that P that someone who holds that not-P is making a
mistake. A mistake will be implicated only if the judgement that not-P is held accountable to the same standards, or
perspective, or whatever, that are implicated in the (hypothetical) supposition that P is true. Very simply: if P is true
by one set of standards, or whatever the relativistic parameter is, and I judge it false by another, then what makes P
true need not be something which, in judging that it is not true, I mistakenly judge not to obtain.

That, then, will be the shape of the true relativist response to the Simple Deduction. The question is whether it can be
made sense of. There is a temptation to think that making sense of it is easier than it really is which we need to
expose straight away. A philosopher seeking to stabilize the Ordinary View should not be interested in relativity—as a
function of context of utterance, or whatever else—in the truth-conditions, and hence the truth-values, of sentences.
The relativity that needs to be made out is relativity in the truth of thoughts, or propositions. If we identify a
proposition by its truth-conditions, the relevant form of relativity is relativity in the question whether those very
truth-conditions are satisfied. Suppose that in the course of a medical procedure, a surgeon says of a scalpel that's
been poorly prepared: ‘This instrument is dangerously blunt.’ Later, when the instrument is about to be re-sharpened
and sterilized, his assistant may warn an inexperienced orderly: ‘Watch out when you handle that—it's dangerously
sharp.’ Granted, it would be crass to say that the surgeon and his assistant mean different things by ‘sharp’ and ‘blunt’
respectively. What is true is that there is a relativity of standard: the surgeon's needs require a much finer edge on
the blade than would suffice to justify his assistant's subsequent warning.

A similar set-up is illustrated by the kind of attributer-contextualist accounts of knowledge proposed by writers such
8
as Keith DeRose and Stewart Cohen.
8
See, for instance, DeRose (1992, 2002), and Cohen (1999).

The point to note, however, is that the kind of relativity involved in these examples—plausible in the case of the
scalpel, more controversial in the case of knowledge-contextualism—is not at all to our present purpose. For while it
would be crass to see them as involving anything comparable to simple ambiguity in ‘sharp’ or ‘knows’, they do
involve that the truth-conditions of ascriptions of sharpness and knowledge are so affected by contextual or other
relevant parameters that there is no single content respectively affirmed and denied by the surgeon's claim and that
of his assistant, or—to cut a long story short—by G. E. Moore's claim that he knows he has a hand and the sceptical
claim that he does not. These views might naturally, if perhaps a little loosely, be described as involving relativism
about sharpness, or knowledge. But true relativism is relativism about truth. It is not the thesis that the content of a
certain kind of ascription can vary as a function of varying standards, or contexts, or other

end p 53

parameters. That's a thesis that, applied to our present problem, simply gives up on the attempt to satisfy
Contradiction and so holds out no comfort to the Ordinary View. True relativism is the thesis—to repeat—that after
the truth-conditions of an utterance have been settled, there can be relativity in the question whether they are
satisfied. It is a thesis that engages at the level of content, rather than at the level of speech-acts. Or if it is not, then
it's merely a slightly more sophisticated cousin of the simple indexical relativist proposal I canvassed at the start—a
variant which holds that while a statement on which a dispute of inclination is targeted is indeed not an ellipsis for
something which explicitly mentions some parametric standard or perspective, etc., it is nevertheless something
whose content is implicitly fixed by reference to such a parameter, so that—as before—Tim and I will have no genuine
conflict of opinion about rhubarb. A true relativist accommodation of the Ordinary View must demand that it is the
very same proposition that Tim affirms and that I deny—and at the same time that neither the affirmation nor the
denial need be mistaken, with this a point which the antagonists themselves can coherently take on board. The latter
point is entirely unproblematical if it is not really the same proposition that is involved. What the relativist has to
explain, in contrast, is how to maintain the point alongside the claim that there is a single proposition affirmed and
denied respectively. What is the relevant notion of propositional identity, and how is it possible rationally to affirm the
truth of such a proposition consistently with allowing that someone else's denial of it is also true?

It is not, it seems to me, at all straightforward to see that the demanded notion of relative truth—relative truth at the
level of propositions—is fully intelligible. But the difficulties are especially daunting if we essay to think of truth as
correspondence, in a robust sense of correspondence with calls for an internal relation between a proposition,
conceived as an articulated abstract entity, and some correspondingly articulated aspect of non-propositional reality.
On any such picture of truth and truth-conferral, it seems impossible to make room for the additional parameter which
relativism posits; the internal structural relationship between propositions and the things that make them true or
false is so conceived as to be essentially dyadic. It's like the congruence in form between a head-and-shoulders
sculpture and the model who posed for it. No doubt the former may be an accurate representation, or not, relative to
the conventions of representation, but we are looking for something to illuminate an alleged relativity which bites
after the conventions of representation have been fixed. And we draw a blank. The unavoidable conclusion seems to
be that, while particular such conventions may allow of degrees of accuracy in representation, the degree to which
there is accuracy is something which supervenes entirely upon the respective physiognomies of the statue and the
9
sitter. There is no place for a third term in the relation.
9 As JC Beall points out in his contribution to this volume, this conclusion is good only if one assumes that a single world furnishes the facts to which the

propositions at issue are liable to (mis)-correspond. If sense could somehow be made of the Goodmanian figure of distinct worlds to correspond to

distinct sensibilities, Tim and I could each be thinking the literal (correspondence-) truth about our respective worlds.

end p 54
 p p p p p y

If that is correct, the immediate lesson to draw is merely the unremarkable one that to attempt to think of truth—
propositional truth—relativistically is to foreclose on thinking of it as correspondence. That's an objection to relativism
only if it's impossible to think of truth in any other viable way. Suppose on the contrary that, at least in some regions
of thought, truth may satisfyingly be construed as consisting in some kind of coherence relation, with coherence an
internal, analytic relationship, fixed by the content of the propositions among which it obtains. Let it be proposed, for
example, that the truth of a proposition consists in its participation within a maximal, coherent system of propositions
incorporating some specified base class of propositions, B. Then depending on the choice of B, a proposition may be
true or not—may be a member of the relevant maximal set of coherent propositions or not—even after its content is
fully fixed. Such a conception of truth may only locally have any attraction at all—one might, for example, think of
truth in pure set theory along such lines—but it provides at least a prima facie model of how a truth predicate for
10
propositions may intelligibly be conceived as relative.
10
For the credentials of coherence, so conceived, to count as truth, and for more general discussion of what it takes for a predicate to express the

concept of truth, see Wright (1998).

No such coherentist model is presumably wanted for the notion of truth that is to engage disputes of taste and other
matters of inclination. Still, the example suggests that once one begins to think of truth along the kind of pluralist
11
lines that a number of philosophers, myself included, have canvassed in recent work,
11 See especially Wright (1992, 1998); and Michael Lynch (2001, 2004).

12
it may be possible to come closer to a stable working-out of true relativism than one might otherwise suspect.
12 The tension between correspondence truth and relativistic truth, and indeed an implicit pluralism about truth, is actually close to the surface, it seems to

me, in several of the treatments canvassed in the recent revival of sympathetic discussion of relativism within analytical philosophy. John MacFarlane, for

example, in a growing series of important discussions (for instance, MacFarlane (2003, 2005 and forthcoming) has tended to promote a conception of the

conferrers of relativistic truth as consisting in ordered pairs of a world and a ‘context of assessment’. Likewise Max Kölbel (2002, 2004) proposes that

truth is relative to what he terms ‘perspective’—where the very word conjures the idea of an argument-place: something the perspective is a perspective

on. These proposals are, of course, partly formal: what may vary with variation in the context of assessment, or the perspective, is just whatever truth is

being conceived as relative to—perhaps standards, perhaps taste, perhaps information, perhaps time. But it is unintelligible what contribution a world is

supposed to make except as providing an input of unreconstructed states of affairs, things standing thus-and-so. The very intellig bility—even prima facie

intelligibility—of the kind of framework MacFarlane explicitly, and Kölbel implicitly, propose thus seems to call for a prior domain of circumstances—the kind

of thing a ‘world’ contributes to the ordered pair, or what a perspective is exercised on—of which, presumably, there is no obstacle in principle to an

independent self-standing statement. For such statements there will then be no need—and, on pain of vicious regress, ultimately no space—for a

relativistic conception of truth.

I'll conclude by outlining one specific suggestion in that direction.

end p 55

8 Relativism and Idealized Assertibility

13
Assertibility
13
I use ‘assertibility’, as is customary if a little unhappy, as a shorthand for warranted assertibility, where the relevant notion of warrant relates just to the

acceptance of the content asserted and has no other bearing on the justifiability of (publicly) asserting it.

is manifestly a relative notion: a statement may be assertible relative to one state of information and not to another.
Might notions of truth arrived at by idealization of assertibility retain this, or a kindred, relativity?

There are two principal such proposed idealizations to be found in the literature. The first, in the Peircean tradition
and associated with Hilary Putnam's latterly renounced ‘internal realism’, idealizes on the state of information: what is
true is what is assertible in a state of information incorporating all possible relevant data for the proposition in
question. It's obvious that this proposal, whatever we might want to say pro- or anti- the credentials of the resulting
truth predicate, holds out no interesting prospect of relativism, since the whole point of the idealization involved is
that it is supposed to ensure convergence. Either a proposition is assertible at the relevant Peircean limit of
information gathering—in which case it is true simpliciter—or, even at the limit, its credentials are matched by a rival,
in which case it is neither assertible nor—for internal realism—true.

Matters may turn out interestingly differently, however, if the idealization assumes the form proposed in the notion of
14
superassertibility.
14 See C. Wright (1992, 1998).

Superassertibility is the property not of being assertible in some ideal—perhaps limiting—state of information, but of
being assertible in some ordinary, accessible state of information and then remaining so no matter what additions or
improvements are made to it. When superassertibility for a given class of statements is taken to be truth, then truth
is held to consist not in assertibility at some ideal limit of information gathering but in enduring assertibility over
indefinite improvements. Does superassertibility offer the prospect of an interesting relativity? More specifically: can
this happen—that in a single world one thinker, Hero, is in position to accept P, and another, Heroine, is in position to
accept not-P, and that each can retain their respective situations no matter what improvements or enlargements are
made to their states of information?

Well, not if Hero's and Heroine's respective bodies of information allow of pooling, and if it is determinate and unique
what the resulting pooled state of information should be, and determinate whether it supports P, or not-P, or neither.
 p p p p p y

But those conditions may not all be met. When Hero and Heroine bring their respective bodies of information
together, it may be that there is more than one equally rationally defensible way for accommodating the components
into a unified state, each maybe involving some discards, with none superior to the others in virtue of the number or
kind of discards involved or the quality of the information remaining. It may also happen that some of the resulting

end p 56

enlarged states of information continue to warrant acceptance of P, and others acceptance of not-P. And once granted
to be possible at all, it's difficult to see how to exclude the thought that such a situation might persist indefinitely. In
that case superassertibility would be relative to a starting point, an initial basis for acceptance or rejection. If one
were satisfied there were no other obstacles to the identification of truth with superassertibility over the region of
discourse in question, that would be a kind of relativity of truth.

However, the kind of case which is our main focus—disputes of taste—is marked by the following peculiarity: that the
basic form of assertibility condition for statements of the relevant kind is given by a subject's finding herself in a
certain type of non-cognitive affective state: liking the taste of rhubarb, for instance. The basic form of assertibility
condition, that is to say, for the impersonal statement—about the vegetable—coincides with that for the self-ascription
of a subjective state that is not conceived—at least not by anyone attracted to the Ordinary View—as a cognitive
response. In that case Hero and Heroine may respectively be in a position to assert P and to assert not-P, not because
they possess differing initial information bases but just by virtue of differing in their non-cognitive responses to
things—and because these responses are non-cognitive, there will be no clear sense to the idea of ’pooling’ their
respective starting points and determining what is warranted by the result. Of course there is such a thing as
enlarging one's information by the addition of the datum that others do not share a particular non-cognitive response.
But if that datum is not treated per se as a defeater, then there will be no immediate threat to the superassertibility
of the original claim.

Much more would need to be said if a satisfying proposal in this direction is to be developed. In particular, if a content
is to be associated with the impersonal statement—‘Rhubarb is delicious’—contrasting with that of a subjective report,
then something has to be said about how the contrast between the two is sustained. Presumably such an account will
give central place to asymmetries in the conditions of defeat, with the assertibility of, for example, ‘I relish eating
rhubarb,’ surviving in circumstances where that of ‘Rhubarb is delicious’ is lost. It's hard to envisage how the story
might plausibly go without some kind of play with intersubjective accord: what purposes could be served by our
having the impersonal form of statement if one could seldom reliably encourage expectations in an audience about
their own affective states and responses? Still, if one's own tastes are not too idiosyncratic—if enough of a
constituency goes along with them—then that may be enough to license a claim, even if significantly many may, with
the same license, dissent from it. And in that case there may be theoretical advantages in representing the situation
as one in which conflicting claims are each true relative to varying parameters of taste, with truth construed as
superassertibility on the basis of a notion of assertibility grounded on the relevant non-cognitive affect.

end p 57

Such a proposal looks to be promisingly placed to handle Faultlessness and Sustainability. But matters may seem less
clear with Contradiction—the claim that genuinely incompatible opinions are involved: how exactly does the proposal
promise a better accommodation of the Ordinary View in this respect than the kind of position, illustrated by the
examples of the blunt scalpel and knowledge-attributions when construed along contextualist lines, which effectively
diagnoses disputes of inclination as illusory? What can be said, in the spirit of the superassertibilist-relativist proposal,
to support the idea that it is the same content that, as it may be, is superassertible for Tim but not for me?

To think of truth in some area of discourse as constituted by superassertibility no doubt leaves considerable latitude
when it comes to theorizing about propositional content. I shall not here attempt such a theory. However, if Tim and I
do have an understanding in common of the proposition that rhubarb is delicious, as it occurs in our respective
affirmation and denial, it would be natural to locate the commonality in a shared conception of basic, sufficient—if
defeasible—grounds for accepting the proposition (one's enjoying rhubarb, presumably) and a shared conception of
the consequences of regarding it as correct. Among the latter might be, for example, the desirability of regular
harvesting of one's rhubarb crop when in season, a high ranking for choosing a dessert in a German restaurant
identified to one as rhubarb crumble, a high priority assigned to the rhubarb patch in the reorganization of the
vegetable garden, and so on. Commonality of understanding will involve that my negative view, by contrast, will lead
to corresponding low priorities and opposed choices. This is the pre-theoretic background against which it seems
intuitive to say that Tim and I have genuinely conflicting views about a single proposition. An explicit theory
subserving the point would be one in the broad tradition deriving from Gentzen's work on the logical constants which
locates the individuation of content in canonical grounds and consequences.

Against this kind of background, it's salient that the situation contrasts with the case of the rejected scalpel. Baldly,
suitable grounds for the attribution of sharpness that the surgeon denies would be quite different to those sufficient
for the attribution of sharpness that the orderly affirms. When the latter asserts that the scalpel is (dangerously)
sharp he is not challenging the surgeon's judgement that it is not—as indeed the surgeon is not challenging the
orderly's judgement that great care is necessary in handling it and preparing it for sharpening and sterilization. But
more: each can quite coherently accept and, in various ways, appropriately act on the other's claim while still
maintaining his own—surely a conclusive consideration in favour of the point that different, and compatible, contents
 p p p p p y

are involved. By contrast, that Tim and I are involved in genuine disagreement is borne out by the fact that we agree
about the, loosely described, consequences of each other's views and then sustain our disagreement through our
respective acceptance or rejection of those consequences and the courses of action involved. Tim orders the crumble;
I don't. Tim designs his vegetable patch

end p 58

15
in a certain way; I don't.
15
Notice that attributor contextualism—which proposes to construe ‘x knows that p’ as comparable to ‘that instrument is dangerously sharp’, rather than

to ‘rhubarb is delicious’, understood as by the Ordinary View—has work to do with this point. A third party can accept not merely that the surgeon's and

the orderly's claims are both correct in context: she can, as it were, take both claims on board—indeed the orderly does so, in effect, by replacing the

knife with a better prepared one for the purposes of the surgery and then taking appropriate personal care while he sharpens and steriliz es the rejected

knife. But what would it be to take on board the claims both of G. E. Moore and a Sceptic: how would one act out a simultaneous acceptance of both

claims?

Rational action on either of the views excludes rational action on the other.

So here's the package: Tim and I are in genuine disagreement about whether rhubarb is delicious. Our opinions are
incompatible. And the common understanding, necessary to ground that incompatibility, is based on a common
conception of the assertibility conditions of the claim—that, absent defeating considerations, it may be asserted just if
one relishes eating rhubarb—and on a shared conception of a range of consequences, both analytical and practical,
which attend its correctness. Our disagreement can be faultless because it can be based on our respectively perfectly
proper responses to our respective non-cognitive propensities. And it can be sustainable because—precisely—neither
claim has been defeated nor has to be defeasible. Finally, the Simple Deduction is blocked in exactly the way
prefigured: when truth is conceived as superassertibility relative to a subject's non-cognitive responses, the
supposition that P is true will be answerable to the corresponding responses of a tacitly understood constituency of
subjects; and it will implicate a mistake in the opinion of one who takes it that not-P is true only if their opinion is
properly held answerable to the responses of the same constituency.

If all this is soundly conceived, then a relativism about truth, fashioned along the indicated lines, may be the natural
companion of non-cognitivist conceptions of competence in particular regions of discourse. But here I must be content
merely to have outlined the approach.

References

Cohen, Stewart (1999) ‘Contextualism, Skepticism and the Structure of Reasons’, Philosophical Perspectives 13:
57–89.

DeRose, Keith (1992) ‘Contextualism and Knowledge Attributions’, Philosophy and Phenomenological Research 52:
913–29.

—— (2002) ‘Assertion, Knowledge, and Context’, Philosophical Review 111: 167–203.

Kölbel, Max (2002) Truth Without Objectivity, London: Routledge.

—— (2004) ‘Faultless Disagreement’, Proceedings of the Aristotelian Society 104: 53–73.

—— (Unpublished) ‘Wright on Disputes of Inclination’, available as a PDF file at: http://www.philosophy.bham.ac.uk

/staff/Kölbel.htm .

Lynch, Michael (2001) ‘A Functionalist Theory of Truth’, in M. Lynch (ed.), The Nature of Truth, Cambridge, Mass.:
Bradford/MIT, pp. 723–49.

end p 59

Lynch, Michael (2004) ‘Truth and Multiple Realisability’, Australasian Journal of Philosophy 82: 384–408.

MacFarlane, John (2003) ‘Future Contingents and Relative Truth’, Philosophical Quarterly 53: 321–36.

—— (2005) ‘Making Sense of Relative Truth’, Proceedings of the Aristotelian Society 105: 321–39.

—— (Forthcoming) ‘The Assessment-Sensitivity of Knowledge Attributions’, in Tamar Szabó Gendler and John
Hawthorne (eds), Oxford Studies in Epistemology, Oxford: Oxford University Press.

Wright, Crispin (1992) Truth and Objectivity, Cambridge, Mass.: Harvard University Press.

—— (1998) ‘Truth: A Traditional Debate Reviewed’, in supplementary volume 24 (1998) of the Canadian Journal of
Philosophy on Pragmatism, guest edited by Cheryl Misak, pp. 31–74.

—— (2001) ‘On Being in a Quandary: Relativism, Vagueness, Logical Revisionism’, Mind CX: 45–98.

—— (2002) ‘Relativism and Classical Logic’, in Logic, Language and Thought, ed. Anthony O' Hear, Cambridge,
Cambridge University Press, pp. 95–118.

end p 60
 p p p p p y

This page intentionally left blank.

 p p p p p y

3 Modelling the ‘Ordinary View’ *

For useful discussion I thank Patrick Greenough, Michael Lynch, Daniel Nolan, Crispin Wright, and
various attendees at the 2004 St. Andrews Truth and Realism conference.

Abstract. This paper is a response to Crispin Wright's attempt to model (what he calls) ‘the ordinary view’ of ‘disputes
of inclination’. Familiarity with Wright's paper (Chapter 2 of this volume) is assumed. I propose and briefly discuss two
models that Wright neglects, a (non-relative) paraconsistent model and a version of truth-relativism where truth is
correspondence.

I Taste-Function Relativism
Consider the following (apparent) dispute:

B RUCE: Vegemite is delicious. JOEY : Vegemite is not delicious.

I believe that the most natural response to this apparent dispute is to treat it as merely apparent, and indeed invoke
some sort of relativism—parameterization—with respect to ‘is delicious’.

The natural response invokes a ‘taste-function’, as it were, which takes some sort of input—say, Vegemite—and yields
a value (which we can take to be a natural number). Simple taste-function relativism maintains that each person has
such a taste-function, and ‘is delicious’ contains an implicit parameter over taste-functions:

‘x is delicioust’ is satisfied exactly if t(x)=n where n m for some threshold m.

In turn, assertibility conditions—which are relative to a state of information or, more generally, a state (or agent)
—likewise invoke such taste-functions (and some threshold m):

that Vegemite is delicioust is assertible by an agent b exactly if b's taste-function t is such that t(Vegemite)=n
(for n m).

end p 61

As a first go, such taste-function relativism, it seems to me, is a viable approach to ‘disputes of taste’—and ‘disputes of
humour’, and so on. No doubt, more needs to be said, but the general idea is clear enough and, it seems to me, fairly
plausible. But Crispin Wright (p. 39) objects.

O BJECTION: That sort of simple relativism about taste clashes against the challenge implicit in what is otherwise
a sensible question, namely: If, as you say, Vegemite is delicious, how come nobody at this conference but you
likes it? Such a question—on the proposed (taste-function) relativism—betrays incompetence (a failure to
recognize the hidden parameter), but such alleged incompetence seems not to be present in such ordinary
questions. REPLY : The given challenge (question) needn't betray incompetence; it can betray a ‘limit sense’ of
taste-relative terms. In particular, it may well be that there's a use of ‘delicious’—an absolute but nonetheless
indexed usage—according to which:

‘x is delicioust’ is satisfied exactly if t (x)=n (where n m for some threshold m) for all ‘accessible’ taste-
functions t , where ‘accessible’ can be cashed out in some standard contextualist fashion.

In a context in which such a ‘limit sense’ of ‘delicious’ is in play—or is taken to be in play by the parties in the
conversation—the given question is straightforwardly competent.

For all that Wright has said, I do not see why such simple taste-function relativism isn't an appropriate account of
‘matters of taste’—or humour or whatnot. But for now, it is best to put Wright's chief task squarely on the table.

2 Wright's Aim: The Ordinary View

Whether taste-function relativism (or some variant) is ultimately the best approach to ‘disputes of taste’ is in many
ways beside Wright's chief task. Wright's task, I take it, is not to find the most plausible account of ‘disputes of taste’
but, rather, to find a ‘coherent’ account of the ‘Ordinary View’ of such (apparent) disputes. What Wright calls ‘disputes
of inclination’ constitute the target of the ‘Ordinary View’, and whether (apparent) disputes of taste are ultimately
disputes of inclination is an open question.

Wright's main task is to give a plausible account—a plausible model—of the Ordinary View, where the Ordinary View
essentially involves four features:
• Existence: There are ‘disputes of inclination’.
• Contradiction: Such disputes involve genuinely incompatible attitudes (and the conjunction of what the disputants
believe is a formal contradiction).

end p 62

• Faultlessness: Nobody need be mistaken in such disputes.

 p p p p p y

• Sustainability: The antagonists may, perfectly rationally, stick to their respective views even after the disagreement
comes to light and impresses (them) as intractable.

Wright's task, then, is simply cashing out the Ordinary View in a ‘coherent’ fashion—either in a non-parametric or a
1
parametric (relativistic) fashion.
1 Some might think that Wright's task is thereby the task of cashing out a coherent version of truth-relativism, but Wright maintains that truth-relativism is

actually a theoretical attempt to cash out the Ordinary View, and so it is worth exploring non-parametric accounts of the Ordinary View.

The given task is constrained by (what Wright calls) the simple deduction, where a and b are agents and, letting
context dictate use-mention, ¬ is the negation of :
1. a accepts [Assumption]
2. b accepts ¬ [Assumption]
3. a and b's disagreement involves no mistake [Assumption]
4. [Assumption]
5. b is guilty of a mistake [2,4]
6. ¬ [3–5, Reductio]
7. a is guilty of a mistake [1,6]
8. The negation of (3) is true [3,5,7].
Note that the conclusion of the simple deduction, at least as I understand it, is:
9. Either a is mistaken or b is mistaken.
And (9) is supposed to follow directly from (8), which is intended to have the form:

¬(¬Ma ¬Mb)
2
with ¬Ma ¬Mb being the (intended) form of (3).
2
Think of M as is mistaken (with respect to ).

So understood, the simple deduction seems to bar the Ordinary View from coherent formulation, apparently showing
that at least one of the disputants is mistaken, contrary to ‘faultlessness’. The constraint, then, is to block the simple
deduction, so that both contradiction and faultlessness may stand up.

Towards cashing out the Ordinary View, Wright proposes two approaches, the first a non-parametric (intuitionistic)
approach, the second a parametric one. I briefly comment on each proposal, but my main aim is to propose a few
options that Wright neglects.

end p 63

3 Intuitionistic Approach
If the task is to give a coherent model of some view, then one should not prejudge the issue with one's favoured
logical theory. Suppose, for example, that we are trying to give a coherent but accurate model of inconsistent fictions.
In that case, a natural first step would not invoke classical logic (or, for that matter, intuitionistic logic) but, rather,
some version or other of paraconsistent logic. In the present case—namely, modelling the ‘Ordinary View’—Wright
pursues an intuitionistic framework.

Wright's intuitionistic response blocks the Simple Deduction by blocking the step from (8) to (9): the step from
¬(¬Ma ¬Mb) to Ma Mb is intuitionistically invalid. Moreover, the intuitionistic model of the Ordinary View can be
motivated in ways that Wright discusses—taking ‘disputes of inclination’ to be situations in which neither party has
suitable warrant for claiming that the other's view is false (or untrue). (See Chapter 2 for further discussion.) Another
virtue of the approach is that it doesn't require a truth-relativism or, for that matter, any significant relativism at all.
The approach is one that cashes out the Ordinary View as a non-relativistic view—a non-parametric view.

Despite its virtues, the intuionistic model confronts problems. The proposal's chief problem, as Wright himself points
out, is (in effect) an explanatory one. We want an explanation of how the disputants can be faultless when they
believe ‘contradictory’ claims, of how they are ‘faultless’ when they have at least contrary attitudes towards one and
the same claim (or proposition, and so on). What is it about such special claims/propositions that affords faultlessness
(and sustainability)? The most natural answer invokes absence of ‘mistake-makers’ (if you will)—the absence of any
‘fact of the matter’ that, as it were, would otherwise make a mistake of your belief. The trouble, as Wright points out,
is that the intuitionistic proposal cannot take that line, since it cannot make the claim that neither party's (given)
3
belief is untrue—the claim equivalent to (3) in the simple deduction.
3
Actually, it isn't clear to me that the intuitionistic line is stuck here. What one needs is some negation-like device that serves—perhaps in concert with

intuitionistic negation—to express ‘no fact of the matter’ but for which some step in the ‘simple deduction’ fa ls (e g., reductio). I suspect that such a

device can be—perhaps has been—constructed, but I will not pursue it here.

What would be better is a model of the Ordinary View that not only blocks the simple deduction but also allows for the
‘no fact of the matter’ explanation of faultlessness. That, in the end, is what Wright's parametric proposal—his ‘true
relativism’—attempts to achieve. But before turning to that proposal, I think it's worth noting a non-parametric option
that Wright neglects—a paraconsistent option.

end p 64
 p p p p p y

4 Paraconsistent Proposal(s)
Recall the challenge of Wright's main task. As he puts it:

The challenge [is] to harmonize the three ingredients—Contradiction, Faultlessness, and Sustainability. And
the point hasn't gone away that if it is insisted that a dispute can be regarded as fault-free only if it's open to
us to suppose that each antagonist has a correct view, then a mere acceptance that the dispute is genuine—so
involves contradictory opinions—precludes regarding it as fault-free. Punkt. (p. 46)

Harmonizing the three ingredients, it seems to me, immediately suggests that the most faithful, natural model of the
Ordinary View is a paraconsistent one.

Recall the task: namely, to give a coherent model of the Ordinary View, one that ‘harmonizes the three ingredients’.
‘Coherent’ need not mean consistent, as far as I can see; it needs to be non-trivial, but needn't be entirely consistent.
And that is what paraconsistency promises: inconsistent but non-trivial theories.

A distinction is worth drawing: weak paraconsistentists

are those who take paraconsistent logics seriously as a means of modelling a particular domain of discourse (e.g.,
naive semantic theories, etc.); they do not accept that such models reflect genuine possibilities, certainly not
actualities. Strong paraconsistentists, on the other hand, are those who believe that such inconsistencies are
genuinely possible—that there may well be truths with true negations. (Strong paraconsistentists are often
dialetheists, those who think that there's some truth the negation of which is true.) For present purposes, Wright's
task calls only for weak paraconsistency—merely modelling the Ordinary View in an inconsistent fashion. Towards that
4
end, two approaches immediately suggest themselves. I (very) briefly consider each in turn.
4
Actually, there are many approaches that suggest themselves, including the Brandom–Rescher approach (Brandon and Rescher 1979) to inconsistent

discourse, as we l as the Da Costa C-systems (Da Costa and Alves 1977), and perhaps especially Batens' so-called ‘adaptive paraconsistent systems’

(References: 3). And others. See also Brown's discussion (Brown 2002).

4.1 Simple Paraconsistent Semantics

For present purposes, we can concentrate on a single (and simple) paraconsistent logic. Let ={1,.5,0} comprise our
semantic values, and let our ‘designated values’ be ={1,.5}. Concentrating on the propositional level, we take
valuations (or interpretations) to be functions from the atomics into , and then extend such valuations along
so-called Strong Kleene lines (see Table 3.1).

With interpretations in hand, a (semantic) consequence relation is defined in the usual (many-valued) way, where
comprises sentences of the language:

end p 65

Table 3.1 K 3 and LP Operator Diagrams

¬ 1 .5 0 1 .5 0
1 0 1 1 .5 0 1 1 1 1
.5 .5 .5 .5 .5 0 .5 1 .5 .5
0 1 0 0 0 0 0 1 .5 0

A iff for any interpretation , (A) if (B) for all B in .

In turn, valid sentences (or logical truths) may be defined in terms of by saying that A is valid iff A.

This is what Priest calls ‘the logic of paradox’ (LP) (Priest 1979), which is just the ‘gap’-free fragment of Anderson and
Belnap's four-valued semantics (Anderson and Belnap 1975). (One ‘transforms’ Strong Kleene into LP by designating
the third value.)

For present purposes, the pressing question is not so much the logic (which will suffice with respect to blocking the
simple deduction) but, rather, the philosophical story—the extent to which the Ordinary View can be ‘coherently
formulated’ along such paraconsistent lines.

4.2 Dialetheic Story

One obvious way to get the three ingredients is via a dialetheic model according to which ‘judgements of taste’ are—to
use Kit Fine's term—gluts, that ‘Vegemite is delicious’ and its negation are true. In that case, we have the ingredient
of ‘contradiction’ in the sense that conjoining the disputants' beliefs yields something of the form ¬ , and the
disputants take conflicting attitudes towards each of the conjuncts. Moreover, we have a straightforward sense of
faultlessness and sustainability: both disputants have true beliefs, and so such beliefs may rationally be sustained.

One problem with the dialetheic model is that it, like Wright's intuitionistic proposal, seems unable to ground
faultlessness in the ‘absence of mistake-makers’, in there being ‘no fact of the matter’ with respect to matters of taste
(or humour, and so on). After all, if both disputants have true beliefs, then, at least in some minimal sense, there's a
fact of the matter ‘grounding’ such true beliefs—at the very least, that Vegemite is delicious and that it's not!

The dialetheic model (logic plus philosophical story) improves on Wright's intuitionistic model. The former, but not the
 p p p p p y

latter, has a straightforward explanation of how ‘beliefs of inclination’ (as it were) can be faultless and, in turn,
sustainable: they're true. (They're also false on the dialetheic line, but that doesn't take away from their truth!) But
both models, as above, fail to achieve the desired ‘no fact of the matter’ feature of the Ordinary View. While that
feature is not

end p 66

essential to the Ordinary View (at least as Wright defines it), it's nonetheless a desideratum. On that score, the
dialetheic model fails.

4.3 Analetheic Interpretation

If we want to employ LP but also have some sense in which ‘there's no fact of the matter’ with respect to the given
class of claims, then we might want to take a different route. In particular, we could take what has been called an
analetheic route (Beall and Ripley 2004), where the analetheist thinks that some ‘untrue’ claims—in this case, ones
for which there's no fact of the matter—are properly assertible.

With respect to LP, dialetheists (informally) read the middle value as both true and false. For our analetheic purposes,
we can (informally) construe the third value as at least not false—or, if you wish, mistake-maker-free.

The idea, in a nutshell, is just this: some claims—for example, those involved in ‘disputes of inclination’—are such that
there's no fact of the matter, and so such claims are at least not false. On the analetheic model, Ordinary Viewers
—those who embrace the Ordinary View—drop the familiar dictum that one ought (rationally) only believe what is
true; the Ordinary View's (analetheic) dictum is:

A NALETHEIC DICTUM: One ought (rationally) only believe what is at least not false.

Coupled with the dictum is a constraint on consistency:

5
DEFAULT CONSISTENCY : One ought (rationally) to minimize inconsistent beliefs.
5
Incidentally, this can—and, when properly filled out, would—be cashed out in terms of what Batens (Batens 2000) calls an ‘adaptive logic’, a

non-monotonic paraconsistent logic. Such adaptive logics provide a clear sense to the idea of ‘minimizing inconsistent be iefs’.

With those two principles, the analetheic model of the Ordinary View has many virtues. Indeed, the virtues of such a
paraconsistent account are as many as the desiderata that Wright propounds:
V1. It upholds all three essential features in a straightforward way:
(a) Contradiction: The disputants’ beliefs are ‘in contradiction’ both in the formal sense and in that the disputants
have ‘opposing’ attitudes towards one and the same claim.
(b) Faultlessness: There is no fault, since there's simply no fact of the matter to make for mistakes. The given
beliefs are at least not false. (Recall the analetheic dictum.)
(c) Sustainability: Since there's no fault, there's nothing blocking sustainability. (Recall the dictum.)
V2. It blocks ‘the simple deduction’ exactly as one would expect: Reductio is invalid over the target class of claims.

end p 67

V3. It gives flat-footed content to the falsity of ‘Rampant Realism’. The Rampant Realist, as Wright construes them,
thinks that there is some fact of the matter that ‘makes true’ or ‘makes false’ the ‘claims of inclination’, but there
is no such fact, according to the analetheist. What makes such claims assertible is that they are at least not false.

Note that I'm not arguing for analetheism with respect to ‘disputes of inclination’. I'm suggesting only that—given
Wright's task—analetheism is a viable model of the Ordinary View (assuming there is such a view), and also that, at
least at first blush, it seems to be superior to the intuitionistic model. But I'll leave that for further debate. For now,
the tentative conclusion is that if we want a non-parametric approach, we ought to pursue a paraconsistent (and, it
seems, analetheic) option.

The question is: why not explore a parametric option? Wright does just that.

5 True Relativism: Superassertibility

Given the problem confronting his intuitionistic (non-parametric) model, Wright pursues a relativistic option—‘true
relativism’—where the aim is to get relativism about propositional truth. Wright suggests that there's no hope in
6
getting relativism about propositional truth if truth is construed along ‘robust correspondence’.
6 But see § 6 of this chapter.

But that is not the end of the project. If, as Wright suggests, we enjoy a truth pluralism, we can construe truth as
something other than correspondence.

Wright suggests that superassertability is a promising option. While superassertibility is clear enough (just think of
Kripke-constructions for intuitionistic logic), I am not entirely clear on the overall proposal. In particular, it isn't clear
how the account—aside from explicitly invoking superassertibility—differs from a simple taste-function relativism that
indexes ‘true’ instead of ‘delicious’.

The debate, which I here leave open, will turn on what superassertibility adds to the Ordinary View that a crude taste-
 p p p p p y

function truth relativism doesn't. For now, a simple taste-function truth relativism would run thus, simply shifting the
parameter from ‘is delicious’ to ‘is true’:

that Vegemite is delicious is truet exactly if t (Vegemite)=n (for n m, threshold m), where ‘truet’ abbreviates
‘true relative to taste-function t’.

How, aside from the differences that may arise from the semantics of ‘truet’ and ‘superassertibility’, is Wright's
account different from above? The answer isn't clear, at least not without further details of ‘true relativism’. Wright's
‘true relativism’ is supposed to afford:

end p 68

representing [a dispute of inclination] as one in which conflicting claims are each true relative to certain
parameters of taste, with truth construed as superassertibility on the basis of a notion of assertibility grounded
on the relevant non-cognitive affect. (p. 57)

Again, aside from explicitly invoking superassertibility, such a representation is exactly as the simple taste-function
relativism suggests—where the ‘parameter of taste’ is simply a taste-function, and assertibility will be tied to the value
of such a function (some of that value, perhaps, being determined in part by ‘non-cognitive affect’).

One place of substantial difference—again, aside from the difference of explicitly invoking superassertibility—might
arise with respect to ‘Contradiction’. How does Wright's ‘true relativist’ achieve ‘Contradiction’? The idea is that the
disputants reflect conflicting attitudes towards the same proposition by assigning different priorities—different
values—to the ‘consequences of regarding it as correct’ (p. 58). For example, one who understands the assertibility-
conditions of ‘Vegemite is delicious’ and who ‘regards it as correct’ will assign high values to practices Down Under
(whatever those practices may be), while one who ‘regards it as incorrect’ will assign low values to such practices. But
what, now, is ‘regarding as correct’? As far as I can see, it is little more than ‘true relative to a taste-function’—true
relative to some (perhaps non-cognitive) assignment of (if you will) ‘taste values’.

If, as I've suggested, simple taste-function truth-relativism differs from Wright's ‘true relativism’ only in that the latter
invokes superassertibility, the merits of the two must be measured on the basis of what superassertibility offers over
simple (taste-function) truth-relativism. On the surface, superassertibility seems a bit more complicated than the
simple (and, indeed, rather crude) truth-relativism tied to a taste-function, but debate will tell whether it affords
significant virtues over the simple-minded sort of relativism. For present purposes, I leave that issue open. I shall turn
to the idea that Wright dismisses: namely, modelling the Ordinary View via a truth-relativism where truth is
propositional, correspondence truth.

6 Relative Correspondence Truth

If there is an ‘Ordinary View’, presumably it is tied to correspondence inasmuch as ‘ordinary intuitions’—before they're
properly tutored in the glories of disquotationalism—tend towards correspondence. Wright (p. 54) claims that the
prospects of achieving relative propositional truth, where truth is robust correspondence, are dim. I suggest that
Wright is wrong on that score. I suggest that there is a route towards achieving relativism about propositional
correspondence truth, one that—suitably tweaked—respects all ingredients of the Ordinary View.

end p 69

The Rampant Realist, as the Ordinary Viewers have her, is one who thinks that there are determinate facts about
‘deliciousness’ just as there are with our best scientific facts. Ordinary Viewers reject that there are such facts—at
least, they're not like those facts, the facts that make our thoughts true independently of what we think. But the
trouble with Ordinary Viewers is that they also think that thoughts about deliciousness are, in some sense, true—it
just depends, a bit, on what we think (or on features of us). The task, as Wright set it out, is to make sense of all this.
In the present context, the task is to get relativism about propositional truth but construe truth as ‘robust
correspondence’.

6.1 The Polarity View

One way to respect the correspondence intuition is via what I've elsewhere called the ‘Polarity View’ (Beall 2000).
Physicists posit all sorts of polarities, and metaphysicians—not that I'm really one of them, I should point out—are
sometimes pulled to do the same. The Polarity View is a view of truthmakers; it posits a ‘positive polarity’ and a
‘negative polarity’, and is perhaps best motivated by concerns over ‘negative truthmakers’ (but I shall not dwell on
that here).
7
The Polarity View can be modelled in a simple way, and it will help to have a basic picture.
7 I borrow this from earlier work (Beall 2000).

Ancestors of the model include van Fraassen's atomic facts (van Fraassen 1969), and more recently the situation
semantics of Barwise and Perry (1983). The model is straightforward. Reality, , comprises a set of properties and
relations, , a set of objects, , and a set of polarities, ={1,0}. Each property rn has a degree, which is
represented by n in ‘rn’. From these ingredients come atomic facts:
 p p p p p y

rn, d1 ,. . ., dn, i

where rn , and d1 ,. . .,dn , and . Intuitively, rn,d1 ,. . .,dn,1 is the fact that d1,. . .,dn are rn-related;
rn,d1,. . .,dn,0 is the fact that d1,. . .,dn are not rn-related. rn,d1 ,. . .,dn,i is a positive fact if and only if i=1;
otherwise, rn,d1 ,. . .,dn,i is a negative fact.

From here, we define what it is for statements to be true in reality or false in reality. Suppose that Pn is an n-place
predicate and c1 ,. . .,cn are singular terms. We let (Pn) be an element of and let (cj) be in . Then the sentence
Pnc1 ,. . .,cn is true in reality if and only if reality comprises the following fact:

(Pn), (c1),. . ., (cn), 1 ;

the given sentence is false in reality if and only if reality comprises the following fact:

(Pn), (c1),. . ., (cn), 0 .

end p.70

Once these atomic sentences have truth-values the compound sentences gain truth-values in the usual way. Letting
8
T and F stand for true in reality and false in reality, respectively, we have the familiar clauses:
8
The clauses for conjunction are the usual dual ones.

T¬ iff F

F¬ iff T

T iff T or T

F iff F and F

So goes the formal picture. The informal story is equally straightforward. In short, truth consists in correspondence
with truthmakers, where the truthmakers are positive and negative facts.

With truth and facts—or ‘states of the world’—so conceived, we can characterize the Rampant Realist as thinking that
all facts are ‘absolutely fixed’ in the sense that the given polarities are fixed independently of us. Truth is simply
correspondence to the facts, with truth of negations being true in virtue of negative facts. There's no call from the
Rampant Realists for ‘relative truth’. The call, as above, comes from Ordinary Viewers.

6.2 Relatively Positive/Negative States

Ordinary Viewers agree that truth is simply correspondence; however, as Wright suggests, they nonetheless call for a
bit of relativity with respect to such (propositional) truth. They can get it, I suggest, by recognizing—for lack of a
better term—relatively positive and relatively negative ‘states’ of the world.

The idea, in short, is that the states making ‘claims (propositions) of inclination’ true (false) are themselves relative to
(say) taste-functions—or some such function that assigns either a positive or negative polarity to the state. In other
words, Ordinary Viewers recognize a proper sub-class of ‘states’—the ones corresponding to ‘claims of inclination’—the
polarity of which is relative to taste-functions.

There are various ways of tweaking the polarity story to add such relativity, and the best account will be a matter of
9
future debate (should there be any interest at all!).
9
After seeing my initial proposal, which is below, Daniel Nolan and Crispin Wright both suggested alternatives, each carrying potential virtues with respect

to modelling the Ordinary View. (Given space considerations, I'll limit the discussion to just one proposal.)

One route, for example, might be to add new polarities that somehow reflect the requisite relativity. For present
purposes, I will add no new polarities but, rather, ‘polarity maps’. The idea runs as follows.

As before, (reality) comprises a set of properties and relations, a set of objects, and a set ={1,0} of polarities.
The difference is that we now also add

end p.71

‘polarity maps’ that comprise pairs t,p , where t is a taste-function and p . States of the world, then, have the
structure:

rn, d1 ,. . ., dn, i

where, now, i (a polarity) or i= (a polarity map). When i=1 we call the state an absolutely positive fact, and
similarly an absolutely negative fact when i=0. When i= , such states are neither positive facts nor negative facts
independently of a given taste-function; they are (as it were) ‘neutral states’ that are (relative) facts—have a
polarity—only relative to some taste-function or other.

The difference between Rampant Realists and Ordinary Viewers is that the former acknowledge no polarity maps in
‘facts’, the latter do. The Rampant Realist thinks that every fact is absolutely positive or absolutely negative; she
needn't be told about ‘taste-functions’ or the like. Ordinary Viewers are different: for some facts, there's no sense to
the question of whether the facts are positive or negative—at least not until you've specified a given taste-function.
Inasmuch as some such facts are relative (to taste-functions), truth—being correspondence—is similarly relative, but
 p p p p p y

it's ‘robust correspondence truth’ for all that.

6.3 The Ordinary View

Whether such an account of relative correspondence truth (with respect to propositions) affords the best model of the
Ordinary View is something I leave open, but I think it's fairly clear that the account affords all three ingredients of
the Ordinary View in a coherent way:
10
• Contradiction: both disputants have incompatible beliefs with respect to the same proposition.
10 To make the incompatibility ‘really’ plain, stipulate that for no do we have and
T F , thereby making the ‘falsity clauses’ in § 6.1

redundant. (This adjustment forces explosion, ex falso quodlibet.)

• Fautlessness: so long as disputants aren't at fault for having their respective taste-functions, fault is hard to press,
in general. After all, we have relative truth, and each disputant's belief is true relative to the way the world is
11
(which, again, is in part a function of their individual taste-functions).
11 Along this vein, the colloquial sense of Wright's quoted proverb—de gustibus non est disputandum—is on target: there's no accounting for tastes.

• Sustainability: short of changing their respective taste-functions, there seems to be little challenge against
sustainability.

The proposed ‘relative correspondence’ model, I think, is promising. On the other hand, recalling the trouble with
Wright's intuitionistic model and, similarly, the dialetheic model, one might worry about the ‘no fact of the matter’
12
desideratum. In the current context, all ‘claims of taste’ are true or false,
12 Here, I'm assuming that we stipulate as much with respect to
T and F.

and so

end p.72

there is a fact of the matter with respect to disputes of inclination. Nonetheless, the desideratum is achieved, or at
least duly respected on the current model:
• There are no absolutely positive (negative) facts of the matter—only relative facts—with respect to ‘claims of
inclination’.
And now the ‘no fact of the matter’ has good sense: Ordinary Viewers accept that there's a fact of the matter as to
whether Vegemite is delicious; it's just that, according to the Ordinary View (as here modelled), there's no absolutely
positive (negative) fact of the matter.

7 Closing Remarks
Wright set the task of modelling the Ordinary View. I have suggested that if we want to model the view via a
non-parametric approach, then a paraconsistent account is probably more natural than Wright's intuitionistic account.
On the other hand, if we want to model the view on a parametric account, there seem to be various options, one of
which, as sketched, takes truth to be relative correspondence—where the relativity shows up only with respect to
‘disputes of inclination’. Inasmuch as Ordinary Viewers tend towards correspondence ‘intuitions’, the relative
correspondence version is a better overall model than either Wright's ‘true relativism’ or the crude (taste-function)
13
relativism that I sketched. But, for space-considerations, I leave the matter there.
13 As for the truth of such matters—matters of taste or the like—I'm still inclined towards a simple taste-function relativism that puts an index on terms

like ‘delicious’. But that—the truth of the matter—is not the issue on the table, despite what some taste-functions might prefer.

References

Anderson, Alan Ross and Belnap, Nuel D. 1975. Entailment: The Logic of Relevance and Necessity, volume 1. Princeton
University Press, Princeton.

Barwise, Jon and Perry, John. 1983. Situations and Attitudes. MIT Press, Bradford Books, Cambridge, MA.

Batens, Diderik. 2000. ‘A Survey of Inconsistency-Adaptive Logics’. In D. Batens, C. Mortensen, G. Priest, and J.-P.
Van Bendegem (eds), Frontiers of Paraconsistency, pages 49–73. Research Studies Press, King's College Publications,
Baldock.

Beall, JC. 2000. ‘On Truthmakers for Negative Truths’. Australasian Journal of Philosophy, 78: 264–68.

Beall, JC and Ripley, David. 2004. ‘Analetheism and dialetheism’. Analysis, 64(1): 30–35.

Brown, Bryson. 2002. ‘On Paraconsistency’. In Dale Jacquette (ed.), A Companion to Philosophical Logic, 628–50.
Blackwell, Oxford.

Da Costa, N. C. A. and Alves, E. H. 1977. ‘A Semantical Analysis of the Calculi Cn’. Notre Dame Journal of Formal
Logic, 18: 621–30.

end p.73

Priest, Graham. 1979. ‘The Logic of Paradox’. Journal of Philosophical Logic, 8: 219–41.
 p p p p p y

Rescher, Nicholas and Robert Brandom. 1979. The Logic of Inconsistency. Rowman and Littlefield, Totowa, NJ.

van Fraassen, Bas C. 1969. ‘Facts and Tautological Entailments’. Journal of Philosophy, 66: 477–87. Reprinted in
Entailment Volume 1 (References: 1).

end p.74
 p p p p p y

end p.76

4 Realism: What's Left?

1. In Search of a Problem
What is realism? Why is realism a philosophical issue? And why is it still an issue today? Or more pointedly, is there a
“realism” issue still worth arguing about? The last question is the one that interests me; but to explain why, I need to
say something about the first two.

Some philosophers present realism as a doctrine concerning the existence of certain kinds of things. A realist is
someone who think that things of the kind in question exist “objectively” or “independently of the mental.” By itself,
however, this characterization does not put us in sight of anything worth calling a philosophical issue. On this
conception of realism, there will be lots of categories of things that one might or might not be a “realist” about. I'm
not a realist about unicorns, the heavenly spheres, or UFOs. Unicorns are mythical; the heavenly spheres are posits of
an exploded theory; and UFOs are illusions (or delusions). No such things exist “independently of the mental”: they
exist in fiction, history books, or people's imaginations. That is, they don't exist. But I doubt that my “anti-realist”
convictions strike anyone as philosophically interesting.

A distinguishing feature of philosophical realism and anti-realism is an odd kind of hyper-generality. As a philosopher,
I may be a realist (or not) about things belonging to certain very broad categories. So for example, I might be a
realist with respect to objects in the external world. Or I might be some kind of non-realist: an idealist perhaps.
Categories like “things in the external world” are rarely (if ever) encountered outside philosophical discussions.

In this spirit, Michael Devitt suggests that we distinguish two forms of realism: common-sense and scientific. Thus:

Common-sense RealismE. Tokens of most current observable common-sense and scientific physical types
objectively exist independently of the mental.
Scientific RealismE. Tokens of most current unobservable scientific physical types objectively exist
1
independently of the mental.
1
Michael Devitt, Truth and Realism, 2nd edn (Princeton: Princeton University Press, 1997). Quotation, p. 24. The subscripts are mine.

end p.77

I could quibble. “Objectively” strikes me as redundant: just another word for “independently of the mental.” As for
2
“physical,” Devitt inserts it to exclude “some of the more bizarre posits of folk theory.”
2 Devitt, Truth and Realism, p. 24.

However, “most” already takes care of them; and there is no common-sense interdict against bizarre physical posits.
That said, however, Devitt's distinction is a good one and I shall work with it. I will even leave “physical” in the
formulation to indicate that I will be most concerned with realism concerning things in the world around us, and not,
for example, with psychological states and process, which obviously do not exist “independently of the mental.”

Some philosophers prefer to state realism in terms of truth. Thus:

RealismT. Most current common-sense and scientific physical claims are true (or approximately true).

We can, of course, divide this doctrine into common-sense and scientific components, as above. The qualification “or
approximately true” applies principally to the latter. No one thinks that our current scientific theories are the last
word, even if they are on right lines.

I don't see much difference between the formulation of realism in terms of existence and that in terms of truth. Things
are identified by their properties. So in claiming that certain things exist, we commit ourselves to the truth of lots of
claims about them. Conversely, in taking claims, including singular claims, about certain things to be
(straightforwardly) true, we imply that those things exist. But from here on, the formulation in terms of truth is the
one I shall have in mind when discussing common-sense or scientific realism (dropping the subscripts). I take this tack
because questions about truth are central to recent discussions of realism, so we might as well have questions about
truth plainly on the table.

Many philosophers who think of themselves as realists would say that, unsupplemented, my truth-centered
formulation of realism is inadequate. To be a “real realist,” in Philip Kitcher's phrase, it is not enough to hold that lots
of common-sense and scientific claims are straightforwardly true: we have to add that “true” is to be understood as
3
“robust correspondence truth.”
3
Philip Kitcher, ‘Real Realism; the Galilean Strategy,’ Philosophical Review, 110/2 (Apr l 2001): 151–97. Some philosophers of this persuasion may hold

that this requirement is not really an addition: for example, if they think that truth as correspondence is implicitly required by genuine “independence from

the mental.” Kitcher himself thinks something of the sort.

So we get:

RealismC. Most current common-sense and scientific physical claims are true (or approximately true) in a
robust, correspondence sense of “true.”
 p p p p p y

I don't much care for talk of “real realism.” This way of talking suggests—and is of course meant to suggest—that
anything less (or other) than this inflated form

end p.78

of realism is somehow ersatz. Since this is just what I shall deny, I will tend to keep the phrase “real realism” in scare
quotes.

I will have a good deal to say about Kitcher's case for “real realism” later in the paper. But for now, the idea of “real
realism” helps with one of my introductory questions. For since the correspondence theory of truth is controversial,
bringing it into a discussion of realism gets us to a philosophical issue. However, we have not yet got to what has
generally been taken to be the issue. To do so, we need to go beyond the metaphysical or semantic doctrines
enunciated so far. We need to bring epistemology into the picture, by going back to the idea that the world exists
independently of the mental.

In one way, it is trivial to claim that some things in the world exist independently of the mental. There is no such
thing as telekinesis, or creation ex nihilo by thought alone. But no contemporary anti-realist supposes we need think
otherwise. It is equally trivial to point out that many objects in the world do not exist independently of the mental.
Many things are human creations; and for good or ill, even the natural world has been reshaped by human activity. As
for the objects of scientific study, these too are often artifacts: there were no stable electric currents before the
invention of the Leyden Jar. The world is full of things that would not have existed but for thoughtful human activity.
But no contemporary realist would disagree. This is because the independence from the mental involved in debates
over realism is not causal or empirical but “transcendental” or “logical.” Objects in the world are not logical
constructions out of sense-data; to talk of the truth about the world is not to talk about what we happen to be able to
verify, even in ideal circumstances; and so on. Hence the well-known members of the anti-realist family: (subjective)
4
idealism, verificationism, “epistemic” conceptions of truth, and all the rest.
4 Wh le it is important to separate the causal and transcendental conceptions of independence, perhaps they are not a together unconnected. Realists tend

to sense an air of spookiness in anti-realist doctrines. The world existed long before we came on the scene (in geological time, an eyeblink ago), and will

be here long after we have gone. It would have been here even if neither we nor any other intelligent life had evolved. It seems strange to say that, in

talking about such things as the world before we showed up, we are really talking about the experiences we would have had if, per impossible, we had

been around after all. Hylas said as much to Philonous, and no realist has ever been satisfied w th Philonous's reply. Of course, Berkeley really did think that

the world was a mental production, though God's rather than ours, going so far as to deny that ideas were produced by the brain. Modern versions of

Berkeley's view, such as phenomenalism, are supposedly less flamboyantly metaphysical. However, common-sense realists are dubious about the gains of

expressing such views in the formal mode.

Anti-realists are anti-skeptics, connecting truth with evidence, or truth-conditions with assertibility-conditions, so as
to secure the possibility of knowledge. Realists are anti-skeptics too, holding that lots of common-sense and scientific
claims are not just true (perhaps in a robust correspondence-to-reality way) but known to be true. But how can this
be—or how can we assure ourselves that it is so—if the world and our beliefs about it are radically independent of

end p.79

one another? This is the traditional problem for realists: to sketch a conception of mind and world that allows for the
world's knowability, without compromising its independence.

This way of putting the problem connects issues of realism with issues of skepticism. But for this very reason, it may
not do such issues any favors. The fact is, many philosophers today are impatient with skepticism. They see skeptical
arguments, not as bringing to light problems inherent in the human condition, but trading on contentious and
5
dispensable philosophical commitments.
5 For an extended defense of this diagnostic approach to skepticism, see my Unnatural Doubts, revised edn (Princeton: Princeton University Press, 1998).

Is this the right attitude towards “realism” too? Is one enough of a “realist” if one avoids “anti-realism” of one sort or
another? This is what I intended when I asked why realism is still an issue today. Not to make a mystery of things, I
think that the answer to my question is “Yes.” Avoiding classical forms of “anti-realism” gives us all the realism we
need. What remains of the “realism” issue is the question of whether we ought to be thoroughgoing naturalists.
Mixing up this question with traditional debates over realism is a bad idea.

My approach to this tangle of issues will involve examining it against the background of an important division in
contemporary anglophone philosophy. This division is between philosophers that I shall call “neo-Pragmatists” and
others that I shall call “neo-Cartesians.” I think that just about all philosophers interested in realism fall into one
camp or the other.

The problem of the relation between “mind and world” is perhaps the central problem of modern philosophy, and
certainly the matrix within which questions about “realism” emerge most clearly. Descartes decisively shaped this
problem by making a strikingly original combination of moves:
• He gave philosophy an epistemological turn, introducing methodological skepticism.
• He promoted the representative theory of mind, at the expense of Aristotelian hylomorphism. One distinctive
feature of the mental was thus intentionality or “aboutness”.
• He attempted to secure the foundations of knowledge by making incorrigibility another distinctive feature of the
mental. Thoughts and sensations were known immediately, by being present to the mind, and so potentially
 p p p p p y

knowable with certainty and completeness. The material world was not immediately knowable. Knowledge of the
world required accurate representation. This is the epistemic face of Descartes's metaphysical dualism.
• He took intentionality—representation—for granted. Intentionality was an intrinsic and irrreducible characteristic of
ideas. In particular, none of the characteristics of ideas—their “aboutness”, the logical mode of combination into
judgments, or the inferential connections between judgments—could

end p 80

be understood in causal-mechanical terms. This is the metaphysical/semantic face of Cartesian dualism.

Neo-Pragmatism and neo-Cartesianism can be understood in terms of what each rejects and preserves from the legacy
of Descartes. In the broadest terms, neo-Pragmatists keep the epistemological orientation, while rejecting the
representative theory of mind. Neo-Cartesians do the opposite. But while this fundamental difference produces
strikingly distinct constellations of philosophical commitments, there is important common ground. Neither school of
thought has much time for either skepticism or dualism. This is because neither is inclined to take intentionality for
granted.

Our question thus becomes whether either outlook preserves enough of the Cartesian matrix to keep issues about
realism alive. My suspicion is that neither does. Given a pragmatist perspective on knowledge, truth and meaning, it is
not clear what “realism” issues are left to argue about. Indeed, there may be none. And while at least some
neo-Cartesians are inclined to criticize neo-Pragmatists for turning away from such issues, they are not well placed to
make their complaint stick.

2. Neo-Pragmatism
Neo-Pragmatism can be understood in terms of a methodological orientation and two general theoretical commitments
that flow naturally from it.

(i) Epistemological orientation. As already suggested, neo-Pragmatism's methodological orientation is fundamentally

epistemological rather than metaphysical. This claim needs explanation, since neo-Pragmatists often show scant
interest in standard-issue epistemological questions. For example, Sellars clearly thinks—and Brandom and McDowell
both follow him in this—that questions in the philosophy of mind and language cut deeper than the questions that
occupy epistemologists. Some neo-Pragmatists—Rorty comes to mind—are even associated with talk of the death of
epistemology. But while my characterization of neo-Pragmatists as epistemologically oriented might seem to be in
tension with this apparent disdain for standard epistemological issues, the tension dissolves when we realize that
“epistemologically oriented” need not mean “skeptically oriented.”

When Descartes gave philosophy an epistemological turn, he did so in a special way: he introduced methodological
skepticism. With respect to this move, neo-Pragmatists are strongly anti-Cartesian. Like their classical forebears, they
are dubious about making skeptical problems philosophy's first order of business. They tend to see concern with
skepticism as either the flip-side of a hankering after foundations for knowledge, or as a regrettable obsession with
problems which, while once possessed of broad philosophical significance, have long

end p 81

since degenerated into scholastic puzzles. Skeptical arguments may continue to intrigue us, by seeming intuitively
plausible. But neo-Pragmatists think that their air of intuitiveness is to be explained away. To the extent that we take
it seriously, skepticism is to be approached diagnostically rather than constructively. However, despite this skepticism
about skepticism, neo-Pragmatists remain epistemologically oriented. In approaching the question of “mind and
world,” they do not start with semantic notions, such as reference or truth, hoping to get a grip on knowledge and
belief. Still less do they start with a notion of truth already linked to metaphysical ideas like “reality” or “the world-
in-itself”. Rather, they start with thoroughly epistemological notions, such as Brandom's “game of giving and asking
for reasons,” offering to illuminate semantic content in terms of them. In general, they approach knowledge, belief,
and meaning by way of locating claims to/attributions of knowledge and belief in human practices of inquiry,
argument, communication, interpretation, and deliberation. This methodological orientation presupposes a
non-skeptical starting point: in trying to make sense of ourselves, we start where we are, making use of whatever
knowledge we have managed to accumulate. But while resolutely anti-skeptical, this methodological orientation is
epistemological (or practical-epistemological) rather than metaphysical.

(ii) Deflationism about truth. The original pragmatists thought that their epistemological orientation mandated an
epistemological account of truth: “truth in action,” the true as whatever was “good in the way of belief” or “assertible
at the end of inquiry.” According to neo-Pragmatists, this was a mistake. Epistemic accounts of truth are both
unnecessarily revisionary of ordinary truth-talk and fraught with metaphysical temptations of their own. The
philosophical aporiai connected with truth are better avoided by approaching the concept of truth in a deflationary
spirit. A typical deflationist view of truth is that the meaning of “true” is fixed by the (non-paradoxical) instances of
some equivalence schema. For Quine, who treats “true” as a predicate of sentences, the schema is:

(TS) “p” is true if and only if p;

yielding as instances:
 p p p p p y

“Grass is green” is true if and only if grass is green; “E=mc2 ” is true if and only if E=mc2 . . . and so on.

For Horwich, who thinks that truth is more properly predicated of propositions, the schema is:

(TP) p is true if and only if p. (“ p ” stands for “the proposition that p.”)

end p 82

But both philosophers are deflationists in maintaining that an underived inclination to accept the relevant
biconditionals exhausts our grasp of the concept of truth. The concept is useful because we often need to endorse or
reject sentences (or propositions) that we cannot severally assert: we may not know exactly what they are, or there
may be too many of them. As a device for semantic ascent, truth-talk lets us do just this. We can say, “Whatever the
Dean told you about the budget, you can be sure that it is true” or “Every sentence (proposition) of the form ‘p v p’
is true.” “True” thus adds significantly to our expressive resources. But that is all it does. The function of truth-talk is
exclusively expressive.

Commitment to deflationism constrains how neo-Pragmatists approach the whole range of semantic concepts. It would
not do to explain meaning or reference in ways that allowed for the construction of a richer concept of truth than they
are willing to countenance. This brings us to the third aspect of the neo-Pragmatist outlook.

(iii) Inferentialism about meaning. Reflecting their practical-epistemological orientation, neo-Pragmatists take an
“interpretavist” approach to meaning and belief, treating meaning as whatever emerges from our practices of
interpretation and rationalization. The picture of belief and meaning that emerges from this approach takes the form
6
that Robert Brandom calls “broad inferentialism.”
6
Brandom develops this approach to meaning in Making It Explicit (Cambridge MA: Harvard University Press, 1994). For “broad inferentialism,” see p.

130ff.

On this view, beliefs derive their content from the ways in which they relate to other beliefs—licensing them and
being licensed in their turn—but also from being observationally licensed, and by licensing actions. The inclusion of
observations and actions among the “conditions and consequences” of beliefs is what makes inferentialism “broad.”
Here neo-Pragmatists remain close to their forerunners. Pragmatists old and new see beliefs as mediating between
perception and action. This idea accords well with the constraints imposed by commitment to deflationism about truth.

It is important to be clear about what the neo-Pragmatist outlook does and does not imply.

First, it is sometimes suggested that, since no two people have identical belief systems, inferentialists must hold that
7
no two people ever hold the same belief as anyone else.
7
Jerry Fodor and Ernest Lepore, Holism: A Shopper's Guide (Oxford: Blackwe l, 1992), p.17ff.

This objection assumes that inferentialists take meanings more seriously than they do (or should). While willing to
talk about what makes for meaningfulness (intentionality-in-general), neo-Pragmatists are generally not well-disposed
to meanings. This is because, for inferentialists, “thinking the same thought” is a matter of sharing contextually
relevant inferential and practical commitments. Outside of particular interpretative contexts, there need be no

end p 83

principled way of separating commitments that fix meaning from those reflecting collateral information.

Second, while inferentialism entails a limited semantic holism—we need lots of beliefs to have any beliefs at all—we
should not exaggerate its holistic aspect. While some inferentialists speak of our “web of belief” or “total view,” the
idea of a single, tightly integrated system is not an essential component of inferentialism. Inferentialism about content
is compatible with our belief system's displaying considerable modularity: that is, with its consisting of a number of
largely independent sub-systems. Inferentialism does not encourage monism, the idea that everything we do and
think should ideally be encompassed by a single, neat conceptual package.

Third, while neo-Pragmatists sometimes present themselves as offering an “anti-representationalist” conception of

mind and language, inferentialism does not imply hostility to the idea of representation as such. Inferentialists need
not eschew representational idioms, or deny that thought and language have a representational dimension. Rather,
their “anti-representationalism” is a matter of the sorts of notions they take as theoretically fundamental. Some
8
neo-Pragmatists muddy the waters here, seeming on occasion to repudiate the very idea of representation.
8
Thus Rorty writes approvingly of “Davidson's central contribution to the philosophy of mind and language: his insistence that the idea of ‘accurate

representation of reality’ is . . . dispensable.” Richard Rorty, ‘Robert Brandom on Social Practices and Representations,’ in Rorty, Truth and Progress:

Philosophical Papers (Cambridge: Cambridge University Press, 1998), pp. 122–37. Quotation p. 128. In the same vein, we find Brandom claiming that

“reference is not a relation,” so presumably not a word-world relation. See Robert Brandom, Making It Explicit, p. 325ff. There is some irony here, since

Brandom has done as much as anyone to make clear that “anti-representationalism” is entirely concerned with the order of explanation.

No such repudiation is required, though a degree of deflation may be in order. (Compare the neo-Pragmatist attitude
to truth-as-correspondence.)

Neo-Pragmatism is radically anti-Cartesian. Not only is it anti-skeptical, substantively as well as methodologically, it is

anti-dualist. Neo-Pragmatism replaces the Cartesian picture of “mind” cut off from and confronting “world” with one in
which our mindedness requires our being always already involved with things around us. Obviously, such an outlook
 p p p p p y

encourages us to wonder whether much remains of debates over realism.

3. Neo-Cartesianism
Neo-Cartesians are at one with neo-Pragmatists in turning away from methodological skepticism and the quest for
foundations for knowledge. Like neo-Pragmatists, they want to understand the relation between mind and world
starting from where we are now, making use of all relevant scientific knowledge.

end p 84

They are not trying—or should not be trying—to ensure mind's contact with the world, viewing mind and world “from
9
sideways on.”
9
A phrase used by John McDowell in Mind and World (Cambridge MA: Harvard University Press, 1994). See p. 34ff.

Again like neo-Pragmatists, neo-Cartesians repudiate both Descartes's foundational epistemology and his dualist
metaphysics. However, where the neo-Pragmatist rejection of Cartesian dualism is bound up with
anti-representationalism, neo-Cartesians want to preserve, or rehabilitate, the Representative Theory of Mind, which
they see as the enduring element in Descartes's legacy, by naturalizing it. Typically, neo-Cartesians hope that some
fundamental semantic relation, such as reference, will prove to be analyzable naturalistically, perhaps in causal-
10
physical terms.
10
Devitt and Kitcher both entertain such hopes.

As for epistemology, they hope to show that knowledge too is “a natural phenomenon,” by explaining it as belief that
11
is truth-reliable.
11
For an extended defense of the idea that knowledge is a natural phenomenon, see Hilary Kornblith, Knowledge and Its Place in Nature (Oxford: Oxford

University Press, 2002). Kornblith has some sharp criticisms of various neo-Pragmatists, myself included. I reply to Kornblith in ‘Is Knowledge a Natural

Phenomenon?’, in The Externalist Challenge, Richard Schantz (ed.) (Amsterdam: De Gruyter, forthcoming).

A naturalistic account of reference promises more than a way of draining Cartesian spookiness from the
Representative Theory of Mind: it promises a “robust” notion of correspondence truth, which, pace deflationists, will
be available for explanatory purposes. Naturally, these ambitions lead neo-Cartesians to favor “real realism.”

Finally, neo-Cartesianism reflects a definite metaphysical orientation. This claim may be more controversial.
Neo-Cartesians tend to think of themselves as hard-headed and scientifically minded, and so in a way they are. But
being anti-dualist does not mean being anti-metaphysical. Rather, neo-Cartesians tend to want to replace dualism
with a physicalistic monism, or more liberally with “scientific naturalism” (if naturalization doesn't demand reduction
12
to the vocabulary of physics).
12
Kitcher notes (op. cit., p. 194, n. 66) that he does not “see a naturalistic account of reference as necessarily involving reduction to a physicalist

vocabulary. Naturalists who are anti-reductionists with respect to some sciences (for example, parts of biology) should allow that the basis for a reduction

of the notion may outstrip the resources of physics.” But presumably not the resources of the natural sciences. Anyway, I use “scientific naturalism” to

acknowledge that not all neo-Cartesians commit themselves to strict physicalism.

Commitment to scientific naturalism amounts to more than being pro-science. It is the view that the facts uncovered
by the natural sciences constitute all the facts there are. This is why epistemology, philosophy of mind, and the theory
of meaning need to be naturalized. This methodological orientation embodies a definite metaphysical commitment,
one that is not mandated by science itself. Neo-Pragmatists can be pro-science too. But they do not think that science
is everything. They are not convinced that ways of talking, or forms of

end p 85

explanation, have to be naturalized to be respectable. Nor are they convinced that “non-natural-scientific” equals
13
“mysterious.” Naturally, this attitude strikes neo-Cartesians as an evasion of “serious metaphysics.”
13
The phrase “serious metaphysics” comes from Frank Jackson, From Metaphysics to Ethics (Oxford: Oxford University Press, 1998), ch. 1.

On the other hand, as scientific naturalists, neo-Cartesians are no more interested in skeptical issues than are
neo-Pragmatists. Accordingly, it is not clear why neo-Cartesians should feel that much remains of traditional,
epistemologically centered worries surrounding realism and anti-realism. Perhaps disputes over reductionism are all
that survive.

4. Realism Again
What issues concerning realism do philosophers find still worth discussing? In a recent and impressive defense of “real
realism,” Philip Kitcher distinguishes two themes. His distinction is helpful. But it also helps us see that there may be
less remaining in debates over realism than Kitcher, and philosophers who think like him, would have us believe.

Kitcher's first theme continues a long-standing debate in which realists confront empiricists (instrumentalists,
fictionalists, positivists) over the question of the proper attitude towards (or understanding of) certain scientific
theories. According to Kitcher, this dispute takes place against a background of shared views. Both sides agree that
statements about observable things (events, processes) are straightforwardly true or false; and to the extent that the
 p p p p p y

disputants agree that certain statements count as “observation statements,” they will attribute the same truth-values
to them. Furthermore, they agree, or at least do not argue about, the account of truth that is appropriate for these
statements. Where they part company is over whether their shared attitude towards observation statements is
appropriately extended to statements that contain “theoretical terms,” apparently referring to unobservables. Where
scientific realists happily countenance the extension, anti-realists resist. In various ways, they are dubious about
reference to unobservables. Accordingly, they are dubious about treating “theoretical” statements as straightforwardly
14
true or false, in the way that observation statements are agreed to be.
14
“Anti-realism” comes in several flavors. Reductionist anti-realists think that statements apparently making reference to unobservables can be translated

into statements that refer only to things we can observe. So while “theoretical” statements may be true or false, they are not about what they seem to

be about. By contrast, instrumentalist anti-rea ists doubt whether theoretical “statements” are straightforwardly true or false, preferring to see theories as

devices for making predictions: i.e. for linking observation statements with further observation statements. A somewhat different view is van Fraassen's

“constructive empiricism.” Whereas reductionism and instrumentalism depart from a “face-value” interpretation of scientific theories, van Fraassen

recommends that scientific theories be understood as “realists” understand them. However, where realists think that we are (often) entitled to believe

that some scientific theory is true, van Fraassen advocates a more modest stance: all we should ever believe of a given, currently successful theory is

that it is “empirically adequate.” See Bas van Fraassen, The Scientific Image (Oxford: Clarendon Press, 1980).

end p 86

Kitcher's second “prominent anti-realist theme” has to do with the semantics of factual statements generally.
Anti-realists of the second type may be happy to extend to theoretical statements the privilege (of being
straightforwardly true or false) that empiricists withhold. Their objection is to that privilege's presumed character.
Their contention is that “realists and empiricists alike went astray in their assumption about the truth and falsity in
the elite class, specifically in the proposal that truth could be understood in terms of correspondence to a
15
mind-independent world.”
15 ‘Real Realism,’ Philosophical Review, 110/2: 152.

. Kitcher takes this talk of truth-as-correspondence very seriously. For Kitcher, not only does truth depend on
reference, but reference is to be understood naturalistically in terms of causal relations between “language and
mind-independent objects.” A “real realist” is thus a realist on both counts: willing to treat theoretical statements as
straightforwardly true, and committed to understanding truth in terms of a robust notion of correspondence. While
Kitcher's two themes are immediately recognizable, and presented in familiar ways, from a neo-Pragmatist standpoint,
his presentation of both themes is misleading, even tendentious.

In presenting his first theme, Kitcher suggests that realists and anti-realists do not argue about observation
statements: their dispute is only over whether to extend the privileges accorded to observation statements to
statements containing “theoretical” terms. This isn't true at all. What neo-Pragmatists object to about empiricism is
precisely its treatment of observation statements, which they see as wrong-headed both epistemologically and
semantically.

In the case of the second theme, the picture of “realists” confronting “anti-realists” is acceptable only if we have
already decided that neo-Pragmatism is somehow anti-realist. The question of how to understand truth is centrally
important here. Avowed anti-realists (of the second type) reject a correspondence theory of truth in favor of an
epistemic theory. Since neo-Pragmatists, as deflationists, reject epistemic conceptions of truth, by this standard they
are not anti-realists. Of course, they reject robust correspondence theories too. But their principal objection is to
robustness. Since they think that a suitable equivalence-schema captures all there is to the idea of truth as
correspondence to fact, they do not so much reject this idea as render it theoretically anodyne. They are not realists
of Kitcher's second type because, in Kitcher's presentation of the second theme, the notion of “realism” already
incorporates the entire neo-Cartesian bag of tricks. Obviously, neo-Pragmatism is “anti-realist” by this standard. But
such definitional maneuvers get us nowhere.

Going back to Kitcher's first theme, we can be “realists” by being anti-empiricists. If we go no farther, choosing to
approach truth in a deflationary

end p 87

16
spirit, we reach a position we can call “minimal” or “deflationary” realism.
16 I don't much care for the phrase “deflationary truth.” What deflationists intend to deflate are philosophical theories of truth. “Deflationary truth”

suggests that truth is prima facie a “substantive” notion, so that the burden is on the deflationist to let the air out of the balloon. At least as far as

questions about realism go, I think that things are the other way around: why does the notion of truth need inflating?

Why isn't his realism enough? A glance at neo-Pragmatist objections to empiricism suggests that arguing for a
more-than-minimal realism is no easy task.

5. Against Empiricism
Anti-realists of Kitcher's first type are traditional empiricists. (There are non-traditional empiricists—van Fraasen, for
example—but the first theme continues “an older debate”.) Traditional empiricists are foundationalists. That is their
problem.

Like neo-Pragmatists generally, Sellars (who still sets the industry standard for neo-Pragmatist critiques of
 p p p p p y

empiricism) regards the quest for empirical foundations as a misguided way of responding to the legitimate
17
demand.
17
Wilfrid Sellars, Empiricism and the Philosophy of Mind, with a Study Guide by Robert Brandom (Cambridge, MA: Harvard University Press, 1978). See

especially section VIII.

Our beliefs about the world must be subject to observational constraint. Without such constraint, they would lack
empirical content, which is to say they would lack content. To this extent, empiricists are right to insist that
observational knowledge is epistemically distinguished. They go wrong in converting the demand for knowledge that is
epistemologically distinguished into a demand that is epistemologically basic. The quest for basic knowledge is a quest
for knowledge that is absolutely non-inferential. Such knowledge will be expressed by judgments that are intrinsically
credible. The idea of intrinsic credibility imposes constraints, epistemic and semantic, so severe as to be unmeetable.

Epistemically, judgments expressing basic knowledge cannot derive their authority from any further commitments
that we might hold. For example, they cannot derive their authority from our knowledge that (in favorable
circumstances, with familiar objects, etc.), we are reliable reporters on the passing scene. To connect the authority of
observation-reports with reliability-commitments (which are non-basic, because general) would compromise the claim
of such reports to be non-inferential. But neither can their authority derive from the mere fact of reliability, as
contemporary externalists suppose. Empiricist foundationalism is driven by the “internalist” demand that the authority
18
of a judgment be (potentially) evident to the person who makes it.
18
Internalism creates pressure to suppose that basic judgments must be not just credible (in the sense of likely to be true) but incorrigible. If the class of

basic judgments contained non-veridical members, which were in no way marked out, in trusting such judgments we would again be relying on our general

reliability. More than that, perhaps, on any occasion of judgment, it would be mere luck if our judgment were true. But knowledge is supposed to exclude

luck.

end p 88

Semantically, if basic judgments are to be the ultimate court of appeal, it must be possible to enter them,
knowledgeably or not, without taking for granted any non-basic commitments. Basic observational knowledge must
therefore be semantically autonomous: it must be possible for a person to have only such knowledge. But even this
isn't sufficient. If basic judgments are to furnish absolute terminating points for justificatory chains, they must be
semantically independent even of other basic judgments. Traditional empiricism takes seriously the idea of a first item
of knowledge. The knowledge expressed by basic judgments is epistemically and semantically encapsulated.

Empiricists try to fulfill the demand for encapsulated knowledge via an account of how basic observational judgments
get their content. First, such judgments embody a demonstrative element: so in making such a judgment, we focus
our attention on something present to consciousness. Second, the descriptive element in such a judgment involves a
term the use of which is fixed ostensively, as, for example, the meaning of “red” is supposedly learned by exposure to
examples of redness. An instance of a basic observation-judgment is thus “This is red now.” In the case of a judgment
like this, if we are paying proper attention, following the rules for the use of “this” and “red”, we cannot be mistaken.

In this account of basic judgments, we find verification by confrontation, justified by meaning-constitution by pure
ostension. Empiricism of this stamp builds in the fully fledged Cartesian idea that mental contents (here the contents
of “immediate experience”) are in some way “known” merely by being “present to the mind.” Suffice it to say that
19
neither neo-Pragmatists nor neo-Cartesians can take this seriously for a moment.
19 There are disturbing signs that such ideas may be making a comeback. See the essay by Laurence BonJour in Laurence BonJour and Ernest Sosa,

Epistemic Justification (Oxford: Blackwell, 2003). Sosa's essay offers a powerful critique of BonJour. For my take on the whole idea of basic beliefs, see

‘Doing without Basic Beliefs,’ in Matthias Steup (ed.), Epistemology: Central Debates (Oxford: Blackwell, 2005).

As an alternative to the empiricist idea of verification by confrontation, Sellars offers what Brandom calls a “two-ply”
20
account of non-inferential observational knowledge.
20 Robert Brandom, Tales of the Mighty Dead (Cambridge, MA: Harvard University Press, 2002), ch. 12.

Sellars develops this account in terms of the making of explicit observation-reports. For simplicity, I shall stick with
this presentation, ignoring how the account can be extended to non-verbal “inner episodes,” such as “seeings.”

The two strands in the account correspond to what Brandom calls the “sentient” and “sapient” aspects of observational
knowledge. According to Sellars, we have a primitive capacity for making discriminative responses to environmental
circumstances. In virtue of this capacity, mere sentience, we are able to acquire reliable reporting dispositions: for
example, the disposition to reliably produce

end p 89

tokens of “That's red” in the presence of red things. But while such reliable reporting dispositions are logically
necessary preconditions for the expression of basic non-inferential knowledge, they are not sufficient.
Unsupplemented, such reactions are mere “parroting,” devoid of content. This brings us to the second strand. Genuine
observation-reports differ from mere conditioned responses in virtue of what we do with them, which, for Sellars,
means using them in inferences. Only with inferential involvement does knowledge, “sapience,” enter the picture.
Sellars argues for involvement in two kinds of inferences: “same-level” and “trans-level.” Involvement in “same-level”
inferences excludes encapsulation. Involvement in “trans-level” inferences excludes semantic autonomy.
 p p p p p y

In same-level inferences, we treat observation-reports as evidence for further claims, some of which will be
non-observational. “That's red” implies “That's not green.” But given suitable collateral information, it can imply more
arcane conclusions: “So it's Schumacher's car.” For traditional empiricists, the latter inference would already involve
level-jumping. But when Sellars refers to trans-level inferences, he has something different in mind. Trans-level
inferences involve commitments concerning our reliability as reporters. In a same-level inference, we use an
observation-report to license a conclusion. In so doing, we treat the report as authoritative: credible. But in treating a
report as credible, we incur a commitment to our reliability as reporters, a commitment that can be codified in a
“reliability inference.” This does not mean that the authority of the report derives epistemically from such
commitments. When produced by accredited language-users, in appropriate circumstances, observation-reports are
21
default-credible, thus genuinely non-inferential.
21
This aspect of Sellars's account of non-inferential reporting raises difficult issues that deserve more extended discussion. I offer a detailed exploration of

the problems, interpretative and philosophical, in ‘Knowledge, Reasons and Causes: Sellars and Skepticism,’ in Skepticism in Context, James Conant and

Andrea Kern (eds) (London: Routledge, forthcoming).

Nevertheless, observation-reports are essentially subject to epistemic evaluation by reporters themselves. This entails
our knowing about the sorts of errors we are subject to, how observation reports can be challenged, how such
challenges can be met, when reports must be withdrawn, and so on. The presupposition of such extensive reliability-
22
knowledge precludes the semantic autonomy that observation-reports were traditionally thought to enjoy.
22
The problem with the metaphor of foundations is that it “keeps us from seeing that if there is a logical dimension in which other empirical propositions

rest on observation reports, there is another logical dimension in which the latter rest on the former” (Empiricism and the Philosophy of Mind, p. 78). In

connection with n. 18 above, our reliability is imperfect and sensitive to circumstances; and competent speakers know this. So incorrigibility goes by the

board too.

Putting these ideas together, no one could have only basic observational knowledge. But—and this is Sellars's main
point—semantic autonomy is not required by epistemic distinction. Observational knowledge can be genuinely
non-inferential without being free-standing, epistemically or semantically. If Sellars is right to insist on combining
reliable responsiveness to circumstances

end p.90

with two kinds of inferential involvement (same-level and trans-level), observation-reports do not constitute an
autonomous stratum of knowledge, and so cannot be foundational, by empiricist standards. At the same time, by
linking genuine sapience with a capacity for epistemic assessment, hence with reliability-knowledge, Sellars excludes
pure externalism, along with more traditional forms of foundationalism, thus posing a challenge to neo-Cartesian
23
hopes for naturalizing epistemology.
23
For further thoughts on this issue, see my ‘Mythology of the Given: Sosa, Sellars and the Task of Epistemology,’ Proceedings of the Aristotelian

Society, supplementary volume (2003); also in Sosa and his Critics, John Greco (ed.) (Oxford: Blackwell, 2005).

Given our prior discussion, this should come as no surprise.

Sellars does not suppose that empiricism is wholly pointless. Rather, he regards the quest for empirical foundations as
a misguided response to a legitimate demand. The legitimate demand is that our beliefs about the world be subject to
observational constraint. In some sense, the tribunal of experience is and must be the ultimate court of appeal for
justifying claims about the world, and for more than narrowly epistemic reasons: without observational constraint, our
beliefs would lack empirical content. Where the quest for foundations goes astray is in presupposing that, to be
genuinely non-inferential, observation-reports (or beliefs) must be logically independent of all non-observational
24
commitments.
24
Empiricism and the Philosophy of Mind, p. 78

Now the problem of our knowledge of the external world is indissolubly linked to a phenomenal conception of
25
evidence, itself the hallmark of traditional empiricism.
25 The phrase “phenomenal conception of evidence” is used by Timothy Williamson in Knowledge and Its Limits (Oxford: Oxford University Press, 2002).

That there is such an essential connection is the main theme of Unnatural Doubts.

Since we could be mistaken about the colour of a physical object, the “red” in “This is red now” must refer to
“phenomenal” redness, so that “is” has more the force of “appears.” Accordingly, in so far as the quest for basic
knowledge enforces a demand for basic judgments, it suggests that “observational” is to be identified as “experiential.”
By contrast, on a Sellarsian account, non-inferential observation-reports concern ordinary objects and events (and
their properties). Everyday things do not lurk behind the veil-of-perception. There is no veil. If the Sellarsian
approach to observational knowledge is on anything like the right lines, the problem of “our knowledge of the external
world” disappears completely, and with it any problems about defending common-sense realism.

Sellars's account of observation does not simply raise the level at which the foundations of the empirical are properly
set. Rather, he challenges the whole foundationalist picture. This picture is not purely structural. It is not the view,
that in any concrete context of inquiry or justification, some commitments are justifiably held, without our having had
to infer them from, prior reasons. Foundationalism is the view that empirical knowledge is arranged in “levels,”
according to a general and natural order of epistemic priority, each level being distinguished

end p.91
 p p p p p y

by some broad content-related feature of beliefs belonging to it. So in the traditional empiricist version, beliefs about
“theoretical” entities depend on beliefs about ordinary objects, which depend on experiential beliefs. At each level,
one can have knowledge, independently of what commitments, if any, one entertains at higher levels. But if
observational knowledge does not constitute a semantically autonomous stratum, independent of all
non-observational commitments, the picture of knowledge as arranged in distinct levels goes by the board.
Furthermore, on the Sellarsian account, the distinction between observational and non-observational is
methodological rather than substantive. What we can observe— in the sense of non-inferentially report on—depends
on what we can be trained to respond to in a sufficiently reliable way. So non-inferential observational knowledge is
not associated with any particular content.

These conclusions bear on issues of “scientific” realism. Traditional empiricist concerns about scientific realism simply
replicate—higher up the structure, as it were—worries about common-sense realism. However, with the idea of such a
structure in abeyance (deflationary) scientific realism becomes an option. Provided that it is subject to observational
constraint, talk about “theoretical” entities is not obviously in worse shape epistemically than talk about anything
else. Some things can be seen with the naked eye, others only with the aid of instruments; and others still are too
little to see at all. But size doesn't matter. Some things are too far away to see (ever), or too old to have been seen.
There is no need for invidious distinctions.

We are not yet done. Sellars's approach to empirical knowledge cuts deeper still, showing the rewards of not taking
intentionality for granted. Descartes says in effect: Here I am with all these beliefs about the world around me—I
wonder if any of them are true? Inferentialists say: without lots of true beliefs, you lack beliefs altogether, since you
lack conceptual capacities. This seems obvious enough with respect to, say, elementary mathematical concepts. A
person who cannot count, or do simple sums, does not have idiosyncratic views about numerical relationships: he has
not learned his numbers. Sellars shows that the same goes for concepts with reporting (observational) uses, hence for
knowledge of things around us. Reliable discriminative responses and the ability to make good inferential use of them
is a possession-condition for concepts of that type. Having lots of justified beliefs about the world around us is a
precondition for having any beliefs about it. We could put it like this: for Sellars, and for neo-Pragmatists generally,
mind is world-involving.

Finally, the two-ply approach to non-inferential reporting neither requires nor allows for a reductive naturalization of
either knowledge or belief. Mere reliability is not sufficient for knowledge. More than that, merely being keyed to
environmental circumstances is insufficient to confer conceptual content—even on a word with a reporting use. So
there is no suggestion that semantic notions could be reduced to causal-physical (or biological) notions. Nor, from the
standpoint of defending common-sense realism from empiricist assaults, is there any reason to

end p.92

seek such a reduction. The limited world-involvement contained in the two-ply account already gives us everything we
need. It even gives us the option of being scientific realists. For the purpose of arguing for “real realism,”
independently of antecedent metaphysical commitments, the bar has been set very high.

6. From Science to (Real) Realism

Kitcher is aware of the “deflationary realist” option and argues that it is unstable. Science itself—not extra-scientific
metaphysical commitments—pushes us beyond minimal realism. This argument has the right shape. That said, it fails:
in superficial ways through equivocation; but more deeply and interestingly through seriously underestimating the
resources available to deflationary realists.

Kitcher finds a version of deflationary realism in Arthur Fine's commendation of the “Natural Ontological Attitude
26
(NOA).”
26 Arthur Fine, The Shaky Game (Chicago: University of Chicago Press, 1986). See also Fine, ‘Unnatural Attitudes: Realist and Instrumentalist

Attachments to Science,’ Mind 95 (1986): 149–79.

In the spirit of neo-Pragmatism, Fine urges us to take science seriously, but in a “homely” way: that is, without
philosophical commentary, especially of the realist or anti-realist variety. However, Kitcher argues that NOA
undermines itself. If we “take the scientific picture of nature seriously,” we are quickly led to what he calls “the
27
Natural Epistemological Attitude” or NEA.
27
‘Real Realism,’ Philosophical Review, 110/2: 153.
28
NEA then leads us to “a position much more ambitious than NOA.”
28
Ibid. 155.

Here is the first step, from NOA-science to NEA:

Physics, physiology, and psychology, even in relatively humdrum and elementary forms, combine to give us a
view of the relations between human beings (and also other sentient organisms) and their environments. We
are animals that form representations of the things around us; that is, the world sometimes puts human beings
into states that bear content. Those states, in turn, guide our behaviour. In observing, or thinking about, other
people, we take it for granted that their representational states sometimes adequately and accurately
represent objects, facts, and events that we can also identify. We understand that among such objects are
 p p p p p y

things that are only detected with difficulty. In some instances, people misrepresent the panoply of things
around them. This Natural Epistemological Attitude has been commonplace at least since the early modern
29
period, and it involves no weighty metaphysics.
29 Ibid. 153–4.

29

29
Ibid. 153–4.

I don't believe a word of it. Shades of the Holy Roman Empire: NEA is neither natural nor epistemological nor an
attitude.

In calling it “natural,” Kitcher means that NEA has become part of common sense. No arcane scientific knowledge is
presupposed; just “humdrum and elementary” ideas. Now, there is a trivial reading of what Kitcher claims here:

end p.93

people have beliefs, which help them deal with the world around them; some of their beliefs come from their
observing things in their surroundings; people in the same place will recognize many of the same things; but people
also go wrong, sometimes badly. Of course, Kitcher does not phrase these commonplace remarks in commonplace
30
terms. He prefers the quasi-technical jargon of “representation.”
30 In his classic deconstruction of the sense-datum theory, Austin complains that the theory depends less on serious arguments than on a certain style of

“blinkered philosophical English.” In Austin's view, philosophers like Ayer operate with ill-defined, technical terms, such as “directly perceive,” whose

character they conceal by putting them into the mouth of “the plain man.” Austin's point applies to Kitcher's use of “representation.”

This allows him to elide the distinction between admitting ordinary “representational” idioms and adopting a
representationalist order of explanation. The latter is a choice of theoretical orientation certainly not mandated by
31
“physics, physiology and psychology.”
31 Kitcher's elision may have been encouraged by careless “anti-representationalist” talk on the part of some neo-Pragmatists.

What is epistemological about NEA? Perhaps there is a concept of “representation” that is useful for psychology. But it
is not immediately obvious that such a notion would be of epistemological significance. Epistemology concerns
knowledge and justification. Commonsensically, epistemological concepts are normative: for example, knowledge
involves belief that is “epistemically appropriate.” While we can argue about what this appropriateness consists in, and
whether it can ultimately be “naturalized,” Kitcher dismisses the normative dimension of epistemological concepts,
using rhetorical sleight-of-hand. Notice the reference to “human beings and other sentient organisms,” as if mere
sentience, and not sapience, were the key to understanding what is involved in the ways that human beings represent
their world. In the same vein, Kitcher tells us that “we are animals that form representations.” So we are. But this
does not establish that we can understand semantic content calling only on concepts developed to describe our animal
nature. We are also animals that engage in complicated financial transactions. It does not follow that economics can
be reduced to biology.

Neo-Pragmatists do not deny any of Kitcher's commonplaces. According to Sellarsian inferentialism, with its two-ply
account of observation-reporting, the world indeed “puts human beings in states that bear content,” since observing
involves causal interaction with the things observed. But this does not mean that the ways in which such states “bear
content” can be reductively explicated in “scientific” terms. Maybe they can and maybe they can't. But the supposition
32
that they can is not required by taking science seriously.
32
Pace Kitcher, the naturalistic view he implies is not even part of early modern formulations of the representative theory of mind.

The idea that they can defines a philosophical research program, not a natural attitude.

Our representations are sometimes accurate and sometimes not. Agreed: sometimes our beliefs are true and
sometimes they aren't. But science alone does not

end p.94

tell us that there is more to “accuracy and adequacy” than having lots of true beliefs, in a deflationary sense of “true.”
Science doesn't tell us that truth either can or needs to be illuminated by talk of accurate representation. To be fair,
at this stage, Kitcher doesn't exactly claim that it does, though he makes it clear later that he favors a causal theory
33
of reference.
33 ‘Real Realism,’ Philosophical Review, 110/2: 194, n. 66.

But with respect to NOA, and to deflationary realism generally, the effect of his way of talking about representation is
to create a strong suggestio falsi, making the transition to “real realism” seem that much easier.

It is perhaps not accidental that Kitcher refers, not to definite scientific results, but to “the scientific picture of
nature.” That seems right. Kitcher uses quasi-scientific jargon to insinuate a philosophical outlook—scientific
naturalism—for which he does not argue and with respect to which science itself is silent. NEA is not a natural
epistemological attitude, but a set of controversial metaphysical commitments. Science doesn't lead to NEA.
Accordingly, NOA doesn't undermine itself by encouraging us to take science seriously.

The argument so far defends deflationary realism from the charge of self-destruction. But the transition from NEA to
“real realism” is also instructive.
 p p p p p y

Once we get to NEA, it is supposed to be a short step to a form of realism much richer than NOA:

Consider a paradigm situation in which we observe . . . another person who is responding to an environment
that we can also observe. . . . The subject may form representations that are either accurate or inaccurate, for
there are entities independent of her of which they may or may not be correct. Further, we appreciate the
force of the suggestion that our subject's successes in responding to and shaping the environment would be
inexplicable unless some of her representations were accurate. . . . What difference does our presence (or
imagined presence) make? If we were absent would her representation vanish, would the relations between
those representations and the independent entities be altered, would her successes disappear, or would the
connection between success and accuracy of representations be affected? It would be presumptuous to think so.
We observe what is going on, but our presence isn't critical to the things the subject does or to the properties
of her representations. But we too could be viewed from the outside and assessed in a similar fashion. We thus
envisage a world of entities independent not just of each of us but of all of us, a world that we represent more
or less accurately, and we suppose that what we identify as our successes signal the approximate correctness of
34
some of our representation. So we arrive at real realism.
34
Ibid. 154–5.

What are we imagining when we envisage “a world of entities . . . independent of all of us”? Perhaps we think that the
world is a big, complicated place, and that maybe there are things about it that we will never get right. Maybe there
are matters about which we will remain permanently in error even if we never realize it. But who claims otherwise?
Inferentialists do not claim that “p” is equivalent to “We all believe that p.” In general, claims about the world, and

end p.95

claims about what we all believe, have distinct inferential powers. But since inferentialists are not real realists, Kitcher
must be looking for a stronger sense of “independence.”

Neo-Pragmatists see limits to how far we can get outside our own skins, even in our imaginations. Consonant with
their deflationary realism, neo-Pragmatists are inferentialists about meaning, and inferentialists cannot suppose that
all our beliefs could be false. Still less can they make sense of the suggestion that our entire system of concepts might
be woefully inadequate for representing reality. Without lots of straightforwardly true beliefs, we would have no
beliefs at all. This commitment to Davidsonian charity is visible in Sellars's two-ply account of observation-reports,
35
though the point is quite general.
35
Here I imply that Davidson's approach to meaning is compatible with a deflationary approach to truth. For a defense of this claim, see my ‘Meaning and

Deflationary Truth,’ Journal of Philosophy, 96 (1999): 945–64.

So to Kitcher's question “What difference does our imagined presence make?”, the answer is that anyone usefully
describable in intentional terms must at least be interpretable, whether or not anyone is around to do the
36
interpreting.
36 Meaning is thus at least public. Whether it is, in some stronger sense, “social” is a question that raises issues I cannot go into here.

Perhaps Kitcher's notion of the “independence” of the world is meant to imply that we can imagine that all our beliefs
are false, or our concepts inadequate. However, this thought points to a strange idea of “the world.” Arthur Fine calls
37
it “the World,” and makes attachment to the idea of the World a defining characteristic of realists.
37 ‘Unnatural Attitudes,’ p. 150.

The World is not our world of familiar things—shoes and ships and sealing-wax—but a world wholly independent of all
of common-sense and scientific beliefs and concepts. As Richard Rorty says, this idea of the World is “the notion of
38
something completely unspecified and unspecifiable—the thing in itself in fact.”
38 Richard Rorty, ‘The World Well Lost,’ Journal of Philosophy, 79 (1972): 649–65. Reprinted in Rorty, Consequences of Pragmatism (Minneapolis:

University of Minnesota Press, 1982).

All pragmatists will agree with Rorty that such a world is “well lost.”

If this is Kitcher's notion of “world,” he has certainly put some distance between him and his deflationary-minded
opponents. But if Kitcher is thinking along these lines, he shouldn't be. No neo-Cartesian should. Neo-Cartesians hope
to naturalize epistemological and semantic notions: for example, Kitcher hopes to flesh out a robust correspondence
conception of truth by developing a causal theory of reference. So consider the semantics of terms with “reporting”
uses, and go back (for the last time) to Sellars's two-ply account of observation-reporting. The first strand in this
account insists on reliable discriminative responsiveness. The second strand adds inferential engagement. Seeking a
more purely causal account, neo-Cartesians will want to modify or even eliminate the second strand. But the first
strand is not up for elimination. Neo-Cartesians are as

end p.96

committed as Sellars to the reliability of our reporting dispositions. Without such dispositions, there would be no
causal patterns to work with.

This said, certain things Kitcher says suggest that his world is indeed Fine's World after all. The World is radically
independent of anything we happen to think about it, or any way that we happen to conceive it. This picture of “mind”
 p p p p p y

confronting “world” seems to invite skeptical questions. That Kitcher is working with this picture is therefore strongly
indicated by his being moved to offer anti-skeptical reassurance, as when he tells us that “what we identify as our
successes” give us reason to suppose that our ways of thinking are not wholly misguided. Notice that he does not
claim that our successes give us reason to suppose, but only “what we think of” as our successes. If skepticism is in
the air, this is as it should be. To credit ourselves with success in dealing with the everyday world would be to leave
the World, begging the skeptic's question. If we are to raise and answer a fully general skeptical question, we cannot
take the everyday world for granted: we have to argue from features of our representations to the conclusion that
these representations more or less accurately mirror the World. We must argue from “what we think of as our
successes”—i.e. our representations of successes—and not from successes themselves, just as Kitcher does.

Such hints notwithstanding, in the end Kitcher's world is not Fine's World, or not officially. First, Kitcher repudiates
the classical Cartesian picture of the mind's “immediately” knowing only its own representations, and so seeking
reassurance that they correspond to a world beyond them. “When we perceive,” he tells us, “we are in causal contact
with physical objects, and although this contact is mediated by our having certain kinds of psychological states, we
39
don't perceive by perceiving those states (or their contents).”
39
‘Real Realism,’ Philosophical Review, 110/2: 157.

If this claim seems to beg the question against the skeptic, so much the worse for skepticism. Thus Kitcher disclaims
any interest in answering general skepticism: “It's impossible to throw away all our beliefs, start from scratch, and
justify the claim that the objects about which we form perceptual beliefs are as we represent them—we need our
40
contemporary amalgam of physics, physiology and psychology to advance any picture of perception.”
40
Ibid. 158.

But if this is so, it is not clear how we reach real realism, since Kitcher's “double extrapolation” to the idea of “an
independent world” now seems to stretch no farther than deflationary realists are willing and able to go.

What is Kitcher's “independent” world? Our familiar world, or Fine's World? I think that the answer is both and
neither. When needing to distance himself from deflationary realists, he edges towards Fine's World. But when
disclaiming interest in the skeptical-foundational view associated with this way of thinking, he rushes back in the
other direction. So the argument from NOA to real realism depends on two elisions: an elision of the distinction
between science and scientific naturalism, to get us to NEA, and an elision of that between world and

end p.97

World, to get us from NEA to real realism. Deflationary realists have no cause for concern.

I have examined Kitcher's argument in some detail because it represents a very recent and very sophisticated attempt
41
to breathe life into “serious” realism.
41
The argument I discuss here amounts to only a small part of Kitcher's subtle and comprehensive defense of “real realism.” However, since he presents

the argument as providing a general strategy for confronting a large variety of “anti-realist” strategies, it deserves special attention. I should add, however,

that I am not sure that all Kitcher's anti-anti-realist moves, several of which I applaud, really depend on his argument for real realism. I would like to detach

his useful strategies against anti-realists from his arguments for an inflationary version of realism.

But Kitcher's problems point to some quite general lessons.

Both neo-Pragmatism and neo-Cartesianism are anti-skeptical to the core. Both schools recommend starting from
where we are, sketching an understanding of “mind and world” that calls on whatever knowledge is to hand. Neither
school supposes that we could throw away all our knowledge of the world, in the hope of legitimating it in a wholesale
way. But this means that Cartesian “real realists” cannot distance themselves from neo-Pragmatist deflationary
realists by claiming to have articulated a more robust notion of an “independent” world than neo-Pragmatists can
allow. At least, they cannot do so if such a notion would have to be so robust as to rehabilitate all the skeptical
conundrums that have supposedly been set aside.

The upshot is that the only “realism” issue left is not the traditional problem of reconciling knowledge with the “mind-
independence” of the world, but simply the feasibility—and more importantly the desirability—of articulating a
thoroughly “naturalistic” conception of the world and our place in it. This issue needs to be detached completely from
traditional epistemologically centered issues. Real realism, in so far as it is advanced on epistemological grounds, is a
blind alley.

Of course, the neo-Pragmatist ideas I favor suggest that scientific naturalism is also something we could do without.
And while I cannot, on this occasion, make a case for this view, I will say a few words about the issue, for it is
important not to suppose that the neo-Pragmatist attitude is just an evasion of an obviously pressing concern.

First, since neo-Pragmatist inferentialism already treats “mind” as “world-involving,” as in Sellars's two-ply account of
observation-reporting, the issue of “naturalizing” the mind is not about whether causal interaction with things around
us needs to be incorporated into our picture of intentionality, but about whether notions like reference and truth need
to be reduced to causal-physical or some other broadly “scientific” notions. However, setting traditional
epistemological issues aside, as I have argued we should, makes the need for such a reduction harder to argue for, at
least in ways that have much chance of making headway against neo-Pragmatists.

end p.98
 p p p p p y

What we run into here is a fundamental difference in philosophical temperament. Neo-Cartesians are monists,
whereas neo-Pragmatists—though anti-Cartesian-dualist—generally aren't. Neo-Pragmatists see intentional
explanation as bound up with rationalization. When we explain a person's actions in terms of his beliefs, desires, etc.,
we represent him as engaged in something that, at least from his point of view, in his situation, and so on, it makes
sense to do. If a person were so wholly illogical that his actions could never be seen as even minimally sensible, there
would be no point in regarding him as having beliefs and desires, in fact no possibility of so doing. But “making sense”
is a normative notion. This is why meaning cannot be “naturalized”: not because mentality involves ghostly doings,
but because intentionality, in Sellars's phrase, is “fraught with ought.” Neo-Pragmatists thus see causal-nomological
explanation and intentional explanation as each sui generis, with distinct uses. Whereas the understanding provided
by natural laws is useful for predicting and controlling natural processes, taking the intentional stance is more
appropriate when dealing with other people or deliberating about one's own actions. Neither form of explanation
needs to be reduced to the other.

Are neo-Pragmatists anti-naturalist? It depends. Neo-Pragmatists see “natural” as having two antonyms:
“supernatural” (physis versus hyperphysis) and “normative” (physis versus nomos). If the question is whether mind
and meaning need to be naturalized, the latter contrast is the relevant one. Neo-Pragmatist anti-reductionism has
nothing to do with a sneaking regard for ghost-stuff. Standards of reasonableness have nothing spooky about them.
We set them and sometimes revise them. We are responsible both to and for them. This does not set us free to
believe what we like. Observational constraint ensures that the world takes a hand in shaping our thoughts.

If we are playing “Who is the naturalist here?”, we need to further distinguish kinds of naturalism: methodological and
metaphysical. Methodological naturalism is just anti-apriorism. In trying to make sense of anything, we can make use
of whatever putative knowledge is to hand, provided we are willing to revise our views with the progress of inquiry. In
this respect, neo-Pragmatists are thoroughgoing naturalists. But scientific (dogmatic) naturalism is a different kettle of
fish, a substantive metaphysical commitment that pragmatists old and new see no reason to make.

end p.99
 p p p p p y

5 Scientific Realism *
Versions of this paper have been delivered in many places, starting with a conference, “Logic and
Metaphysics,” held in Genoa in September 2001. I am indebted to these audiences for comments.
I am also indebted to Radu Dudau for helpful advice on the literature, to Peter Godfrey-Smith for a
helpful prior exchange on the topic, to the members of my graduate class on scientific realism in
Fall 2001, and to the following for comments on a draft: Jeff Bub, Radu Dudau, Frank Jackson,
Mikael Karlsson, Andre Kukla, Jarrett Leplin, David Papineau, Stathis Psillos, and Howard Sankey.
The paper was first published in Jackson and Smith 2005 .

What is scientific realism? The literature provides a bewildering variety of answers. I shall start by addressing this
question (section 1). I shall go on to discuss the most influential arguments for and against scientific realism. The
arguments for are the “success argument” and related explanationist arguments (section 2). The arguments against
are the “underdetermination argument” which starts from the claim that theories always have empirically equivalent
rivals; and the “pessimistic meta-induction” which starts from a bleak view of the accuracy of past scientific theories
(section 3). My approach is naturalistic.

1 What is Scientific Realism?

Science appears to be committed to the existence of a variety of unobservable entities—to atoms, viruses, photons,
and the like—and to these entities having certain properties. The central idea of scientific realism is that science really
is committed and is, for the most part, right in its commitments. As Hilary Putnam once put it, realism takes science at
“face value” (1978: 37). So, for the most part, those scientific entities exist and have those properties. We might call
this the “existence dimension” of realism. It is opposed by those who are skeptical that science is giving us an
accurate picture of reality.

Scientific realism is about unobservable entities. Science appears also to be committed to lots of observable
entities—to a variety of plants, molluscs, moons, and the like. Folk theory appears to be committed to observables like
stones, trees, and cats. A skepticism that extends to observables is extreme, “Cartesian,” skepticism. It yields the
issue of “realism about the external world.” This issue is both different from and prior to the issue of scientific realism.
It addresses doubts

end p.100

about the very clearest cases of knowledge about observables, doubts occasioned by skeptical hypotheses such as that
we are manipulated by an evil demon. The issue of scientific realism arises only once such doubts about the
observable world have, somehow or other, been allayed. Given the obvious truth of the following weak
underdetermination thesis:

WU: Any theory has rivals that entail the same actual given observational evidence,

allaying those doubts will involve accepting some method of nondeductive ampliative inference. Not even a theory
about observables can be simply deduced from any given body of evidence; indeed, not even the very existence of an
observable can be deduced “from experience.” If we are to put extreme skepticism behind us and gain any knowledge
about the world, we need some ampliative method of inference. Armed with that method, and confident enough about
the observable world, there is thought to be a further problem believing what science says about unobservables. So
the defense of scientific realism does not require that we refight the battle with extreme skepticism, just that we
respond to this special skepticism about unobservables. We shall see that this point has not been kept firmly enough
in mind.

The general doctrine of realism about the external world is committed not only to the existence of this world but also
to its “mind-independence”: it is not made up of “ideas” or “sense data” and does not depend for its existence and
nature on the cognitive activities and capacities of our minds. Scientific realism is committed to the unobservable
world enjoying this independence. We might call this the “independence dimension” of realism. The very influential
philosophers of science, Thomas Kuhn and Paul Feyerabend, think that scientific entities are not independent but are
somehow “constructed” by the theories we have of them. This “constructivism” has its roots in the philosophy of Kant
and is extremely influential. An important feature of constructivism, for the purposes of this paper, is that it applies in
the first instance to observables: there is no special problem about the independence of unobservables (as there is
1
thought to be about their existence).
1 But what about quantum theory? The notorious Copenhagen interpretation responds to the mysterious picture of the world suggested by quantum

theory by taking the quantum world to be observer-dependent. This would offend against the independence dimension of realism. But, of course, this

interpretation of the theory is not obligatory. Many interpretations have been proposed that do not involve observer dependence. I'm told that all of these

are somewhat weird in one way or another. Some philosophers respond to the mysteries by taking quantum theory instrumentally and hence not as an

accurate guide to rea ity; so, the existence dimension of rea ism is not embraced for the theory. But the enormous success of the theory makes this a

difficult choice (see section 2 below on the significance of success). In brief, controversy rages (see, for example, the papers in Cushing and McMullin 1989

and Cushing, Fine, and Goldstein 1996). This situation is both fascinating and worrying. What conclusions should we draw from it about scientific realism?

In my view, we should draw none until the dust begins to settle.

 p p p p p y

The struggle between constructivism and realism is

end p.101

2
appropriately conducted at the level of observables. I shall therefore not engage in it here.
2
Elsewhere (1991, 1999, 2001) I take a very dim view of constructivism.

Before attempting a “definition” of scientific realism, some further clarification is called for. First, talk of the
commitments “of science” is vague. In the context of the realism debate it means the commitments of current
scientific theories. The realist's attitude to past theories will be the concern of section 3.2. Second, the realist holds
that science is right, “for the most part.” It would be foolhardy to hold that current science is not making any mistakes
and no realist would hold this. Third, this caution does not seem to go far enough: it comes too close to a blanket
endorsement of the claims of science. Yet scientists themselves have many epistemic attitudes to their theories. These
attitudes range from outright disbelief in a few theories that are useful for predictions but known to be false, through
agnosticism about exciting speculations at the frontiers, to a strong commitment to thoroughly tested and
well-established theories. The realist is not less skeptical than the scientist: she is committed only to the claims of the
latter theories. Furthermore, realism has a critical aspect. Theories may posit unobservables that, given their
purposes, they need not posit. Realism is committed only to “essential” unobservables. In brief, realism is a cautious
and critical generalization of the commitments of well-established current theories.

More clarification would be appropriate but this will have to do. Utilizing the language of the clarification we can
define a doctrine of scientific realism as follows:

SR: Most of the essential unobservables of well-established current scientific theories exist
mind-independently.

With a commitment to the existence of a certain unobservable goes an implicit commitment to its having whatever
properties are essential to its nature as that unobservable. But, beyond that, SR is noncommittal on the properties of
the unobservables, on the scientific “facts.” Yet the scientific realist is often committed not only to the entity realism
3
of SR but to a stronger “fact” realism:
3 The scare quotes around “facts” are to indicate that the use of the term can be regarded as a mere manner of speaking, not reflecting any commitment

to the existence of what many regard as very dubious entities. Ian Hacking (1983) calls this sort of doctrine “theory realism.” I prefer to talk of “facts”

rather than theories to emphasiz e that the doctrine is about the world itself not our account of it.

SSR: Most of the essential unobservables of well-established current scientific theories exist
mind-independently and mostly have the properties attributed to them by science.

The existence dimensions of these doctrines are opposed by those who are skeptical of what science is revealing; the
independence dimension is opposed by the constructivists.

end p.102

Although not generally skeptical of scientific theories, SR and SSR do reflect some skepticism. By varying the amount
of skepticism, we could define some other doctrines; for example, instead of claiming that most of the unobservables
exist we could claim that a large proportion do or, even weaker, that some do. Clearly there is room for argument
about how strong a position should be defended against the skeptic. Related to this, but less interesting, there is room
for argument about which doctrines warrant the label “scientific realism.” But this does not prepare one for the
bewildering variety of definitions of scientific realism in the literature, many of them very different from SR and SSR.

SR and SSR are about what the world is like, they are metaphysical (or ontological). Some philosophers favor
4
epistemic definitions of scientific realism.
4 For example, Kukla (1998: 10); see also Psillos (1999: xix–xxi).

Thus, instead of claiming that most of the unobservables of science exist, one could claim that a belief that they do is
justified; or, instead of claiming SSR, one could claim that SSR is justified. This illustrates that epistemic definitions
are generally parasitic on metaphysical ones. And although the epistemic ones are clearly different from metaphysical
ones, they are not different in a way that is significant for the realism debate. For, if one believes that, say, SSR is
justified, one should believe SSR. On the other hand, if one believes SSR, one should be able to produce a
justification for it. And someone who urges SSR in the realism debate would produce (what she hopes is) a
justification because she would argue for SSR. A metaphysical doctrine of scientific realism and the epistemic one that
5
is parasitic on it stand or fall epistemically together.
5
Indeed, one can generalize: (p)((p is justified) is justified iff p is justified). (Talk of justification here should be construed broadly so that “externalist”

accounts of knowledge are not ruled out.) Jarrett Leplin (1997: 26) has defined an epistemic doctrine, “minimal epistemic realism,” that is not parasitic on

a metaphysical one. This doctrine does not claim that a belief in any of the claims of science is justified, just that such a belief could be justified. A realism

that concedes this much to the skeptic is indeed minimal (although still too strong for an antirealist like Bas van Fraassen).

It is common to propose what may seem to be semantic definitions of scientific realism, definitions using the terms
6
“refer” and “true”.
6 For example: Hesse (1967: 407); Hooker (1974: 409); Papineau (1979: 126); Ellis (1979: 28); Boyd (1984: 41–2); Leplin (1984b: 1–7); Fales

(1988: 253–4); Jennings (1989: 240); Matheson (1989).

 p p p p p y

For example, we might propose: “Most of the theoretical terms of currently well-established scientific theories refer to
mind-independent entities and the theories' statements about those entities are approximately true.” This should be
seen as simply a paraphrase of the metaphysical SSR, exploiting only the “disquotational” properties of “refer” and
“true” captured in the schemas “ ‘F’ refers iff Fs exist” and “ ‘S’ is true iff S.” Such paraphrases are often convenient
but they do not change the subject matter away from atoms, viruses, photons, and the like. They are not in any
interesting sense semantic. In particular, they do not involve commitment to a causal theory of reference or a
correspondence theory of truth, nor to any other

end p.103

theory of reference or truth. Indeed, they are compatible with a totally deflationary view of reference and truth: a
7
deflationist can be a scientific realist.
7 Horwich (1998).

So there are epistemic and apparently semantic definitions of scientific realism which do not differ in any significant
way from straightforwardly metaphysical definitions like SR and SSR. However, there are others that do differ
significantly. Most important are the ones that really have a semantic component. “Scientific realism” is often now
8
taken to refer to some combination of a metaphysical doctrine like SSR with a correspondence theory of truth.
8 See, for example, Putnam (1978: 18–20, 123–25; 1987: 15–16); Fine (1986a: 115–16, 136–7); Miller (1987); Kitcher (1993: 127–33); Brown

(1994). Arthur Fine dismisses the realism issue a together because he takes it to involve an issue about truth. Despite this dismissal, Fine urges the

mysterious “Natural Ontological Attitude” (1986a, 1986b) which often seems to be a realist doctrine like SSR! However, some passages (1986b: 163–5)

make it hard to take Fine as a realist.

The combination is strange. Skepticism about unobservables, which is indubitably at the center of the realism debate,
is simply not about the nature of truth. The issue of that nature is surely fascinating but is orthogonal to the realism
9
issue.
9
This objection also counts against definitions of realism that include the idea that truth is the aim of science (e g. van Fraassen (1980: 8)) wherever this

talk of truth is taken to commit realism to correspondence truth. Even if the talk does not have that commitment, and so is acceptable to a deflationist,

such definitions have problems. On the one hand, if the idea that truth is the aim of science is added to a doctrine like SSR, the addition is uninteresting: if

science is discovering the truth, nobody is going to propose that it is not aiming to, that truth is a happy accident. On the other hand, if the idea is not

added to a doctrine like SSR, the definition will be too weak: realism will require that science aims for truth w thout any commitment to it ever having

achieved that aim. Indeed, if science had never achieved that aim despite the efforts of the last few centuries, it would hardly be rational now to have the

aim. In the distant past, of course, the situation was different (as Howard Sankey has emphasized to me). Then the realistically inclined philosopher should

have had the aim without the commitment. But now she should have the commitment with the result that the aim goes without saying.

Of course there may be evidential connections between the two issues: there may be evidential connections between
10
any issues (Duhem–Quine). But no doctrine of truth is constitutive of metaphysical doctrines of scientific realism.
10
For more on this and other matters to do with defining realism, see my 1991, particularly chs 2–4; 1997, part I; and 1999, part I.

Here are two excuses for the intrusion of semantics into the definition of scientific realism. (1) We seem to need semantics to capture the “nonfactualist”

antirealism of classical positivistic instrumentalism (for a learned account of the history of this instrumentalism, see Psillos (1999, ch. 3)). This

instrumentalism is like the moral antirealism of “noncognitivism” in claiming that what appear to be descriptive and factual statements are not rea ly so. So

these “statements” are not really committed to what they appear to be committed to. I argue that these antirea isms are, nonetheless, at bottom

metaphysical not semantic (1997: part II). Aside from that, positivistic instrumentalism is no longer a player in the dispute over scientific realism (although

instrumentalism in general certainly is, for it simply involves doubting the theoretical claims of science without reinterpreting them). (2) Although a doctrine

like SSR need not be combined with a correspondence theory of truth it very likely cannot be plausibly combined with an epistemic theory of truth. But still

a nonepistemic theory is not constitutive of SSR (1991: 4.3).

In what follows I shall be concerned simply with the latter, using SR and SSR as my examples.

end p.104

We move on to consider the explanationist arguments for scientific realism, and the underdetermination argument
and the pessimistic meta-induction against realism.

2 Arguments for Scientific Realism

2.1 The Success Argument

The most famous argument for realism is the argument from the success of science. The argument has its origins in
the work of Grover Maxwell (1962) and J. J. C. Smart (1963) but its most influential expression is by Putnam (1978:
18–19) drawing on Richard Boyd. Scientific theories tend to be successful in that their observational predictions tend
to come out true: if a theory says that S then the world tends to be observationally as if S. Why are theories thus
successful? The best explanation, the realist claims, is that the theories’ theoretical terms typically refer—SR—and the
11
theories are approximately true—SSR: the world is observationally as if S because, approximately, S.
11
Note that although this argument is usua ly stated using “refer” and “true,” this is not essential. And such usage should be seen as exploiting only the

disquotational properties of the terms with no commitment to a robust correspondence relation between language and the world. The realist argument

should be that success is explained by the properties of unobservables, not by the properties of truth and reference. (“Truth, like Mae West's goodness,

has nothing to do with it”; Levin (1984: 124).) So the argument could be urged by a deflationist (Devitt (1991: 113–17)).
 p p p p p y

For example, why are all the observations we make just the sort we would make if there were atoms? Answer:
because there are atoms. Sometimes the realist goes further: it would be “a miracle” that theories were so successful
if they weren't approximately true. Realism does not just have the best explanation of success, it has the only good
explanation.

Larry Laudan (1981, 1984, 1996) has mounted a sustained attack on this argument. In the first prong of this attack,
Laudan offers a list of past theories—phlogiston theory is a favorite example—that were successful but are now known
not to be approximately true. The realist has a number of responses. First, the success of a theory can be challenged:
although it was thought to be successful, it was not really so (McAllister 1993). But unless the criterion of success is
put so high that not even contemporary theories will qualify, some theories on Laudan's list will surely survive.
Second, it can be argued that a theory was not, in the appropriate sense, well-established and hence not the sort that
the realist is committed to; or that entities it posited were not essential to its success (Kitcher 1993: 140–9). But
surely some theories on the list will survive this test too. Third, the realist can insist that there are many other past
theories, ones not on Laudan's list, for which the realist's explanation of success works fine (McMullin 1984).

end p.105

Still, what about the theories that survive on Laudan's list? The realist must offer some other explanation of their
success. So even if the approximate truth of most theories is the best, perhaps only, explanation of their success, it
cannot be so for all theories. But then the realist should not have needed to struggle with Laudan's list to discover this
need to modify the success argument. After all, the sensible realist does not suppose that no well-established scientific
theory has ever been very wrong in its entities and its claims about them. And some theories that have been very
wrong have surely been successful; indeed, scientists sometimes continue to use theories known to be false simply
because they are so successful. So the realist must offer some explanation of this success that does not depend on the
rightness of the theories.

Here is a suggestion that is very much in the spirit of the original success argument. The success of a theory T that is
12
very wrong is explained by the approximate truth of a replacement theory T .
12 Laudan is scornful of the realists’ appeal to the explanatory role of approximate truth. He complains that the notion is undefined. He acknowledges that

many scientifically useful notions are undefined, but thinks that approximate truth is so specially unclear that the realists’ appeal to it is “so much mumbo-

jumbo” (1981: 32). This is excessive. First, approximately true theories can have fully true parts that do the explaining. Second, the talk of approximate

truth is simply a convenience (note 11). The claim that the approximate truth of “a is F” explains an observation amounts to the claim that a's being

approximately F explains it. Science and life are replete with such explanations; for example, a's being approximately spherical explains why it rolls.

It is because the unobservables posited by T exist and have approximately the properties attributed to them that T is
successful. Indeed, we expect the very same theory that shows T to be wrong also to explain T's observational
13
success.
13
This expectation should not be construed as a requirement on a replacement theory: that no theory should replace T until it has explained T's success.

Laudan rightly objects to this requirement (1981: 44–5).

So the realist modifies the success argument: the best explanation of a theory's success is mostly that its
unobservables exist and have approximately the properties specified by the theory; otherwise, the best explanation is
that the unobservables of a replacement theory exist and have approximately the properties specified by that theory.
Furthermore, the realist may insist, the only way to explain the success of a theory is by appeal to its unobservables
or those of another theory.

The three earlier realist responses greatly reduce the challenge that Laudan's list poses for the original success
argument. Still, some theories on the list will survive these responses. The modified argument offers an explanation of
14
the success of those theories and is sufficient to support SR and SSR.
14
Insofar as we have reason to believe that the success of a current theory is to be explained by some as yet unknown future theory, that success

provides no reason to believe in the current theory's entities or approximate truth. So the modified success argument's support for SR and SSR does

depend on the three earlier responses showing that it is mostly the case that the success of a past theory can be explained by the reference of its terms

and its approximate truth. (I am indebted to Mathias Frisch for drawing my attention to this.)

end p.106

15
Now, of course, this modification that accepts past mistakes raises the specter of the pessimistic meta-induction.
15 Peter Lewis (2002) summarizes Laudan's discussion as a meta-induction (which differs from the meta-induction we shall be considering):

Many false past theories were successful.

So the success of a theory is not a reliable test for its truth.

Lewis argues persuasively that Laudan's meta-induction exemplifies “the false positives paradox.” To establish his conclusion, Laudan needs to find periods

of science where most theories were successful and yet most were false.

We shall consider that later (section 3.2). Meanwhile the modification does seem to save the success argument.

But perhaps antirealists can explain success? There have been attempts. (1) Bas van Fraassen has offered a
Darwinian explanation: “any scientific theory is born into a life of fierce competition, a jungle red in tooth and claw.
Only the successful theories survive” (1980: 39). But this explanation is not relevant because it is not explaining the
 p p p p p y

same thing as the realist's success argument. It is explaining why we humans hold successful theories. It is not
explaining why those particular theories are successful (Lipton 1991: 170–2; Devitt 1991: 116; Kitcher 1993: 155–7;
Leplin 1997: 7–9).

(2) Arthur Fine claims that antirealism can explain success as well as realism can by appealing to a theory's
instrumental reliability (1986b; Fine is not committed to this antirealist explanation). Jarrett Leplin (1987, 1997)
develops this proposal and labels it “surrealism.” The basic idea is that although the world has a “deep structure” this
structure is “not experientially accessible.” “The explanation of the success of any theory . . . is that the actual
structure of the world operates at the experiential level as if the theory represented it correctly” (1997: 26). Leplin
goes on to argue, in my view convincingly, that the surrealist explanation is not a successful alternative to the realist
16
one.
16
For more on this, see Kukla (1998: 20–4); Psillos (1999: 90–3).

In the second prong of his attack on realism, Laudan (1981: 45–6) has criticized the realist's success argument for its
dependence on inference to the best explanation, or “abduction.” Fine (1986a: 113–22) has made a similar
17
criticism.
17 Realists tend to be fond of abduction; for example, Glymour (1984), Boyd (1984), McMullin (1984), Musgrave (1985), and Leplin (1997).

In presenting this criticism they charge the realist with “question-begging.” This charge is not apt. The realist
argument could be question-begging only if it assumed abduction and the dispute with the antirealist was over
abduction. But the primary dispute, at least, is not over abduction but over a doctrine like SSR. So, presented in this
way, the criticism seems to miss its target. (One wonders if the cause of this mistake is that the many definitions of
scientific realism have left the target unclear.)

end p.107

The criticism should be simply that the realist argument relies on abduction and this is a method of inference that an
antirealist might reject. Van Fraassen (1980, 1989), for one, does reject it. Is the realist entitled to rely on abduction?
Boyd (1984: 65–75) has argued that the antirealists are not in a position to deny entitlement because scientists
regularly use abduction to draw conclusions about observables.

Boyd's argument illustrates an important, and quite general, realist strategy to defend unobservables against
18
discrimination, to defend “unobservable rights.”
18 For examples of this strategy, see Churchland (1985), Gutting (1985), Musgrave (1985), Clendinnen (1989), Devitt (1991: 147–53), and Ps llos

(1999: 186–91). Van Fraassen (1985) responds to the first three of these.

The realist starts by reminding the antirealist that the debate is not over extreme skepticism: the antirealist claims to
have knowledge of observables (section 1 above). The realist then examines the antirealist's justification for this
knowledge. Using this justification she attempts to show, positively, that the epistemology it involves also justifies
knowledge of unobservables. And, she attempts to show, negatively, that the case for skepticism about unobservables
produced by the antirealist is no better than the case for skepticism about observables, a skepticism that all parties to
the scientific realism dispute have rejected.

So the antirealist's criticism of the success argument leaves him with the task of showing that he can save his beliefs
about observables without using abduction. If he cannot manage this, the criticism fails. If he can—and van Fraassen
(1989) has made an attempt—then the realist seems to face the task of justifying abduction.

How concerned should the realist be about this? Perhaps not as much as many suppose. After all, the antirealist must
rely on some methods of ampliative inference, even if not on abduction, to overcome extreme skepticism. How are
those methods justified? The antirealist may well have little to say about this, relying on the fact that these methods
are widely and successfully used in science and ordinary life and on there being no apparent reason to abandon them.
But, of course, that seems to be true of abduction as well. If further justification for a method is required, where could
19
we find it?
19
One problem in finding it is that we cannot give a precise specification of any of these methods: as Georges Rey says, “no one yet has an adequate

theory of our knowledge of much of anything” (1998: 29).

The naturalistically inclined will have trouble with any attempt at an a priori justification. And it is hard to see how an
a priori approach could be effective either for or against abduction. What about an empirical justification within a
naturalized epistemology? Any justification must of course use some methods of inference. If it uses the very method
it seeks to justify, circularity threatens. Stathis Psillos argues that this circularity is not vicious (1999: 81–90). Be
that as it may, Neurath's famous image of rebuilding a boat while staying afloat on it suggests another procedure for
justifying a method: we hold fast to all other methods and use them to justify the method

end p.108

in contention. So, perhaps we can justify abduction using the methods of induction and deduction. Indeed, perhaps
the very success of abduction in science and everyday life—its tendency to produce conclusions that are later
20
observationally confirmed—provides the basis for such an inductive justification.
20 The suggestion is that experience, according to the empiricist “the sole legitimate source of information about the world” (van Fraassen (1985: 286)),

supports abduction. For arguments in favor of abduction see Boyd (1984), McMullin (1984), Lipton (1991), Devitt (1991: 111–13), and Leplin (1997:
 p p p p p y

116–20). Analogous problems arise, of course, over the justification of deduction; see Field (1996, 1998); Rey (1998); Devitt (1998).

Van Fraassen (2000: 261–71) seems to misunderstand the relation of naturalized epistemology to science. It goes without saying that epistemology

implies the methods of science. But van Fraassen seems to take the naturalist view to be that basic science, or special sciences like biology, medicine,

and psychology, imply the methods of science, a view that he rejects. This view misrepresents naturalism. Naturalism holds that epistemology is itself a

special science. As such it is no more simply imp ied by another science than is any other special science: it has the same sort of relative autonomy, and

yet dependence on basic science, as other special sciences. Naturalized epistemology, like any special science, applies the usual methods of science,

whatever they may be, mostly taking established science for granted, to investigate its special realm. In the case of epistemology that realm is those very

methods of science. The aim is to discover empirically how we humans learn, and should learn, about the world (Devitt (1991: 75–9)). We have no

reason to suppose that the methods that have yielded knowledge elsewhere cannot yield knowledge in epistemology.

In any case it is not obvious that the justification of abduction will be more problematic than the justification of the
methods of inference relied on by the antirealists.

2.2 The Success of Methodology Argument

Our scientific methodology is “instrumentally reliable” in that it leads to successful theories, theories that make true
observational predictions. Everyone agrees that our methodology does this. Why does it? What is the explanation?
Boyd (1973; 1984; 1985) has posed this question and offered an answer that is both realist and naturalist: the
methodology is based in a dialectical way on our theories and those theories are approximately true. He argues that
antirealists of various sorts cannot explain this methodological success satisfactorily and so his realist explanation is
the best. I think that he is probably right.

Like the earlier success argument this argument relies on abduction, but it has a different explanandum. Where the
earlier argument sought to explain the success of theories, this one seeks to explain the success of scientific
methodology in producing successful theories.

2.3 The Basic Abductive Argument

The two abductive arguments for realism that we have considered are somewhat sophisticated. A more basic
argument is strangely overlooked: by supposing that the unobservables of science exist, we can give good
explanations of the behavior and characteristics of observed entities, behavior and characteristics which

end p.109

would otherwise remain inexplicable. This basic argument differs from the success argument in the following way.
Where the success argument uses realism to explain the observational success of theories, the basic argument uses
realism to explain observed phenomena. This is not to say that observational success is unimportant to the basic
argument: the explanation of observed phenomena, like any explanation, is tested by its observational success. So
21
according to the basic argument, realism is successful; according to the popular one, it explains success.
21
Devitt (1991: 113–17). Hacking's arguments (1983) for the reality of entities manipulated in experiments and perceived under a microscope are

persuasive examples of the basic argument (a though he, strangely, does not regard them as abductions). Hacking's point about manipulation is clearly

related to Alan Musgrave's (1988) insistence that novel predictions give us the best reason for believing a theory. Ernan McMullin (1991), responding to

Fine (1991), provides some nice examples of the basic argument in geology, biology, and astrophysics. An advantage of the basic argument is that it

makes clear that, contrary to Fine's frequent claims, the use of abduction to justify realism is not at some “philosophical” level above science: “the

argument is properly carried on at one level only, the level of the scientist” (McMullin (1991: 104)).

In sum, there are some good arguments for scientific realism provided the realist is allowed abduction. Some critics
reject abduction but this rejection seems dubious. Perhaps our knowledge of observables depends on abduction. In
any case, abduction seems to be on an equal footing, at least, with other ampliative methods of inference.

3 Arguments against Scientific Realism

3.1 The Underdetermination Argument

There is an appealing and influential empiricist argument against scientific realism that starts from a doctrine of
22
empirical equivalence.
22
My discussion of this argument draws on a more detailed one in my 2002.

Let T be any theory committed to unobservables. Then:

EE: T has empirically equivalent rivals.

This is taken to imply the strong underdetermination thesis:

SU: T has rivals that are equally supported by all possible observational evidence for it.
23
So, realist doctrines like SR and SSR are unjustified.
23
The argument has no one clear source. But see Duhem (1906), Quine (1960, 1961 (“Two Dogmas”), and 1975); van Fraassen (1980), Putnam

(1983 (“Equivalence”)).

Some preliminaries. First, what exactly is it for two theories to be “empirically equivalent”? The basic idea is that they
 p p p p p y

have the same observational consequences. We shall later see the importance of looking very closely at this basic
idea.

end p.110

Second, where EE talks simply of T having equivalent rivals, the premiss of the argument is sometimes that T has
indefinitely many rivals (e.g. Kukla 1998: 58) and sometimes, that it has at least one (e.g. Psillos 1999: 164). For
convenience, I shall mostly treat EE as if it were only committed to one rival because its commitment to more does
not seem to make a significant difference to the conclusions we should draw.
24
Third, SU should not be confused with various other underdetermination theses
24 My presentation reflects the influence of Laudan's excellent discussion of this variety (1996: ch. 2).

including the weak and obviously true one, mentioned in section 1, that leads to the challenge of extreme skepticism:

WU: Any theory has rivals that entail the same actual given observational evidence.

SU is stronger than WU in two respects. First, SU concerns an ampliative relation between theories and evidence and
not merely a deductive one. Second, SU is concerned with T's relation to all possible evidence, not merely to the given
25
evidence.
25 So the premiss about empirical equivalence that is supposed to support SU in the underdetermination argument must also concern all possible

evidence. Psillos’ version of the argument fa ls on this score. It starts: “for any theory T and any body of observational evidence E, there is another theory

T such that T and T are empirically equivalent with respect to E” (1999: 164). The quantifiers need to be reordered if this is to support SU: for any

theory T, there is another theory T such that for any (possible) body of observational evidence E, T and T are empirically equivalent with respect to E.

If we are to avoid skepticism in the face of WU, we noted, some ampliative method of inference must be accepted. But
if SU is true, we face a further challenge: ampliative methods do not support T over its rivals on the given evidence
nor even on all possible evidence. So what T says about the unobservable world can make no evidential difference.
Surely, then, commitment to what the theory says is a piece of misguided metaphysics. Even with extreme skepticism
behind us, realism is threatened.

Now, consider EE. A good reason for believing EE is that there is an empiricist algorithm for constructing an
equivalent rival to T. Consider To, the theory that the observational consequences of T are true. To is obviously
empirically equivalent to T. Still, it may not count as a rival because it is consistent with T. That is easily fixed: T* is
the theory that To is true but T is not. T* is an empirically equivalent rival to T. So EE is established.

It is tempting to respond that T* is produced by trickery and is not a genuine rival to T (Laudan and Leplin 1991;
Hoefer and Rosenberg 1994). But this response seems question-begging and unconvincing, as Andre Kukla argues
(1998: 66–81). A better response is that, in counting theories generated by the empiricist algorithm as rivals, EE, as
it stands, is too weak to sustain SU. For, with extreme skepticism behind us, we are justified in choosing T over T*.

end p.111

In considering this choice, the first half of T*, To, is key. In van Fraassen's terminology, To is the claim that T is
“empirically adequate.” He has some famous remarks comparing this claim with the bolder claim that T is true: “the
empirical adequacy of an empirical theory must always be more credible than its truth” (1985: 247); “it is not an
epistemological principle that one may as well hang for a sheep as for a lamb” (p. 254). The extra boldness of T
comes, of course, from its realist commitment to certain truths about unobservables. Because van Fraassen thinks
that T takes no further empirical risk than To, he claims that this extra boldness “is but empty strutting and
posturing,” a “display of courage not under fire” (p. 255). We should prefer the weaker To.

Now if van Fraassen were right about this, no evidence could justify a move from To to the bolder T. So it could not
justify a preference for T over its rival T* (= To & not-T). SU would be established.

Here is a reason for thinking that van Fraassen is not right. If it were really the case that we were only ever justified
in adopting the weakest theory compatible with the possible evidence for T, we would have to surrender to extreme
skepticism. For, To is far from being the weakest such theory. For example, consider Te , the theory that T is
“experientially adequate.” Where To claims that the observable world is as if T, Te claims only that the observable
world appears to be as if T. Te is much weaker than To: it does not require that there be an observable world at all;
perhaps an evil demon is at work. Those, like van Fraassen, who believe theories of the observable world, are
26
displaying courage not under fire all the time.
26
I develop this argument more thoroughly in my 1991: 150–3.

This argument exemplifies the negative side of the realist strategy described earlier: arguing that the case for
skepticism about unobservables produced by the antirealist is no better than the case for skepticism about
observables.

We can apply the positive side of the strategy too. Any methods of ampliative inference that support the move from Te
to To and free us from extreme skepticism must justify the dismissal of the evil-demon hypothesis and a whole lot of
others. The methods must justify many singular hypotheses about unobserved objects and many general hypotheses
that cover such objects (“All ravens are black” and the like). Whether or not these methods alone support the further
move to T, hence support scientific realism, they will surely justify the dismissal of T's rival T*, produced by the
 p p p p p y

empiricist algorithm. And they will justify the dismissal of another empirically equivalent rival produced by Kukla's
algorithm according to which the world changes when unobserved (1993). It would be nice to know, of course, what
these methods are. But it is a strategic error for the scientific realist to attempt to say what such methods are in
responding to the antirealist. For, the antirealist believes in observables and whatever

end p.112

ampliative inferences support that belief will justify the dismissal of the likes of T*.

The antirealist might, of course, simply insist that inferences that work for observables do not work for unobservables.
27
Certainly there is no logical inconsistency in this insistence.
27 Kukla emphasizes this (1998: 25–6, 84).

28
Nevertheless, the insistence is arbitrary and unprincipled. The realist need say no more.
28
However, I think that an examination of the epistemic significance of observation helps to bring out the arbitrariness (1991: 143–7).

We conclude that EE as it stands cannot sustain SU: T is indeed justified over empirically equivalent rivals like T*. If
the underdetermination argument is to work, it needs to start from a stronger equivalence thesis, one that does not
count any theory as a rival to T that can be dismissed by whatever ampliative inferences enable us to avoid extreme
skepticism. Let us say that the rivals that can be thus dismissed are not “genuine.” T* and the output of Kukla's
algorithm are surely not genuine. Precisely how far we can go in thus dismissing rivals remains to be seen, of course,
pending an account of how to avoid extreme skepticism. And, given the realist strategy, the account that matters is
the one given by the antirealist.

With EE now restricted to genuine rivals, the next step in assessing the underdetermination argument is a careful
consideration of how to interpret EE's talk of empirical equivalence. The basic idea is that empirically equivalent
theories have the same observational consequences. What does this amount to? A natural first stab at an answer is
that the theories entail the same observations. This yields the following version of EE:

EE1: T has genuine rivals that entail the same possible observational evidence.

Whether or not EE1 is true, it is easy to see that it is inadequate to support SU. This inadequacy arises from the fact
that T is likely to entail few observations on its own and yet the conjunction of T with auxiliary hypotheses, theories of
instruments, background assumptions, and so on—briefly, its conjunction with “auxiliaries”—is likely to entail many
observations. T does not face the tribunal of experience alone (Duhem–Quine). By failing to take account of these
joint consequences, EE1 leaves many ways in which evidence could favor T over its rivals, contrary to SU. To sustain
SU and challenge realism, we need another interpretation of EE.

Consider Laudan and Leplin's influential critique of the underdetermination argument (1991). They propose the thesis,
“The Instability of Auxiliary Assumptions” according to which “auxiliary information providing premises for the
derivation of observational consequences from theory is unstable in two respects: it is defeasible and it is
29
augmentable” (p. 57).
29
See also Ellis (1985) and Devitt (1991: 117–21).

As the accepted

end p.113

auxiliaries that can be conjoined with T change, so do its consequences. So, any determination of T's empirical
consequence class “must be relativized to a particular state of science,” the state that supplies the auxiliary
hypotheses. Thus “any finding of empirical equivalence is both contextual and defeasible” (p. 58). To determine the
consequences of T we need more than logic, we need to know which auxiliaries are acceptable, an “inescapably
epistemic” matter (p. 59).

To avoid the consequences of this argument, Kukla (1993) proposed an answer to our interpretative question along
the following lines: for two theories to be empirically equivalent at time T is for them to entail the same observations
when conjoined with At, the auxiliaries that are accepted at T. This yields:

EE2: T has genuine rivals which are such that when T and any of the rivals are conjoined with At they entail
the same possible observational evidence.

Set aside for a moment whether or not EE2 is any threat at all to realism. It is clearly too weak to sustain the threat
posed by SU. Let T be an empirically equivalent rival to T according to this interpretation. So T& At and T & At entail
the same observations. This sort of equivalence is relative to At, to the auxiliaries accepted at a certain time. It
amounts to the claim that T and T cannot be discriminated observationally if conjoined only with those auxiliaries.
But this does not show that T and T could not be distinguished when conjoined with any acceptable auxiliaries at any
time. And that is what is needed, at least, to sustain the claim that T and T cannot be discriminated by any possible
evidence, as SU requires. SU demands a much stronger answer to the interpretative question: for two theories to be
empirically equivalent is for them to entail the same observations when conjoined with any (possible) acceptable
30
auxiliaries.
30
This demand arises out of a liberal and, it seems to me, intuitive view of what counts as “possible evidence.” Quine and van Fraassen have a more
 p p p p p y

resticted view which I discuss in my 2002, sec. 13.

This yields:

EE3: T has genuine rivals which are such that when T and any of the rivals are conjoined with any possible
acceptable auxiliaries they entail the same possible observational evidence.

If T and T were thus related they would be empirically equivalent not just relative to certain auxiliaries but tout
court, absolutely equivalent. Only then would they be observationally indiscriminable. So if EE is to support SU, it
31
must be interpreted as EE3.
31 Tim Williamson pointed out to me a problem with EE3 as it stands. Suppose that T1 and T2 are two allegedly equivalent rivals and that A1 is an

acceptable auxiliary relative to T1 but not T2 and A2 is an acceptable auxiliary relative to T2 but not T1. Thus A1 might be a theory of a testing instrument

from the perspective of T1 and A2 a theory of that instrument from the perspective of T2. So the acceptability of the auxiliaries is not independent of the

theories being tested. Now suppose that T1&A1 and T2&A1 have different observational consequences. That alone should not show that T1 and T2 are

not empirically equivalent. For, T1&A1 and T2&A2 might have the same observational consequences. Clearly, what needs to be assessed for empirical

equivalence are theories together with their dependent auxiliaries. And EE3 should be taken as referring to any possible independently acceptable

auxiliaries.

end p.114

32
The main point of Laudan and Leplin's critique can be put simply: we have no reason to believe EE3.
32 Note that this is not the claim that EE3 is “demonstratively false”; cf. Kukla (1998: 58).

If T and T cannot be discriminated observationally relative to, say, currently accepted auxiliaries, they may well be so
relative to some future accepted auxiliaries. Some currently accepted auxiliaries may cease to be accepted and some
new auxiliaries are likely to become accepted. This point becomes particularly persuasive, in my view (1991: 119),
when we note our capacity to invent new instruments and experiments to test theories. With a new instrument and
experiment comes new auxiliaries, including a theory of the instrument and assumptions about the experimental
situation. Given that we can thus create evidence, the set of observational consequences of any theory seems totally
open. Of course, there is no guarantee of successful discrimination by these means: a theory may really face a
genuine empirically equivalent rival. Still, we are unlikely to have sufficient reason for believing this of any particular
33
theory.
33
For some theories where we may have sufficient reason, and for some past ones where we wrongly thought we had, see Psillos (1999: 166–8) and

the works he cites.

More importantly, we have no reason at all for believing it of all theories, as EE3 requires. We will seldom, if ever,
have a basis for concluding that two genuine rivals are empirically equivalent in the absolute sense required by EE3.

This argument against EE3 does not depend on any assumption about the breadth of T. So EE3 cannot be saved by
taking it to apply to “total sciences” (Boyd 1984: 50). Should such a broad conjunction of theories seem to face an
equivalent rival at a certain time, we are unlikely to have sufficient reason for believing that experimental
developments will not enable us to discriminate the conjunction from its rival by supplying new auxiliaries. There is no
known limit to our capacity to generate acceptable auxiliaries.

I have argued that we have no reason to believe EE3. But suppose, nonetheless, that EE3 were true. Would this
34
establish SU and undermine scientific realism? It might well do so.
34 Laudan and Leplin (1991: 63–8) think not, arguing that T can be indirectly supported over its rival by evidence that confirms another theory that entails

T but not its rival; and that some consequences of T and its rival might support only T. But, as Kukla points out (1998: 84–90), this argument begs the

question: if EE3 really were true, this evidential support would seem to disappear.

If EE3 were true, realists would have to appeal to “nonempirical virtues” to choose between empirically equivalent
theories. Empirical virtue is a matter of entailing (in conjunction with accepted auxiliaries) observational truths and
not entailing observational falsehoods. The nonempirical virtues are explanatory power, simplicity, and the like. For
the reason indicated earlier in discussing abduction (section 2), I think that the realist is entitled to appeal to
explanatory virtues, at least. But if it really were the case that all theories faced genuine rivals equally compatible
with all possible evidence, the appeal to these

end p.115

35
virtues would seem epistemologically dubious.
35 I emphasize that since it has not been established that all theories do face such rivals, it might well be appropriate to appeal to explanatory virtues, or

indeed to the evidential support mentioned by Laudan and Leplin (1991: 63–8), to prefer some theory that does face such a rival.

For, in those circumstances, there could be no way to judge the empirical success of these virtues, no way to show, for
example, that theories that provide the best explanation tend to be observationally confirmed. So the defense of
realism might well depend on there being no good reason for believing EE3.

What about EE2? We have already seen that EE2 will not sustain SU. But perhaps it is otherwise threatening to
realism. So, first, we need to consider whether it is true; then, whether, if it were, it would undermine realism.

There are surely some theories that face a genuine rival that is empirically equivalent relative to the accepted
auxiliaries at a certain time. But do all theories face such rivals at that time, let alone at all times? EE2 guarantees
 p p p p p y

that all theories do at all times. But the ampliative methods, whatever they may be, that support our knowledge of
the observable world and avoid extreme skepticism will count many rivals as not genuine, so many as to make this
guarantee seem baseless. How could we know a priori that T must always face such a genuine rival?

Suppose, nonetheless, that EE2 were true. So, if T and its rivals are restricted to the accepted auxiliaries at a certain
time, T could not be justified over some rivals on the basis only of the observations that the theories and auxiliaries
entail and the ampliative methods that save us from extreme skepticism. So, without recourse to some further
ampliative methods, T would be underdetermined by the evidence that the restriction allows into play. Of course, once
new acceptable auxiliaries were discovered and the restriction changed, the further methods might well not be needed
to justify T over those old rivals. So this underdetermination would not be as serious as SU, but it would be serious
enough: at any time, we would not know what to be realist about. But then perhaps the realist would be entitled to
the further ampliative methods that would remove this underdetermination. For the reasons already indicated, and
36
given that the case for EE3 has not been made, I think that the realist might be so entitled.
36 In a reply to Kukla (1993), Leplin and Laudan (1993: 10), in effect, doubt EE2 but in any case emphasiz e that EE3 is what matters to the

underdetermination argument. Kukla disagrees, claiming, in effect, that EE2, when applied to total sciences, “brings in its train all the epistemological

problems that were ever ascribed to the doctrine of EE” (1998: 64). According to my discussion, EE2 would bring some epistemological problems if it

were true, but they are not as extreme as those that would be brought by EE3 if it were true.

In sum, we have no reason to believe EE2 or EE3 and so the underdetermination argument fails. However, if EE3
were true, it might well undermine scientific realism and if EE2 were true, it could. Once we have set aside extreme
skepticism, then, contrary to received opinion, the nonempirical virtues are not central to defending realism from the
underdetermination argument; the rejection of the equivalence thesis is. In drawing these conclusions I have mostly
construed EE2 and EE3 as if they were committed only

end p.116

to T having at least one genuine empirically equivalent rival. Their actual commitment to more rivals does not
significantly change the conclusions we should draw.

3.2 The Pessimistic Meta-Induction

The most powerful argument against scientific realism, in my view, is what Putnam (1978) calls a “meta-induction.” It
does not rest on a prejudice against abduction or exaggerated concerns about underdetermination. It rests on
plausible claims about the history of science. The basic version of the argument is aimed at an entity realism like SR:
the unobservables posited by past theories do not exist; so, probably the unobservables posited by present theories do
not exist. Another version, largely dependent on the basic one, is aimed at a “fact” realism like SSR: past scientific
theories are not approximately true; so, probably present theories are not approximately true. (This is an example of
the convenience of exploiting the disquotational property of “true” to talk about the world.) Both versions of the
37
argument rest on a claim about past theories from the perspective of our present theory.
37
So there is a “tension” in the argument: it seems to rest on a realist view of present science and yet concludes that this realist view is mistaken; see

Leplin (1997: 141–5). I suppose that we should see the meta-induction as some sort of reductio.

And the pessimistic suggestion is that, from a future perspective, we will have a similarly critical view of our present
theories. Laudan (1981, 1984, 1996) has supported these claims about the past with a list of theoretical failures.

Laudan's list is the one used to discuss the realist's success argument (section 2.1), but the purpose of the list is
different here. The purpose before was to show that past theories were successful without being true, thus
undermining the argument for realism that “realism explains success.” The list's purpose here is to show that past
theories were not approximately true and their unobservables did not exist, thus establishing the premiss of an
argument against realism, against the view that present theories are approximately true and their unobservables
exist.

Scientific realism already concedes something to the meta-induction in exhibiting some skepticism about the claims of
science. It holds that science is more or less right but not totally so. It is committed only to well-established theories
not exciting speculations. It leaves room for a theoretical posit to be dismissed as inessential to the theory. According
to the meta-induction, reflection on the track record of science shows that this skepticism has not gone nearly far
enough.

The realist can respond to the meta-induction by attacking the premiss or the inference. Concerning the premiss, the
realist can, on the one hand, resist the bleak assessment of the theories on Laudan's list, claiming that while some of
the unobservables posited by these theories do not exist, others do; or claiming that while there is a deal of falsehood
in these theories, there is a deal of truth

end p.117

38
too (Worrall 1989;
38
Worrall takes the truth to be not about the nature of entities but about structures that contain the entities. For a critical discussion of this “structural

realism,” see Psi los (1999: ch. 7).

Kitcher 1993: 140–9; Psillos 1999: chs 5–6). On the other hand, the realist can claim that the list is
 p p p p p y

unrepresentative, that other past theories do seem to be approximately true and to posit entities that do exist
(McMullin 1984).

In the light of history, some skepticism about the claims of science is clearly appropriate. The argument is over how
much, the mild skepticism of the realist, or the sweeping skepticism of the meta-induction. Settling the argument
requires close attention to the historical details. This is not, of course, something that I shall be attempting. However,
I shall make some general remarks about the attempt.

How can we tell whether Fs, posited by a past theory, exist? Given the disquotational schema, “ ‘F’ refers iff Fs exist,”
many approach this question by considering another: How do we tell whether “F” refers? This common approach would
be harmless if it exploited only the disquotational property of “refer” captured by the schema, a property acceptable to
the deflationist. For then the reference question is just a paraphrase of the existence question. However, it is usual
(and, in my view, right) to take “refer” to pick out a substantive semantic relation between “F” and the world, a
relation that needs to be explained by a theory of reference. So it is natural to take the reference question to concern
this substantive relation and to be answered by appealing to some theory of reference. But then the common
approach is far from harmless.

The first problem is that the theory of reference appealed to on this approach is usually a description theory.
According to this theory, the reference of “F” depends on the descriptions (other terms) that its containing theory
associates with it: it refers to whatever those descriptions pick out. It is likely that, from our present scientific
perspective, those descriptions do not pick anything out. So, the conclusion is drawn that “F” does not refer and hence
39
there are no Fs.
39
See Kuhn (1962) for an argument along these lines. Stephen Stich (1983) and others have argued similarly for various forms of eliminativism about

the mind. Stich has since recanted (1996: 3–90).

Yet, the arguments of Saul Kripke (1980) and others have made it likely that reference for some terms, at least, is to
be explained not by a description theory but by a theory that links a term to its referent in a more direct causal
40
way.
40
For a summary and development of these and other moves in the theory of reference see Devitt and Sterelny (1999).

So it may well be that a description theory is the wrong theory for “F”. This points to the second, deeper, problem with
the common approach: in attempting to answer the existence question by answering the reference question, the
approach has its epistemic priorities all wrong. For, we know far less about reference, particularly about when to apply
a description theory and when to apply a causal theory, than we know about what exists. In light of this, the rational
procedure is to let our

end p.118

41
view of what exists guide our theories of reference rather than let our theories of reference determine what exists.
41
My 1991 argues for this prior ty. There is, of course, a truth underlying the mistaken approach: to determine whether the posits of a theory exist we

have to know what those pos ts are and for that we have to understand the language of the theory (pp. 50–3). But understanding a language is a

practical skill that does not require theoretical knowledge about the language, else we would understand very little (pp. 270–5).

So, we should not use a theory of reference to answer our existence question. How then should we answer it?
Consider how, in general, we argue directly for the nonexistence of Fs. On the basis of the established view of Fs, we
start, implicitly if not explicitly, with an assumption about the nature of Fs: something would not be an F unless it
were G. Then we argue that nothing is G. So, there are no Fs. Very often this argument is persuasive and generally
accepted. But someone might respond by denying the assumption about nature. “Fs do not have to be G, they are just
mistakenly thought to be G. So the argument proves nothing.” How do we settle this disagreement? It may be
difficult. We can try saying more about the established view of Fs, but this may not do the trick. After all, the
responder does not deny that Fs are thought to be G, just that being G is part of the nature of being an F. And the
established view may not be clear on the nature issue. We may be left with nothing but a “clash of intuitions” over
that issue. In such a situation, we should wonder whether there is a genuine issue to settle: there may be no
determinate matter of fact about the nature issue. If there is not, then there is no determinate matter of fact about
whether the absence of G things establishes the nonexistence of Fs.

Consider two humdrum examples. Most people are antirealist about witches because they believe that nothing casts
spells, rides on a broomstick through the sky, and so on. Some people may be antirealist about God because they are
convinced by the Problem of Evil that nothing is both all powerful and all good. But these are grounds for antirealism
only if casting of spells, riding on broomsticks, and so on, and being all powerful and all good, are essential to witches
and God, respectively. There may be disagreement about that. And there is room for worry that disagreement may
not be entirely over matters of fact.

In light of this, we can expect that close attention to the historical details about past unobservables will reveal some
ontologically determinate cases but very likely some indeterminate ones too. The determinate cases will surely include
some of nonexistence; phlogiston is a good candidate. But it will surely also include some of existence; the atoms
42
posited in the nineteenth century are good candidates.
42
Also molecules and microbes; see Miller (1987).

So, we should conclude that the premiss of the meta-induction is overstated, at least. But how much is it overstated?
 p p p p p y

That depends on the “success ratio” of past theories, the ratio of the determinately existents to the determinately
nonexistents+indeterminates. Where is this ratio likely to leave scientific realism? To answer this we need to consider
the meta-induction's inference.

end p.119

I think (1991: 162–5) that there is a good reason for being dubious about the inference. Suppose that our past
theories have indeed failed rather badly to get the unobservable world right. Why would that show that our present
theories are failing similarly? It clearly would show this if we supposed that we are no better at finding out about
unobservables now than we were in the past. But why suppose that? Just the opposite seems more plausible: we are
now much better at finding out about unobservables. A naturalized epistemology would surely show that science has
for two or three centuries been getting better and better at this. Scientific progress is, to a large degree, a matter of
improving scientific methodologies often based on new technologies that provide new instruments for investigating the
world. If this is so—and it seems fairly indubitable—then we should expect an examination of the historical details to
show improvement over time in our success ratio for unobservables. If the details do show this, it will not matter to
realism that the ratio for, say, two centuries ago was poor. What will matter is that we have been improving enough
43
to now have the sort of confidence reflected by SR.
43 So this realist response does not take the failures of “immature” science to be irrelevant to the defense of rea ism, thus threatening the defense with

“vacuity” (Laudan (1981: 34)). Rather, it takes the relevance of a science's failures (and successes) to that defense to increase with the degree of that

science's maturity, a degree assessed by an empirical epistemology.

And if we have been improving, but not fast enough for SR, the realist can fall back to a more moderate commitment
to, say, a high proportion of the unobservables of currently well-established theories.

Improvements in scientific methodologies make it much harder to mount a case against realism than seems to have
been appreciated. For the appeal to historical details has to show not only that we were nearly always wrong in our
unobservable posits but that, despite methodological improvements, we have not been getting significantly righter. It
seems to me most unlikely that this case can be made.

4 Conclusions
Scientific realism is best seen as a straightforwardly metaphysical doctrine along the lines of SR or SSR. Various
explanationist arguments for scientific realism succeed provided that the realist is entitled to abduction. I have
suggested that the realist is entitled to this. The underdetermination argument against realism fails because we have
no good reason to believe an empirical equivalence thesis that would serve as its premiss. The pessimistic
meta-induction, with its attention to past theoretical failures, does pose a problem for realism. But the problem may
be manageable. For, the antirealist must argue that the historical record shows not only that past failures are
extensive but also that we have not improved our capacity to describe the unobservable world sufficiently to justify
confidence that

end p.120

the accounts given by our current well-established theories are to a large extent right. This is a hard case to make.

References

Boyd, Richard N. 1973. ‘Realism, Underdetermination and a Causal Theory of Evidence.’ Nous, 7: 1–12.

Boyd, Richard N. 1984. ‘The Current Status of Scientific Realism.’ In Leplin, 1984a: 41–82.

Boyd, Richard N. 1985. Lex Orandi est Lex Credendi. In Churchland and Hooker, 1985: 3–34.

Brown, James Robert. 1994. Smoke and Mirrors: How Science Reflects Reality. New York: Routledge.

Churchland, Paul M. 1985. ‘The Ontological Status of Observables: In Praise of the Superempirical Virtues.’ In
Churchland and Hooker, 1985: 35–47.

Churchland, Paul M., and Hooker, Clifford A. (eds). 1985. Images of Science: Essays on Realism and Empiricism, with a
Reply from Bas C. van Fraassen. Chicago: University of Chicago Press.

Clendinnen, F. J. 1989. ‘Realism and the Underdetermination of Theory.’ Synthese, 81: 63–90.

Cushing, James T., and McMullin, Ernan (eds). 1989. Philosophical Consequences of Quantum Theory. Notre Dame:
University of Notre Dame Press.

Cushing, James T., Fine, A., and Goldstein, S. (eds). 1996. Bohmian Mechanics and Quantum Theory: An Appraisal,
Boston Studies in the Philosophy of Science, vol. 184. Dordrecht: Kluwer.

Devitt, Michael. 1991. Realism and Truth. 2nd edn (1st edn, 1984). Oxford: Blackwell.

Devitt, Michael. 1997. ‘Afterword.’ In a reprint of Devitt, 1991. Princeton: Princeton University Press: 302–45.

Devitt, Michael. 1998. ‘Naturalism and the A Priori.’ Philosophical Studies, 92: 45–65.
 p p p p p y

Devitt, Michael. 1999. ‘A Naturalistic Defense of Realism.’ In Metaphysics: Contemporary Readings, ed. Steven D.
Hales. Belmont, CA: Wadsworth Publishing Company (1999): 90–103.

Devitt, Michael. 2001. ‘Incommensurability and the Priority of Metaphysics.’ In Incommensurability and Related
Matters, eds P. Hoyningen-Huene and H. Sankey. Dordrecht: Kluwer Academic Publishers: 143–57.

Devitt, Michael. 2002. ‘Underdetermination and Realism.’ In Realism and Relativism: Philosophical Issues, 12, 2002,
eds Ernest Sosa and Enrique Villanueva. Cambridge, MA: Blackwell (2002): 26–50.

Devitt, Michael, and Sterelny, Kim. 1999. Language and Reality: An Introduction to the Philosophy of Language. 2nd
edn (1st edn, 1987). Oxford: Blackwell.

Duhem, P. 1906. The Aim and Structure of Physical Theory, trans. P. Wiener. Princeton: Princeton University Press
(1954)

Ellis, Brian. 1979. Rational Belief Systems. Oxford: Blackwell.

end p.121

Ellis, Brian. 1985. ‘What Science Aims to Do.’ In Churchland and Hooker, 1985: 48–74.

Fales, Evan. 1988. ‘How to be a Metaphysical Realist.’ In Midwest Studies in Philosophy, Volume XII: Realism and
Antirealism, eds Peter A. French, Theordore E. Uehling, Jr., and Howard K. Wettstein. Minneapolis: University of
Minnesota Press: 253–74.

Field, Hartry. 1996. ‘The A Prioricity of Logic.’ Proceedings of the Aristotelian Society, 96: 1–21.

Field, Hartry. 1998. ‘Epistemological Nonfactualism and the A Prioricity of Logic.’ Philosophical Studies, 92: 1–21.

Fine, Arthur. 1986a. The Shaky Game: Einstein, Realism, and the Quantum Theory. Chicago: University of Chicago
Press.

Fine, Arthur. 1986b. ‘Unnatural Attitudes: Realist and Instrumentalist Attachments to Science.’ Mind, 95: 149–77.

Fine, Arthur. 1991. ‘Piecemeal Realism.’ Philosophical Studies, 61: 79–96.

Glymour, Clark. 1984. ‘Explanation and Realism.’ In Leplin, 1984: 173–92.

Gutting, Gary. 1985. ‘Scientific Realism versus Constructive Empiricism: A Dialogue.’ In Churchland and Hooker,
1985: 118–31.

Hacking, Ian. 1983. Representing and Intervening: Introductory Topics in the Philosophy of Natural Science.
Cambridge: Cambridge University Press.

Hesse, Mary. 1967: ‘Laws and Theories.’ In The Encyclopedia of Philosophy, vol. IV, ed. Paul Edwards. New York:
Macmillan: 404–10.

Hoefer, C., and Rosenberg, A. 1994. ‘Empirical Equivalence, Underdetermination, and Systems of the World.’
Philosophy of Science 61: 592–607.

Hooker, Clifford A. 1974. ‘Systematic Realism.’ Synthese, 51: 409–97.

Horwich, Paul. 1998. Truth. 1st edn 1990. Oxford: Clarendon Press.

Jackson, Frank, and Smith, Michael (eds). 2005. The Oxford Handbook of Contemporary Philosophy. Oxford: Oxford
University Press.

Jennings, Richard. 1989. ‘Scientific Quasi-Realism.’ Mind, 98: 223–45.

Kitcher, Philip. 1993. The Advancement of Science: Science without Legend, Objectivity without Illusions. New York:
Oxford University Press.

Kripke, Saul A. 1980. Naming and Necessity. Cambridge, MA: Harvard University Press.

Kuhn, Thomas S. 1962: The Structure of Scientific Revolutions. Chicago: Chicago University Press. 2nd edn, 1970.

Kukla, Andre. 1993. ‘Laudan, Leplin, Empirical Equivalence, and Underdetermination.’ Analysis 53: 1–7.

Kukla, Andre. 1998. Studies in Scientific Realism. New York: Oxford University Press.

Laudan, Larry. 1981. ‘A Confutation of Convergent Realism.’ Philosophy of Science, 48: 19–49. Reprinted in Leplin,
1984a.

Laudan, Larry. 1984. Science and Values. Berkeley: University of California Press.

Laudan, Larry. 1996, Beyond Positivism and Relativism: Theory, Method and Evidence. Boulder, CO: Westview Press.

Laudan, Larry, and Leplin, Jarrett. 1991. ‘Empirical Equivalence and Underdetermination.’ Journal of Philosophy, 88:
 p p p p p y

449–72. Reprinted in Laudan, 1996. (Page references are to Laudan, 1996.)

Leplin, Jarrett (ed.). 1984a. Scientific Realism. Berkeley: University of California Press.

end p.122

Leplin, Jarrett. 1984b. ‘Introduction.’ In Leplin, 1984a: 1–7.

Leplin, Jarrett. 1987. ‘Surrealism.’ Mind, 96: 519–24.

Leplin, Jarrett. 1997. A Novel Defense of Scientific Realism. New York: Oxford University Press.

Leplin, Jarrett, and Laudan, Larry. 1993. ‘Determination Underdeterred: Reply to Kukla.’ Analysis 53: 8–15.

Levin, Michael. 1984. ‘What Kind of Explanation is Truth?’ In Leplin, 1984a: 124–39.

Lewis, Peter J. 2002. ‘Why the Pessimistic Induction is a Fallacy.’ Synthese, 129: 371–80.

Lipton, Peter. 1991. Inference to the Best Explanation. London: Routledge.

McAllister, J. W. 1993. ‘Scientific Realism and Criteria for Theory-Choice.’ Erkenntnis, 38: 203–22.

McMullin, Ernan. 1984: A Case for Scientific Realism. In Leplin, 1984a: 8–40.

McMullin, Ernan. 1991. ‘Comment: Selective Anti-Realism.’ Philosophical Studies, 61: 97–108.

Matheson, Carl. 1989. ‘Is the Naturalist Really Naturally a Realist?’ Mind, 98: 247–58.

Maxwell, Grover. 1962: The Ontological Status of Theoretical Entities. In Minnesota Studies in the Philosophy of
Science, Volume III: Scientific Explanation, Space and Time, eds H. Feigl and G. Maxwell. Minneapolis: University of
Minnesota, 1962: 3–27.

Miller, Richard W. 1987. Fact and Method: Explantion, Confirmation and Reality in the Natural and Social Sciences.
Princeton: Princeton University Press.

Musgrave, Alan. 1985. ‘Realism versus Constructive Empiricism.’ In Churchland and Hooker, 1985: 197–221.

Musgrave, Alan. 1988. ‘The Ultimate Argument for Scientific Realism.’ In Relativism and Realism in Science, ed. R.
Nola. Dordrecht: Kluwer Academic Publishers: 229–52.

Papineau, David. 1979: Theory and Meaning. Oxford: Clarendon Press.

Psillos, Stathis. 1999. Scientific Realism: How Science Tracks Truth. New York: Routledge.

Putnam, Hilary. 1978. Meaning and the Moral Sciences. London: Routledge & Kegan Paul.

Putnam, Hilary. 1983. Realism and Reason: Philosophical Papers, vol. 3. Cambridge: Cambridge University Press.

Putnam, Hilary. 1987. The Many Faces of Realism. LaSalle: Open Court.

Quine, W. V. 1960. Word and Object. Cambridge, MA: MIT Press.

Quine, W. V. 1961. From a Logical Point of View. 2nd edn (1st edn, 1953). Cambridge, MA: Harvard University Press.

Quine, W. V. 1975. ‘On Empirically Equivalent Systems of the World.’ Erkenntnis, 9: 313–28.

Rey, Georges. 1998. ‘A Naturalistic A Priori.’ Philosophical Studies, 92: 25–43.

Smart, J. J. C. 1963: Philosophy and Scientific Realism. London: Routledge & Kegan Paul.

Stich, Stephen P. 1983. From Folk Psychology to Cognitive Science: The Case Against Belief. Cambridge, MA: MIT
Press.

Stich, Stephen P. 1996. Deconstructing the Mind. Oxford: Oxford University Press.

van Fraassen, Bas C. 1980. The Scientific Image. Oxford: Clarendon Press.

end p.123

van Fraassen, Bas C. 1985. ‘Empiricism in the Philosophy of Science.’ In Churchland and Hooker, 1985: 245–308.

van Fraassen, Bas C. 1989. Laws and Symmetry. Oxford: Clarendon Press.

van Fraassen, Bas C. 2000. ‘The False Hopes of Traditional Epistemology.’ Philosophy and Phenomenological Research,
40: 253–80.

Worrall, J. 1989. ‘Structural Realism: The Best of Both Worlds.’ Dialectica, 43: 99–124.
 p p p p p y

end p.124
 p p p p p y

6 Scientific Realism as an Issue in Semantics *

These comments follow in outline but not always in detail the comments that I made at the Truth
and Realism conference. As a result of my discussions with Michael Devitt during the conference, I
discovered that I had to work a lot harder to make the point I had been trying to make in section
I . As formulated in the comments presented, section III might have been taken as an
alternative epistemology for realism; now it serves to question in another way the possibility of
formulating a general thesis of scientific realism.

No doubt the question whether atoms exist is not a question about the nature of representation, but it is not very
clear to me that we can formulate some general thesis that we might call “scientific realism” that is not a doctrine
concerning the nature of representation. Further, if we do formulate it as such, against Michael Devitt's will, then I am
not sure Devitt's own conception of semantics helps sustain a realistic attitude toward the unobservables posited by
science. Finally, it is not clear to me that the reason to believe in unobservable entities is always abductive inference,
as Devitt supposes. Sometimes something more is required, such as an ability to count and track. These will be my
themes.

I. Scientific Realism and Semantics

Suppose I am J.J. Thomson in 1897 and as a result of my experiments with cathode ray tubes, I declare that electrons
exist. Does that make me a scientific realist? I believe that electrons exist, but to be a scientific realist, surely I have
to accept some more general doctrine such as we might derive from the example of my belief in electrons. Or suppose
I am Thomas Hunt Morgan in 1917 and on the basis of my breeding experiments I conclude that genes exist. Does
that make me a scientific realist? Perhaps that makes me a realist about genes, but in order to be a scientific realist,
surely I have to believe some more general proposition that my belief in genes exemplifies. So what is this more
general doctrine that we might call “scientific realism”?

Michael Devitt has argued very forcefully, in the present paper (this volume, Ch. 5) and elsewhere, that, whatever it
is, it is not a semantic issue. He writes:

Skepticism about unobservables, which is indubitably at the center of the realism debate, is simply not about
the nature of truth. (this volume, section 1)

end p.125

Devitt thinks that, while the use of semantic terminology in formulations of scientific realism should be
unobjectionable in principle, it can nonetheless be very misleading. We can make it very plain that scientific realism is
not an issue concerning semantics by just not using terms like “true” and “refers” in our formulation of the general
thesis of scientific realism at all. Thus, he defines what he calls “entity realism” as follows:

SR: Most of the essential unobservables of well-established current scientific theories exist
mind-independently. (this volume, section 1)

What he calls “fact realism,” SSR, adds to this that these unobservables “mostly have the properties attributed to
them by science.”

But what is an “unobservable of a scientific theory”? Doesn't that mean, an unobservable kind of thing posited by the
theory? And doesn't that mean, an unobservable kind of thing such that, according to the theory, things of that kind
exist? And when we cash out in direct discourse the indirect discourse of the phrase “according to,” we find that this,
in turn, means:

an unobservable kind K such that the theory implies “Instances of” “kind exist” and refers to kind K.

So if we reformulate Devitt's SR to make its meaning explicit, what we get is this:

SRS: For most (all?) of the well-established current scientific theories T and most of the terms of T such that
T implies “Instances of” “kind exist”, if refers to unobservable kind K, then instances of kind K exist
mind-independently.

But “refers” is a semantic term. So scientific realism is a thesis we formulate in semantic terms. We can even say, it is
a thesis about the nature of truth. Inasmuch as a sentence of the form “Instances of” “kind exist” is true if and
only if instances of the kind that refers to exist, SRS tells us the following about the nature of truth: If a theory T is

well-established, then its consequences of the form “Instances of” “kind exist” are probably true. Inasmuch as
the property of being well-established is an epistemic property, which attaches to a theory only insofar as its
proponents are in a certain state of mind, we could even say that Devitt's scientific realism makes truth
mind-dependent—contrary to his intention.

In the present paper and elsewhere, Devitt has stressed that scientific realism, as he understands it, is a contingent
thesis. One of his answers to the pessimistic metainduction (“Scientific Realism”, section 3.2) is that even if scientific
 p p p p p y

theories were often wrong in the past, we can accept the commitments of our current theories, because we are now
better at discovering the truth than we were in the past. In his book Realism and Truth, he calls his thesis “realism
(now)” (1997, p. 20). What he means is that his realism is not a general policy that would commit a person to
whatever entities may be posited by the theories of his or her age, but

end p.126

only a commitment to the entities posited by the theories that have currency now. He goes on to say that realism
(now) could be formulated as a long disjunction of conjunctions. The conjuncts would be existence claims such as
“Electrons exist” and “Curved space-time exists.” In that passage, he is talking about a realism that includes
common-sense realism; so he includes sentences like “Cats exist.” We cannot be confident that all of the existence
claims of contemporary science and common sense are true; but we could express our confidence that most of them
are true with a long disjunction of conjunctions, each conjunction in which comprised most of these existence claims.

So let us see whether we can avoid semantic terminology in the formulation of scientific realism by formulating it as
such a disjunction of conjunctions. Suppose, for simplicity, that contemporary science makes only three existence
claims about unobservable entities, E1 , E2 and E3. Then Devitt might claim that the correct formulation of his scientific
realism is just the following disjunctive theory:

DT: (E1 and E2 ) or (E2 and E3 ) or (E1 and E3).

Of course, Devitt is not defending DT on specifically empirical grounds. We can imagine a “scientific skeptic” who
thinks that the latest experiments with nuclear accelerators, or the latest astronomical observations, show that all of
contemporary physics, and indeed all of contemporary science, rests on a fundamental error about the nature of
energy. Against such a scientific skeptic, DT will require a kind of defense far from Devitt's purview. Nonetheless,
Devitt can rightly claim that he is doing his part in the defense of DT by defending it on philosophical grounds against
other, more philosophical challenges.

However, I am not sure that we really can take DT as the whole of Devitt's scientific realism. If there are philosophers
who argue on purely philosophical grounds that we have no reason to believe contemporary scientific theories, then
surely they are confused and their confusions need to be exposed. But I am not sure that that is ever the correct
formulation of the antirealist's thesis, namely, that we have no reason to believe contemporary scientific theories. Is it
not always something more subtle, such as that we have no reason to believe that contemporary theories are
“literally true,” as opposed to being only “empirically adequate”? If the opposition always formulates its thesis using
semantic terminology, such as “true” and “refers,” and what we have to show is that no skepticism about
contemporary science can be properly formulated in those terms, then we cannot deny that the debate is, at least in
part, a debate about semantics.

Moreover, it is not true that Devitt's defense of scientific realism is nothing more than an exposé of philosophical
confusions. Section 2 of the present paper surveys and endorses several general arguments in defense of scientific
realism. The reason Devitt focuses on in this paper is that realism explains the fact (or purported fact) that our
scientific theories yield predictions that tend to come true (section 2.1, first paragraph). (Or maybe it would be safer
to say that we have

end p.127

theories that improve our success rate from one low rate to a slightly better but still low rate.) So we may ask: what
is the realistic thesis that these arguments are arguments for? It cannot be DT. It is not DT that explains the fact that
our theories are successful in generating predictions. Part of the explanation of why the electron theory enables us to
predict what will happen in a cathode ray tube might be that electrons exist. But since “electrons exist” will be absent
from some of the disjuncts, DT does not even say that electrons exist. DT is a disjunction of conjunctions, many of
which are quite irrelevant to the explanation of our success with cathode ray tubes. So should we say that a
disjunction of theories explains a disjunction of predictions?

In any case, I think any such appeal to DT would be a bluff. None of us can actually list the sentences to be included
among the conjuncts in the conjunctions that are the disjuncts in DT. When we ask Devitt to explicitly state the thesis
he is defending, the sentence he speaks in response will not be DT. When he thinks about scientific realism in deciding
what to say about it, what he thinks to himself will not be DT. On the assumption that Devitt can actually state the
thesis he is defending, the thesis he is defending just cannot be DT!

Devitt himself is not totally opposed to the use of semantic vocabulary in formulations of scientific realism.

For example, we might propose: “Most of the theoretical terms of currently well-established scientific theories
refer to mind-independent entities and the theories' statements about those entities are approximately true.”
(this volume, section 1).

But, he goes on to say:

This should be seen as simply a paraphrase of the metaphysical SSR, exploiting only the “disquotational”
properties of ‘refer’ and ‘true’ captured in the schemas “ ‘F’ refers iff F's exist” and “ ‘S’ is true iff S.” Such
paraphrases . . . are compatible with a totally deflationary view of reference and truth . . . (this volume, section
1)
 p p p p p y

One way of taking this caveat is as saying that if, in discussing scientific realism, the terms “true” and “refer” become
objects of contention, then we can always revert to a formulation that does not employ those terms. But if that is
what Devitt means, then the claim is so far not established. He has not given us a clear formulation that does not
employ semantic vocabulary.

Deflationists about truth do not have to say—indeed, they generally deny—that the term “true” is everywhere
eliminable, that for any claim that we formulate by means of that word, we can find an equivalent claim that does not
employ that word. The thesis of deflationism is, rather, twofold. First, it claims that every sentence of the form:

[s] is true,

where “[s]” stands for the quotation-name of a sentence s, is in some sense equivalent to the sentence s. Second, it
says that this equivalence is not in turn to be

end p.128

explained in terms of a semantic theory that treats semantic properties as, in some sense, honest-to-goodness real
properties. (This is comparable to Devitt's own characterization of deflationism in his 2001, section 6.)

So perhaps we could charitably interpret Devitt as holding, not that terms like “true” and “refers” are eliminable from
the formulation of scientific realism, but only that if, in the course of arguing over scientific realism, the scientific
realist's use of terms like “true” and “refers” becomes an issue, then the scientific realist can explain what he or she
means by those terms by citing the deflationist's equivalence thesis. In that way, we can prevent the issue of scientific
realism from degenerating into an argument over the virtues of the Correspondence Theory of Truth. It will not have
that inoculating effect if we then have to take up the second half of the deflationist's thesis, namely, that the
equivalence between “[s] is true” and s can be explained without appeal to honest-to-goodness real properties. But
perhaps Devitt can maintain that all we need in order to inoculate scientific realism is the first half of the deflationist's
thesis, the equivalence thesis itself.

Unfortunately, that will not work. I do not see that there is any sense in which we can avoid taking up the second half
of the deflationist's thesis if it is just false. We cannot just go around insisting that certain sentences are equivalent,
or that certain arguments are valid, and not expect to have to explain that equivalence in light of some conception of
the nature of language or mental representation. That is not to say that we have to explain the equivalence
specifically in terms of truth or reference. But we will have to have some conception of how by thinking we solve
problems, or of how by talking we facilitate cooperation, and that conception will have to say something about how
what we ought to think or say depends on how the world is at that time. In terms of this conception of language or
thought we will explicate any purported equivalance between sentences.

The need to provide some such account of the equivalence between “[s] is true” and s is particularly pressing, since
there are powerful reasons to deny that they are equivalent at all. Consider the following identity sentence:

(*) =“ is not true”

According to the equivalence thesis, “ ‘ is not true’ is true” equivalent to “ is not true.” But if we accept that they
are, then from the identity sentence (*), it seems we can derive the contradiction, “ is true and is not true.” The
fault here may not lie in the equivalence thesis. It could lie in the rules of Conditional Proof or Reductio ad Absurdum
1
that we will use in deriving the contradiction.
1 The T-schema and (*) yield a contradiction via a single identity substitution. So if we are to maintain the equivalence thesis and avoid paradox, then the

equivalence thesis cannot be formulated as a claim about the logical status of the T-schema (compare Devitt (2001)). However it can be understood as

the claim that [s] is true and s logically imply one another. In that case, we can preserve the equivalence thesis while avoiding contradiction by rejecting

the likes of Conditional Proof and Reductio ad Absurdum. For further discussion, see my 2005.

end p.129

But we cannot address this issue without doing real semantics. I know that many proponents of deflationism think
that the semantic paradoxes are somebody else's business, but I think the paradoxes clearly demonstrate that
deflationism, as I have defined it here, is mistaken. (That is not to say that the honest-to-goodness property in terms
of which we explain the equivalence has to be truth, as that has classically been conceived.)

So we cannot both insist on the equivalence thesis and avoid being drawn into a debate about semantics. One could
allow that scientific realism is essentially a semantic thesis, inasmuch as it ineliminably employs semantic terminology
that requires a substantive explication, and at the same time maintain that the semantic aspects of the thesis do not
matter inasmuch as the debates between defenders and critics never turn on those semantic aspects. But that is not
right either. What SRS says is that if a theory is well-established, then it is probably true; and whether we should
agree to that does depend on substantive questions about the nature of truth. Moreover, a fairly persuasive reason to
be a scientific antirealist is that it is hard to see how a theory of reference might extend to the scientific terms for
unobservables. In fact, as I will argue in the next section, it is hard to see why Devitt should not be a scientific
antirealist himself on precisely those grounds.

II. Scientific Realism and the Causal Theory of Reference

It's not that Michael Devitt doesn't want to talk about semantics. He's the leading exponent of the causal theory of
 p p p p p y

reference. It's just that he doesn't think we should have to talk about semantics in debating scientific realism.
Suppose nonetheless that we succeed in dragging Michael, kicking and screaming, into a debate about semantics. Can
he win the scientific realism debate on his semantic platform?

Just to remind you of the sort of theory at issue, let me quote this very brief expression of the causal theory of
reference:

At a dubbing, a name is introduced by grounding it in an object. There is a causal chain linking the ability
gained at the dubbing to the object. In virtue of that link, the reference of the name is fixed as the object.
(Devitt and Sterelny, Language and Reality, 1987, p. 56)

Moreover, what goes for names, goes for some other sorts of nouns as well.

Will this sort of theory work as well for theoretical terms such as “electron”? Even Devitt seems to have his doubts. In
his book Realism and Truth, Devitt writes:

Can we hope for a causal theory of those scientific terms which refer to unobservables? Any such theory faces
a difficulty over and beyond those that all causal theories face. The

end p.130

relation between a person and an object she perceives seems a promising one on which to build any causal
theory. But this relation is not available for scientific terms, because their referents cannot be perceived. What
then is the causal relation to these referents that determines reference? (Realism and Truth, 2nd edn, 1997, p.
122)

Perhaps, then, a proponent of the causal theory should allow a description theory of reference, at least in the case of
theoretical terms. But, no:

On this view, our term “electron” would refer to electrons in virtue of the fact that our theory associates the
term with various descriptions (a weighted most of ) which picked out electrons. . . . From our present
perspective we can see that, according to the description theory, most of the terms of past theories do not
refer, because there is nothing that the associated descriptions pick out; our past theories were mostly wrong.
So it would be probable that the terms of our of present theories did not refer. Thus Scientific Realism would be
false. (Realism and Truth, 2nd edn, 1997 p. 122)

In this passage from Realism and Truth, Devitt refers the reader to p. 201 of his book, Designation, where we find
these words:

Our problem with a theoretical term is that its referent cannot be perceived. . . . I suggest that what we seek
here is a relation consisting of an instrument “perceiving” the referent and our “reading” of the instrument: we
are “perceiving the referent through the instrument.” It would, of course, be difficult to fill out the details of
this suggestion. I shall not attempt to do so. (Designation, 1981, pp. 200–1)

There follow two more paragraphs addressed to the matter, but Devitt does not pretend that those two paragraphs
contain the solution. As far as I am aware, Devitt has not addressed the matter in greater detail anywhere else.

Devitt's causal theory of reference faces an even more basic problem, which he himself has formulated very clearly (in
his book co-authored with Kim Sterelny). He calls it the qua-problem. Here I paraphrase from my own 1987 review of
Devitt and Sterelny's book (p. 269): Of all the things involved in the causation of the act of perception that leads to
the introduction of a term, which will be the term's referent? What makes it the case that “Nana” comes to be the
name of a cat and not the name of a head of a cat or the name of a time-slice of a cat? Devitt and Sterelny reply that
the type of the referent is determined by a descriptive element in the thought that leads to the introduction of the
term (pp. 64–5). Thus they reject the “pure-causal theory” in favor of what they call a “descriptive-causal theory.”
But as Devitt and Sterelny themselves point out, the qua-problem arises as well in accounting for the reference of the
relevant descriptive elements. Since a solution along the same lines as before would beg the question, they are forced
to concede in the end that they have no comprehensive theory of reference at all (p. 75).

Nonetheless, Devitt is confident that in the end we will have the naturalistic theory of reference that we need for the
Correspondence Theory of Truth. In his book, Coming to Our Senses, he lists three strategies for explicating the
reference relation in a naturalistic way: to appeal to the historical cause of token

end p.131

representations (that's his own favored approach); to appeal to the reliable cause of token representations of a given
type (an approach he attributes to Stampe and Dretske); and to appeal to the teleological cause of the given type of
representation (an approach he attributes to Papineau and Millikan). He then says:

All known theories adverting to these three relations seem to face difficulties, and I doubt that any of them
gives a generally satisfactory explanation of the ultimate referential link. Nevertheless, I think that we should
assume that a satisfactory theory or theories along these lines can be found. (Coming to Our Senses, 1996, p.
161)
 p p p p p y

On the contrary, I think that at this stage in history we should be very pessimistic about the prospects for a
naturalistic theory of reference. It is always good to have different researchers pursuing various theoretical strategies.
But the utter failure of any of these theories to produce a remotely plausible account of the representation relation for
any kind of (conceptual) representation whatsoever should persuade most of us to give up.

The Correspondence Theory of Truth, as Devitt understands it, with its commitment to a naturalistically explicable
reference relation is one expression of the hope that the way the world really is exercises some kind of control over
what we say. I share that hope, but I do not think it commits me to the Correspondence Theory of Truth. To realize
that hope, what we need is some conception of the norms of discourse that meets the following two conditions. First,
whether a speaker speaks in conformity to those norms is at least in part a matter of how things really are in the
world. Second, speakers of the language have some ability to tell whether a speaker is speaking in accordance with
those norms. My own view is that we can have a conception of the norms of discourse that will meet these conditions
2
but which does not explicate truth in terms of real reference relations (see my 2003).
2 Of the participants in the Truth and Realism conference, the one who earliest recognized that the naturalistic theory of reference was a lost cause was

Richard Rorty. I should acknowledge that he got to me at a young and impressionable age. My first exposure to the realism issue was in 1977, when I

took a course from him, in Frankfurt, Germany, titled Der Realismusstreit in der zeitgenössischen amerikanischen Sprachphilosophie.

III. The Epistemology of Realism

There is another difficult term in Michael Devitt's formulation of scientific realism, and that is the term “well-
established.” It is only the well-established theories that we can assume are probably true. It is only the
well-established theories whose terms presumably do refer. For Devitt, what it means to say that a theory is
well-established is that it is supported by abductive inference, by which he means inference to the best explanation.
(He certainly does not mean that the institutionally authorized authorities do in fact believe it.) In any given case, the
reason

end p.132

to believe that unobservable entities exist is that the hypothesis that they do is supported by abduction. He writes:

The basic argument for the unobservable entities is simple. By supposing they exist, we can give good
explanations of the behavior and characteristics of observed entities, behavior and characteristics which would
otherwise remain completely inexplicable. (Realism and Truth, 2nd edn, 1997, p. 108)

Devitt denies that scientific realism is primarily an epistemological thesis (this volume, chapter 5, section 1), but he
will have to admit that one of the main issues between scientific realists and scientific antirealists has been the
validity of the method of abduction as a test of truth.

In fact, I do not think that the persuasive reason to believe in unobservable entities is always abduction. I am no
historian, but my impression is that very often in the history of science, the question whether a certain theory gives a
good explanation of some observable phenomenon has come down to precisely the question of whether the entities
posited by the theory do exist. If the question whether a theory does provide good explanations comes down to the
question whether those things exist that the theory says exist, then the question whether those things do exist cannot
be resolved by deciding whether the theory gives good explanations.

The obvious move to make in response to this objection would be to try to prise off some aspect of goodness in
explanations that does not depend on the ontological commitments of the explaining theory and then to claim that the
ontological commitments of the theory can be satisfied for free provided that the explanations are good in these
special respects. I cannot show that such a strategy cannot succeed. But perhaps I can persuade you not to try it by
pointing out that there are other tests for the existence of theoretical entities, over and above the explanatory
success of the theories that posit their existence, that have been employed in addition. In particular, the scientific
community has sometimes required in addition that the entities posited could be literally counted or could be tracked
as individuals.

Let me give you one example. Throughout the nineteenth century, many accomplished chemists doubted the
existence of chemical atoms, what we now call molecules. Here I follow the account in Mary Jo Nye's excellent book,
Molecular Reality (1972). Nye quotes the German chemist Friedrich August Kekulé (the discoverer of the ring
structure of the benzene molecule), who in 1867 wrote the following:

The question whether atoms exist or not has little significance from a chemical point of view; its discussion
belongs rather to metaphysics. . . . From a philosophical point of view, I do not believe in the actual existence
of atoms, taking the word in its literal significance of indivisible particles of matter—I rather expect that we
shall some day find for what we now call atoms a mathematico-mechanical explanation which will render an

end p.133

account of atomic weight, of atomicity, and of numerous other properties of the so-called atoms. (Quoted in
Nye, pp. 4–5)

Such doubts were put to rest in the scientific community only after the French chemist Jean Perrin succeeded in
 p p p p p y

estimating Avogadro's number, the number of molecules in a single mole of a substance. By 1912, Henri Poincaré,
who himself had long been one of the skeptics, could write the following words:

The long-standing mechanistic and atomistic hypotheses have recently taken on enough consistency to cease
almost appearing to us as hypotheses; atoms are no longer a useful fiction; things seem to us in favor of
saying that we see them since we know how to count them. . . . The brilliant determinations of the number of
atoms made by M. Perrin have completed this triumph of atomism. . . . In the procedures deriving from
Brownian movement or in those where the law of radiation is invoked . . . in the blue of the sky . . . when
radium is dealt with . . . the atom of the chemist is now a reality. (Quoted in Nye, p. 157)

If we examine Perrin's method, we will find that it is very far from what philosophers usually think of as abduction.
Perrin relied on observations of very small particles suspended in water. He prepared small batches of tiny spherules
from something called “gamboge,” which is a kind of gum resin. Using microscopes and centrifuges, Perrin determined
the size and density of individual gamboge spherules, and he measured the density of their distribution at various
points in a tiny column of water. Recall that the characteristic weight of a substance is a number that can be
determined on the basis of the behavior of the substance in forming compounds. If M is the characteristic weight of a
substance, then M grams of the substance is called a mole of the substance. Let us assume that the number of
molecules of a substance in one mole—Avogadro's number—is always the same. Let us take for granted also the ideal
gas law, PV=kT, where P is pressure, V is volume, T is temperature, and k is a constant, the same for one mole of any
gas. And let us assume that the ideal gas law applies as well when P is osmotic pressure, that is, the pressure that a
substance suspended in a solvent exerts on a semipermeable membrane that allows the solvent to pass through.
Moreover, we help ourselves to basic assumptions about the relation between pressure, gravity, and mass.

Suppose we have a column A containing hydrogen and a column B of gamboge spherules suspended in water. And
suppose we choose two points, xB and x B, on B (the latter being higher up) and measure the number of spherules per
unit volume at each of these two points. Call the results nB and n B, respectively. Moreover, we can measure the
pressure of the gas in column A at various points. So suppose we choose two points on column A, xA and x A, such that
the pressures at xA and x A are pA and p A, respectively, and (p A/pA)=(n B/nB). Let HA be the distance between xA and
x A. Let HB be the distance between xB and x B. Let mA be the mass of an individual molecule of hydrogen (since we
are supposing for the sake of argument that molecules exist). And let mB be the mass of a

end p.134

single spherule of gamboge. Finally, let d be the density (mass per volume) of the gamboge, and let D be the density
of water. From the assumptions that we have laid down and the fact that (p A/pA)=(n B/nB), we can derive, by
mathematics alone, the conclusion that (mA×HA)=(mB(1 d/D)×HB). But HA, HB, mB, d and D can all be measured by
observation. So we can calculate mA, the mass of an atom of hydrogen. For the details of the argument (requiring a
certain amount of reading between the lines), see Perrin 1990 (1913).

What I expect this summary of Perrin's method to suggest is that the basis for accepting the existence of unobservable
entities is not aptly described as inference to the best explanation. It is not the case that scientists are reliably
persuaded that unobservable entities exist when they find that the hypothesis that such entities exist is their best
explanation of the phenomena that they observe. In the case of chemical molecules, the tide of informed opinion
turned in favor of existence only once it became possible to count their numbers and measure the properties of
individual ones. Granted, Perrin's calculation rested on the prior assumption that hydrogen molecules exist. And what
impressed Poincaré (and Perrin too) was not just that Perrin got a result but that his result was approximately the
3
same as that which could be obtained by other means.
3 At the ‘Truth and Realism’ conference James Ladyman pressed this point on me.

So one might say that the hypothesis that molecules exist and exist in certain quantities was the best explanation of
the fact that these several methods all converged on approximately the same number. But that is no defense of the
claim that the basis for believing in unobservables is inference to the best explanation. What persuades is not that the
hypothesis that molecules exist explains various observable phenomena. What persuades is that we get approximately
the same result when we try in various ways to count their numbers. That argument might be called an inference to
the best explanation, but then what matters is the kind of fact that is explained.

Here is why I think it is important to make this correction to Devitt's account of the epistemology of realism. Despite
Devitt's attempt to deny it, I suspect that an important motive for those who call themselves scientific realists is the
thought that there is a certain relation to the evidence such that if a theory stands in that relation to the evidence,
then we ought to accept that the theory is true. That might be plausible so long as we can give a simple, uniform
account of that relation. But when we start to look at the history of science and the arguments that have brought
about those changes in theory that we now, in retrospect, acknowledge as great advances, we may not find any single
relation that has reliably served as the test for truth. In that way too we may be deprived of any general thesis that
we can call scientific realism.

end p.135

References

Devitt, Michael. 1981. Designation. New York: Columbia University Press.

 p p p p p y

Devitt, Michael. 1996. Coming to Our Senses: A Naturalistic Program for Semantic Localism. Cambridge: Cambridge
University Press.

Devitt, Michael. 1997. Realism and Truth, 2nd edition with a new afterword. Princeton, NJ: Princeton University Press.

Devitt, Michael. 2001. ‘The Metaphysics of Truth.’ In Michael Lynch, ed., The Nature of Truth: Classic and
Contemporary Perspectives. Cambridge, MA: MIT Press, pp. 579–611.

Devitt, Michael. 2005. ‘Scientific Realism.’ This volume, ch. 5.

Devitt, Michael and Sterelny, Kim. 1987. Language and Reality: An Introduction to the Philosophy of Language.
Cambridge, MA: MIT Press.

Gauker, Christopher. 1987. Review of Devitt and Sterelny, Language and Reality. Teaching Philosophy, 10: 269–71.

Gauker, Christopher. 2003. Words without Meaning. Cambridge, MA: MIT Press.

Gauker, Christopher. 2005. ‘Semantics for Deflationists.’ In J. C. Beall and Bradley Armour-Garb (eds), Deflationism
and Paradox. Oxford: Oxford University Press. 148–76.

Nye, Mary Jo. 1972. Molecular Reality: A Perspective on the Work of Jean Perrin. American Elsevier, New York.

Perrin, Jean. 1990. Atoms, translated by D. Ll. Hammick. Ox Bow Press Woodbridge, CT. (Originally published in
French in 1913.)

end p.136
 p p p p p y

7 Abundant Truth in an Austere World

What is real? Less than you might think. We advocate austere metaphysical realism—a form of metaphysical realism
claiming that a correct ontological theory will repudiate numerous putative entities and properties that are posited in
everyday thought and discourse, and also will even repudiate numerous putative objects and properties that are
posited by well-confirmed scientific theories. We have lately defended a specific version of austere metaphysical
realism which asserts that there is really only one concrete particular, viz., the entire cosmos (see Horgan and Potr
(2000, 2002), Potr (2003)). But there are various potential versions of the generic position we are here calling
austere metaphysical realism; and it is the generic view that constitutes the ontological part of the overall approach to
realism and truth that we will describe here.

What is true? More than you might think, given our austere metaphysical realism. We maintain that truth is
semantically correct affirmability, under contextually operative semantic standards. We also maintain that most of the
time, the contextually operative semantic standards work in such a way that semantic correctness (i.e., truth) is a
matter of indirect correspondence rather than direct correspondence between thought or language on the one hand,
1
and the world on the other.
1 This paper is a broad-brushstroke presentation of the general picture we advocate concerning truth and realism. To some extent we will be summarizing

and somewhat re-packaging aspects of the position that has been developing in a series of papers. See, for instance, Horgan (2001b) and several of

Horgan's earlier papers cited therein, Horgan and Potr (2000, 2002), Horgan and Timmons (2002), Barnard and Horgan (forthcoming), Potr (2003),

Potr and Strahovnik (2004). But we will also incorporate some new ideas, in a way that rests in part on work that Horgan has done with John Tienson

on connectionism and the philosophy of psychology, and on ongoing discussions between Horgan and Potr about particularism, semantic normativity,

common-sense realism, and scientific realism.

When correspondence is indirect rather than direct, a given statement (or thought) can be true even if the correct
ontology does not include items answering to all the referential commitments (as we will here call them) of the
2
statement.
2 We use the expression “referential commitments” in order to evoke the idea that such commitments arise via the use of logico-grammatical “referential

apparatus” in language and thought—e.g., the use in language of singular terms and existential quantifiers, and the mental use of thought-constituents

expressible inguistically by such linguistic items. Read “referential” here as “arising from the use of logico-linguistic referential apparatus”—not as

“purporting to denote some putative item within the correct ontology of the world.” In Barnard and Horgan (forthcoming), the alternative expression

“ideological commitments” is employed instead of “referential commitments”. (Also, Barnard and Horgan employ “mediated correspondence” instead of

“indirect correspondence.”)

This means that even if a putative object is repudiated by a correct ontological

end p.137

theory, ordinary statements that are putatively about that object may still be true. For instance, the statement “The
University of St Andrews is in Scotland” can be semantically correct (i.e., true) even if the right ontology does not
include any entity answering to the referring term “The University of St Andrews,” or any entity answering to the
3
referring term “Scotland.” This general approach to truth is what we call contextual semantics.
3
Here we use “semantics” to cover not only language/world relations but also thought/world relations. Semantics concerns matters like intentionality,

reference, and truth, both for public language and for thought.

We will here call our package-deal position, comprising austere ontology and contextual semantics, austere indirect-
correspondence realism (for short, AIC realism). In the first section we will elaborate somewhat upon our version of
AIC realism, stressing some prima facie advantages of AIC realism over various alternative approaches to truth and
realism that are currently on offer in philosophy. The remainder of the paper will be organized around three kinds of
skeptical challenge that can be raised against AIC realism. We will seek not only to meet these challenges, but also to
use them as guides for further developing—and for motivating—the specific version of AIC realism that we seek to
defend.

Briefly, the three challenges are these. First, our appeal to contextual variability of semantic standards faces an
unhappy dilemma: the need to make a choice between (i) claiming (implausibly) that thoughts and statements that
seem categorical (e.g., “The University of St Andrews is in Scotland”) are really implicitly relativistic and
non-categorical in content, or instead (ii) claiming (again implausibly) that thought and discourse harbor extensive
unnoticed semantic ambiguity (since the specific meaning of one's thoughts and statements depends so heavily,
according to contextual semantics, upon implicit contextual factors). Second, AIC realism, because it advocates
austere ontology, threatens to be grossly contrary to common-sense beliefs about the world, and/or grossly contrary
to what the best science seems to tell us about the world. The AIC realist, after all, denies the existence of vastly
many of the entities posited in ordinary thought/talk and in scientific thought/talk. Third, skeptical doubts are apt to
arise about whether an adequate general account—in terms of general, systematic, normative principles—can be
given of matters like (i) indirect-correspondence semantic standards of various sorts, and (ii) the dynamics of
contextual variation in semantic standards.

Concerning the first challenge, we will bring to bear some ideas recently deployed by Horgan and Timmons (2002) in
 p p p p p y

an effort to make sense of Putnam's notion of conceptual relativity. We will harness these ideas as a proposed way of
going between the horns of the dilemma.

end p.138

Concerning the second challenge, we will have quite a lot to say about how common-sense beliefs and scientific beliefs
fare, under the approach we advocate. We will be arguing (i) that AIC realism accommodates most of common sense
and science very well, (ii) that AIC realism plausibly explains why common sense balks so strongly at austere
ontology, and that the balking reaction (because it is thus explainable) does not constitute good grounds for rejecting
AIC realism, and (iii) that common sense itself, when it turns ontologically reflective, actually generates strong
theoretical grounds for an austere ontology.

Concerning the third challenge, we will argue that semantic normativity in most contexts, and likewise the dynamics
of contextual variability in semantic normativity, very likely are too subtle and complex to conform to general,
systematic, principles. As we will put it, semantic normativity is very likely quasi-particularistic, rather than rule-like
(cf. Potr 2000, Potr and Strahovnik 2004). The case for quasi-particularism about semantic normativity rests partly
upon ontologically reflective common sense: certain common-sense reflections that motivate an austere ontology also
motivate semantic quasi-particularism. We will also argue that those who maintain that there must be general
rule-like semantic principles, in order for humans to be capable of mastering semantic normativity, are making highly
dubious empirical assumptions about the workings of human cognition.

1. Contextual Semantics, Metaphysical Realism, and Ontological Austerity

There are various ways to call into question the idea that all referential commitments, in language and thought, are
really ontological commitments. There are paraphrase strategies of one sort or another. There are fictional
approaches, and error theories. There are epistemic theories that seek to reduce truth to epistemically warranted
affirmability, or to some idealized variant of it. There are global irrealist approaches which effectively deny that
there's any such thing as genuine ontological commitment at all, understood in a metaphysical realist way. And there
are approaches to truth according to which issues about truth and issues about ontological commitment have very
little directly to do with one another. Let us briefly say something about various prima facie advantages of AIC realism
over these various other approaches. Along the way we will bring into focus some further features of our favored
version of AIC realism.

There are various ways of trying to “paraphrase away” discourse that is referentially committed to putative entities
that one considers ontologically dubious. One way is to offer paraphrases that just drop out the relevant referring
terminology altogether (as in paraphrasing “She has a charming smile” as “She smiles charmingly”). Another is to
offer paraphrases that effectively identify the erstwhile offending entities with entities one considers ontologically
more respectable (as in identifications of numbers with sets of one sort or another).

end p.139

But there are reasons to be very dubious about the paraphrase approach as a general strategy for avoiding
questionable-looking ontological commitments. One problem is that often there are no terribly plausible candidates for
paraphrasing. How, for instance, might one plausibly paraphrase the statement “The University of St Andrews is in
Scotland” into some statement that eschews university-talk and nation-talk? A further problem is that often, among
the marginally eligible candidate paraphrases, there will be far too many that look equally (albeit marginally) eligible.
For instance, there will be too many equally eligible ways to paraphrase talk about the University of St Andrews into
talk about things—or sets or mereological sums of things—like people and buildings and computers and vehicles and
such.

An obvious advantage of AIC realism, in comparison to the paraphrase approach, is that the former eliminates the
need for systematic “paraphrasing away” of discourse with referential commitments that one has reason to think are
not genuine ontological commitments. Instead of reformulating the relevant claims in an ontologically austere
discourse, one instead goes “soft on truth” for the relevant discourse: one claims that the discourse operates under
indirect-correspondence semantic standards (IC semantic standards)—and that truth, for a discourse governed by such
standards, is just semantic correctness under those standards. The upshot is truth without ontological commitment,
and without the need for systematic paraphrasing.

Lately there has been some enthusiasm for “semantic pretense” theories of ontologically questionable thought and
discourse: theories that treat such thought/discourse as being effectively a form of fiction (e.g., Walton 1990,
Crimmins 1998). We take it that according to these views, numerous beliefs that we common-sensically hold true are
not really true at all—just as it's not really true that there's a person named Santa Claus who lives at the North Pole
and dispenses presents at Christmas, or that there was a person named Sherlock Holmes who lived on Baker Street in
London and was a brilliant sleuth. An obvious prima facie advantage of AIC realism, in comparison to semantic
pretense theories, is that the former allows us to respect the persistent belief that numerous ordinary beliefs are
literally true, whereas semantic pretense theories do not. AIC realism has the same advantage over error theories of
common-sense thought and talk.

Epistemic reductionist theories of truth seem to exert a perennial attraction for some philosophers. Recently
 p p p p p y

influential versions include Putnam's some-time contention that truth is “ideal warranted assertibility” (Putnam 1981),
and Wright's some-time suggestion that truth is identical to a form of idealized warranted assertibility that he calls
“superassertibility” (Wright 1987). But such approaches face at least two very serious objections, not faced by AIC
realism. First, unless one idealizes all the way up to something like a “God's-eye epistemic vantage point,” there will
be persistent cases where truth and idealized warranted assertibility evidently diverge—for instance, (i) statements
about the distant past

end p.140

for which no extant evidence exists one way or the other, (ii) statements about certain goings-on in distant portions
of spacetime outside the light-cone of the human race, etc. Second, even if one idealizes so much that idealized
warranted affirmability coincides—or necessarily coincides—with truth (given the extent of idealization), it will
nevertheless be the case that when a statement is ideally warrantedly affirmable, this will be because it is
true—rather than its being true because it is ideally warrantedly affirmable. (This is what Wright (1992) calls “the
Euthyphro contrast.”) As we say, AIC realism does not face these problems—which constitutes a very considerable
advantage of AIC realism over epistemically reductionist accounts of truth.

Here we should pause to stress a certain feature of contextual semantics, as we construe it: semantic standards of
correct affirmability are likely to be closely intertwined with epistemic standards of warranted affirmability, even
though the former are not reducible to the latter. Such intertwining is entirely to be expected: since systematic true
belief is an evaluative ideal with respect to normative epistemological notions like justification and warrant, semantic
standards and epistemic standards ought to “fit” one another. Warranted thoughts/statements will be ones that are
likely to be true given the available evidence—i.e., likely to be semantically correct given the available evidence. Small
wonder, then, that contextually operative epistemic standards of warranted affirmability will be closely intertwined
with contextually operative semantic standards of correct affirmability—notwithstanding the fact that one can't reduce
4
semantically correct affirmability to epistemically warranted affirmability (or to some idealization of the latter).
4
There is thus intertwining of semantic and epistemic contextual standards. But the contextually operative epistemic standards can vary, to some extent,

even when the contextually operative semantic standards don't. This is because there can be other reasons for contextual variation of epistemic

standards. You may be acquainted with the departure time of your airplane. But if your precise arrival at a destination is of extreme importance, you will be

taking several additional precautions before you really trust your airplane itinerary. Your epistemic standards will be set much higher. Intuition tells you that

semantic standards do not necessarily get higher in such a case.

Another, related, point to stress is this: facts about contextually operative standards of epistemic warrant can be
expected to be a good guide to contextually operative semantic standards (given the fit between them). Consider, for
instance, what would need doing in order to obtain good epistemic warrant for the claim that Warner Brothers Films is
owned by the Miramax Entertainment Corporation. The relevant evidential standards just don't require obtaining good
evidence that there are entities, viz., WARNER BROTHERS FILMS and MIRAMAX ENTERTAINMENT CORPORATION , included among
the furniture of the universe. (Here and occasionally below, we resort to Putnam's capitalization convention, as a way
of signaling ontological uses of terms that, according to our general view, often are employed without ontological
commitment because the contextually operative semantic standards governing these terms can be indirect-
correspondence standards.)

end p.141

One way to stop worrying about which referential commitments of thought and language are genuine ontological
commitments is to claim that none of them are, and that the idea of a mind-independent, language-independent,
world is a metaphysical extravagance. This is global metaphysical irrealism (which is often combined with
epistemically reductionist views of truth). But we ourselves find global irrealism itself to be a doctrine so
metaphysically extravagant as to be not only wildly implausible but also well nigh unintelligible. Since these claims
have been argued for elsewhere (Horgan 1991, 2001b), we will not repeat the arguments here.

Lately there has been much enthusiasm for one or another version of “minimalism” or “deflationism” about truth—
roughly and generically, the view that the various instances of the Tarskian schemas T and F pretty much exhaust all
there is to the concepts of truth and falsity (e.g., Field 2001, Horwich 2001). One major disadvantage of
minimalism/deflationism is this: although there certainly are schema-T uses of the truth predicate, there are
important correspondence uses too. For instance, for someone who holds that moral judgments/statements have
nondescriptive overall content, one important right thing to say about them with respect to truth—under one
legitimate usage of “true”—is that they are neither true nor false. A significant advantage of AIC realism over
minimalism/ deflationism—more specifically, an important advantage of the contextual semantics component of AIC
realism over minimalism/deflationism—is that contextual semantics can smoothly accommodate both schema-T uses
of the truth predicate and correspondence uses, for modes of discourse for which (according to nondescriptivist
treatments of these modes of discourse) these uses diverge.
5
Here we will mention a few aspects of how this accommodation works.
5
For more details, see Timmons (1999), Horgan (2001b), Horgan and Timmons (2000).

Horgan and Timmons distinguish between tight and non-tight contextual semantic standards. A judgment/statement is
governed by tight semantic standards if those standards conspire with how the world is to render the
statement/judgment semantically correct or semantically incorrect; otherwise, the semantic standards are non-tight.
 p p p p p y

(If, for example, humor judgments are partially expressive of the judge's own sense of humor and thus can vary
among several people without anyone's being mistaken, then the semantic standards governing thought and talk
about what's funny are non-tight.) Second, contextually variable semantic standards govern the truth predicate itself.
Third, under one usage of “true” that is sometimes contextually appropriate—a correspondence usage—a
statement/judgment whose governing semantic standards are non-tight counts as neither true nor false. (Horgan and
Timmons claim that under this usage, moral statements/judgments are neither true nor false in their standard usage,
since normally they are governed by non-tight semantic standards.) Fourth, on another usage of the truth predicate
that is sometimes contextually appropriate,

end p.142

the truth predicate conforms to schema T. (Horgan and Timmons claim that on this usage, truth ascriptions to moral
judgments/statements are morally engaged meta-level judgments/statements, as are the first-order
judgments/statements to which they are ascribed. Such a truth predication is a fusion of semantic and moral
evaluation.)

Vagueness also makes trouble for standard versions of deflationism/minimalism about truth. It is extremely plausible
that thought-contents and statements applying a vague category to a “borderline case” are neither true nor false.
Suppose, for instance, that Jones is borderline-bald (and thus is also borderline non-bald). Then the natural thing to
say about truth and falsity is that the statement “Jones is bald” is not true, and also is not false. But the
deflationist/minimalist is hard pressed to avoid reasoning in the following way:

Suppose that “Jones is bald” is not true. By schema T, “Jones is bald” is true iff Jones is bald. Hence Jones is
not bald. By schema F, “Jones is bald” is false iff Jones is not bald. Hence, “Jones is bald” is false.

But the conclusion of this reasoning seems wrong. Since Jones is a borderline case of baldness, the statement “Jones
is bald” is neither true nor false.

In sum: We have briefly set forth some significant prima facie advantages of AIC realism, in comparison to a range of
other approaches to truth and realism now on offer. Doing so has involved some degree of further elaboration of the
position beyond what we said at the outset. In particular, we have stressed that under contextual semantics, semantic
and epistemic normative standards are apt to be closely intertwined, even though semantic standards are not
reducible to epistemic ones (or to some idealization thereof). We have also stressed the tight/non-tight distinction,
and the way it allows for contextually variable semantic standards to govern the truth predicate itself. In particular,
given a nondescriptivist treatment of moral discourse, it allows for both correspondence uses of the truth predicate
and schema-T uses.

2. Contextual Semantic Variation: Content Relativity vs. Identity-Preserving Difference

We turn next to some remarks about how to think about the workings of contextually variable semantic standards,
drawing upon material from Horgan and Timmons (2002). One idea that commonly gets invoked in connection with
context dependence is that contextual factors are an implicit aspect of the very content of a thought or sentence
involving such factors—so that the actual content is really relativistic content. For example, when one says that the
table is flat, one really means that it is flat relative to such-and-such standards. When one says that Miramax
Entertainment Corporation owns Warner Brothers Films, one is

end p.143

saying that “relative to such-and-such semantic standards,” Miramax Entertainment Corporation owns Warner
Brothers Films.

This approach seems just wrong. The trouble is that even when contextually variable semantic parameters are in play,
typically the discourse/thought governed by such parameters operates categorically: one is speaking/judging from
within a semantically committed stance in which one accepts the semantic standards as operative. When one says that
something holds relative to such-and-such standards, on the other hand (even if the “relative to such-and-such
standards” part is implicit), one need not be accepting those standards at all. People do sometimes talk/think in a
6
relativistic way, but when they do so they typically employ explicit relativization.
6
Indeed, we ourselves talk that way some of the time in setting forth our account of contextual semantics—as when we say, “Truth is semantically

correct affirmability under contextually operative semantic standards.” Here we should emphasize that we regard the contextual-semantics story as

applicable to virtually all the concepts/terms usable to tell the story, and in particular to the concept truth. Very often, indeed typically, one doesn't use

“true” in a relativistic way. On the contrary, one uses it categorically, from within a stance in which one not only accepts certain specific semantic-

correctness standards as contextually operative, but one lets them govern one's truth ascriptions too—the familiar (though not inevitable, on our view)

“schema T” usage of “true.”

But if indeed discourse/thought governed by contextual parameters is normally categorical in content rather than
implicitly relativistic, then the following question naturally arises: What is the semantic relation between a statement/
thought that is affirmed under such-and-such contextually operative semantic standards, and one (expressed the
same way) that instead is affirmed under so-and-so alternative semantic standards? A looming worry is that this
contextualist approach is effectively positing massive equivocation in the use of concepts and terms that are governed
 p p p p p y

by contextual parameters. Commonality of content, across different uses, is threatened.

Our position on this matter is as follows. Inter-context variation in operative semantic parameters (variation in
aspects of what Lewis (1979) called the “score in the language game”) constitutes identity-preserving
change/difference in meaning and concepts. This is what Horgan and Timmons (2002) call differance in
meanings/concepts (adapting Derrida's terminology), rather than what's ordinarily called “difference in
meaning/concept.” The idea is to think of meanings and concepts (whatever exactly they are) as items that can
exhibit, from one context of usage to another, certain identity-preserving differences—much as other kinds of entities
(e.g., persons) can differ in certain ways over time (e.g., in hair color).

To illustrate such differance consider the following two-part remark that might be made by philosophers while
discussing ontology and common-sense belief: “Are there corporations? Of course! Are there really corporations?
7
No!”
7
This is a specific philosophical usage of “really”—not the only one that might be appropriate in philosophical contexts, to be sure, and one that is different

from most uses of “really.” This specific usage is employed to overtly signal a shift into a mode of thought/discourse governed by direct-correspondence

semantic standards.

One can

end p.144

assert both parts of this remark in close succession to one another, but not in the same breath. There is a subtle
semantic variation at work across the two uses of “corporation,” but not a semantic difference that constitutes
anything nearly as great as ordinary semantic ambiguity (as with “bank” as financial institution or edge of a river).
The word “corporation” has the same meaning when employed under different semantic standards governing the first
part of the remark and the second part, even though this contextual variation is indeed a differance: it's an identity-
preserving difference in meaning. Likewise, mutatis mutandis, concerning the concept expressed by “corporation.”

Another way one might put the point is this. Certain terms, as used in two different token statements made under
somewhat different settings of the contextually variable score in the language game, are weakly synonymous but not
strongly synonymous. One can also use this terminology about the respective thought-contents expressed by the
statements. Statements/thoughts that are weakly but not strongly synonymous do exhibit certain meaning/content
differences, but these (again) are much less stark than ordinary semantic ambiguity—and the differing uses exhibit
much more semantic commonality despite these differences.

3. Accommodating Common Sense, Mostly

Let us now consider the status of common-sense thought and talk, within a picture that weds contextual semantics to
an austere ontology. First, we will argue that common sense gets very largely reconciled with austere ontology, under
our proposed contextual semantics. Second, we will argue that contextual semantics plausibly explains why and how
common sense balks at the claims of austere ontology—an explanation that accommodates something importantly
right about thus balking, while also revealing the subtle ways that common sense confuses itself in thus balking (and
in addressing ontological issues). Third, we will argue that there are powerful considerations from within common
sense itself that generate serious theoretical pressure toward some kind of austere ontology—i.e., an ontology that
eschews numerous entities to which common sense itself is referentially committed. We will take up the first of these
themes in this section, and the next two in sections 4 and 5 respectively.

It is important to realize that AIC realism delivers full-fledged, bona fide, truth despite the austerity of the ontology.
Truth under indirect-correspondence semantic standards (IC semantic standards) is not some sort of second-class
semantic status, something lesser than full-fledged or genuine truth. To think otherwise is to be too much in the grip
of the very picture of truth that is here being repudiated. Indirect correspondence is perfectly fine, as a mode of
correspondence. There are numerous ways that people engage in referential commitments that do not actually involve
ontological commitment, and doing so often serves people's thinking/communicating/living purposes very well. Think,
for instance,

end p.145

of talk of universities, nations, corporations, religions, or musical works of art. We “quantify over” such entities freely
in various modes of thought/discourse, and in general the contextually operative semantic standards certainly do
require of the world that it be one way rather than another, in order for a thought/statement involving such
referential commitments to count as semantically correct—i.e., as true. But in order for the world to be one of the
ways it might be that would render true our ordinary thoughts/claims about such entities (in typical contexts of
usage), it simply is not required that the right ontology should include entities answering to terms like “St Andrews
University,” “Scotland,” “General Motors Corporation,” “Catholicism,” or “Mozart's 27th piano concerto.”

How then is one to think of truth, qua indirect correspondence? How is one to understand the contention that the
world makes the statements true, but not in a way that involves ITEMS answering to the statements' referential
commitments? There is actually a familiar conception of truth conditions that can here be invoked. Think of truth
conditions for a thought/statement as constituted by something like a set of ways the world might be (“possible
worlds”) such that if the world is one of those ways then the statement/thought is true (false).
 p p p p p y

There are various ways one might “do the ontology” of such a conception of truth conditions. For example, each of
these ways the world could be may get construed as a maximal cosmos-instantiable property. One would probably
want to refine such a picture in various ways, we suspect—for instance, to accommodate vagueness and indexicality.
(So-called “centered” ways the world might be, with a designated location to be thought of as the location of the
thinker/speaker, could perhaps accommodate indexicality. One might need to soften up on the idea of a truth
condition as a determinate set of ways the world might be, to accommodate vagueness.) But the core idea remains: a
statement S with various referential commitments in it has a truth condition TC(S)—roughly, a set of maximal
cosmos-instantiable properties—such that S is true provided that the cosmos instantiates one of these properties in
TC(S), and is false provided that the cosmos does not instantiate one of these properties in TC(S). The truth
conditions for a statement S with referential commitments to universities, nations, family trusts, or the like can
perfectly well be satisfied—i.e., the cosmos can perfectly well instantiate one of the properties in the set TC(S)—even
if the furniture of the cosmos does not include such putative entities as UNIVERSITIES, NATIONS, or CORPORATIONS.

4. Common Sense vis-à-vis Austere Ontology: The Role of Scorekeeping Confusions

Common sense balks, and balks strongly, when the claims of some austere ontology are put forward. If we say to you,
“There aren't really such things as tables, chairs, or people,” you are apt to react by judging such a claim
preposterous—

end p.146

not just false, but screamingly false. That this reaction is apt to occur, and to occur with this kind of strength and
intensity, is a datum—one that needs to be dealt with, within the overall position we are advocating.

Our position does not directly accommodate the datum, in a straightforward manner; i.e., our position does not treat
the balking reaction as correct, and does not vindicate it. Rather, on our view you are making a mistake if you reject
our claim that there are no such things as tables, chairs, or people. For, as we intend the claim (viz., as a claim that is
being made under direct-correspondence semantic standards (DC standards)), what we are saying is true if indeed
some austere ontology is the correct one.

We take it that one serious obligation of a philosophical theory is to give a suitably respectful treatment of intuitively
plausible judgments/statements that the theory rejects—all the more so if the judgments are as intuitively plausible
as are judgments like “Of course there really are tables, chairs, and people” (thought/said in response to our own
claim that there are not really such things). Respectfulness means, inter alia, that although the judgments/statements
in question get treated as embodying some kind of error, it's not a silly or stupid or merely careless error. Rather, the
account will be better if the error is treated as particularly subtle, and one that people are particularly apt to make
given the normal workings of their cognitive apparatus.

Let us introduce the idea of a competence-based performance error. The thought is that although the cognitive system
falls short of its own competence in generating the given judgment or intuition or belief, nonetheless it is producing
8
this performance error in a way that in some sense emanates from competence.
8 For further discussion of this idea, with applications of it to philosophical puzzlement about the freedom/determinism issue and about the issue of mental

causation, see Graham and Horgan (1994) and Horgan (2001a).

Think, for instance, of the Mueller-Lyer illusion. The cognitive system is working the way it is evolutionarily designed
to, in producing a perceptual presentation of the one horizontal line as longer than the other. If you form the
perceptual belief that one of the horizontal lines is longer than the other, then you are committing a
competence-based performance error. Your visual system is working the way it has evolved to work (although it
happens here to be yielding up a visual illusion), and you are forming a mistaken perceptual belief by reliance on the
presentational content of your visual experience.

There's something right and something wrong, in your belief-forming process. The something right is that the lines
really do look to be not the same length. Someone not knowing better would justifiably form the belief that they are
not the same length. Something has gone wrong, though. The lines have in fact the same length, as can be
determined by ordinary measurement techniques. You are competent to tell whether they are the same length or not;
they are the same length; and yet here you are, judging them to be different in length.

end p.147

It is important, too, that the lines are apt still to look the different lengths even after the measurement has been
carried out. They will still look that way even when you realize that they are actually the same length. The
appearance will persist (or at least there will be a strong tendency for it to do so), even after you come to believe that
it is non-veridical.

Return now to the fact that common sense balks, seriously balks, at the claims of austere ontologies. We propose to
treat this as also a competence-based performance error, of a certain sort. Although it is mistaken to repudiate
austere ontologies on common-sense grounds, in a very important sense this is a mistake that is rooted in
9
semantic/conceptual competence itself.
9 We do not necessarily want to lean heavily on the analogy with the Mueller-Lyer illusion here. That example does put one kind of meat on the bones of

the idea of a competence-based performance error. We will be describing another kind of meat now, leaving it open how much our account does or does
 p p p p p y

not parallel the Mueller-Lyer case.

We will describe the basic picture in eight steps. First: as competent users of concepts and terms that are governed by
various implicit contextual parameters (aspects of what Lewis called the score in the language game), people typically
deal with these parameters so smoothly and competently that they do not even notice them. Witness, for example,
Lewis's charming little spiel about Bruce the cat, in the section of Lewis (1979) in which he is illustrating how well we
accommodate to the current contextually relevant score that determines which contextually eligible referent of a
definite description “the F” (e.g., “the cat”) counts as most the salient F (and hence as the referent of “the F”). He
writes:

Imagine yourself with me as I write these words. In the room is a cat, Bruce, who has been making himself
very salient by dashing madly about. He is the only cat in the room, or in sight, or in earshot. I start to speak
to you:

The cat is in the carton. The cat will never meet our other cat, because our other cat lives in New
Zealand. Our New Zealand cat lives with the Cresswells. And there he'll stay, cause Miriam would be sad
if the cat went away.

At first, “the cat” denotes Bruce, he being the most salient cat for reasons having nothing to do with the
conversation. If I want to talk about Albert, our New Zealand cat, I have to say “our other cat” or “our New
Zealand cat.” But as I talk more and more about Albert, and not any more about Bruce, I raise Albert's salience
by conversational means. Finally, in the last sentence of my monologue, I am in a position to say “the cat” and
thereby denote not Bruce but rather the newly-more-salient Albert. (Lewis 1979, p. 241)

We accommodate to score-changes without necessarily realizing that we are doing so, and without necessarily
noticing that contextual parameters are in play at all. Witness, for other examples, the other kinds of phenomena
Lewis canvasses in that landmark paper, all of which he persuasively argues are ones in which implicit, contextually
variable, parameters of “score in the language game” are in play. Humans appear to be hugely good at scorekeeping
in most ordinary

end p.148

contexts—so good that they do it without noticing that implicit semantic parameters are in play, and that they vary.

Second: normally when one employs certain categories, and certain terms expressing them, there is a very strong
default presumption in play that the contextually operative semantic standards are ones under which certain
affirmative, referentially committing, thoughts/statements employing those categories will be true (i.e., semantically
correct).

Third: on the view we advocate, contexts in which DC semantic standards are being employed are extremely rare and
rarefied: they are contexts of serious ontological inquiry.

Fourth: when we say to you “There are no tables, chairs, or people,” there is a strong tendency for you to react to this
statement as though it were being made under relatively typical contextually operative standards for categories like
“table,” “chair,” and “person.” This is partly because of the default presumption just mentioned (the second point), and
partly because of the extremely rarefied nature of thought/discourse that's conducted under DC semantic standards
(the third point). There is a strong tendency, that is, for you not to accommodate properly to the contextually
operative score in the language game that governs our own remark (even though normally people accommodate to
one another's remarks smoothly and naturally), and for you instead to respond to us with a remark that itself is
governed by more ordinary semantic standards (IC standards) for categories like “table,” “chair,” and “person.”

Fifth: so there is something importantly right in your balking reaction. Your claim “Of course there are tables, chairs,
and people!” is certainly true, under the contextually operative standards governing your own usage. And your claim
is also in line with the strong default presumption that a discourse in which familiar, commonly used categories are in
play is governed by semantic standards in which certain affirmative claims employing those categories can turn out to
be true (i.e., semantically correct).

Sixth: but even though your balking reaction is in these ways competence-based, there is something importantly
wrong in it nonetheless. To wit: you have failed to accommodate to the semantic standards that are contextually
appropriate, given that serious ontological inquiry is the game we mean to be playing.

Seventh: also, you fail to appreciate that implicitly contextually variable semantic parameters are in play at all, and
that our remark was made under a different language-game score than your own. Here, the fact that you normally
accommodate to score-changes so smoothly and naturally that you do not even notice score-parameters and their
variation is contributing to the mistake you are making.

Eighth: the sense of oddness and screaming falsity of claims like “There are no people” is apt to persist, even for
someone who accepts (as we do) an ontology that does not include people. (More below on why we do.) The various
factors cited above explain why this should be so.

end p.149

This kind of mistake is what we will call a scorekeeping confusion. Given the overall picture we are presenting, such
 p p p p p y

scorekeeping confusions are entirely to be expected. This is because of the ways they emanate so naturally from one's
own semantic/conceptual competence, including one's competence at normally handling implicit contextual parameters
so smoothly and naturally that one typically does not even notice that they are there.

So the upshot is that your balking reaction is being respectfully explained away as a certain kind of competence-based
performance error—a scorekeeping confusion on your part. Although this aspect of common-sense belief gets
repudiated by our account, rather than being vindicated, our account has the significant virtue of explaining
nonetheless why this aspect of common-sense thinking occurs—and also why it will likely continue to exert a strong
intuitive pull, even upon those who come to accept that the right ontology is some kind of austere ontology.

The larger upshot so far, concerning common sense, is (1) that common sense is mostly accommodated within the
picture we are pushing (as argued in the preceding section), and (2) that common sense's mistaken tendency to balk
at austere ontological claims can be plausibly explained as a subtle kind of competence-based performance error—viz.,
a scorekeeping error.

5. Common-Sense Pressures toward Austere Ontology

Let us now argue that common sense itself, when it goes reflective about matters ontological, generates strong
pressure in favor of an austere ontology. This can and does happen even without the realization that semantic
standards are contextually variable, or the realization that contextually operative semantic standards can require
either direct correspondence or some form of indirect correspondence, or the realization that most contexts of thought
and discourse are ones in which indirect-correspondence standards prevail. In sections 5.1 and 5.2 we will be
incorporating material from Horgan and Potr (2000), lightly edited to play up the common-sensical sources of the
considerations being advanced.

5.1. Metaphysically Lightweight Posits

We will use the phrase “metaphysically lightweight posit” in a deliberately vague way. Under this rubric we include
“socially constructed” institutional entities like corporations, universities, nations, and multinational organizations
(e.g., NATO). We also include various non-concrete cultural artifacts, like Beethoven's fifth symphony (as distinct from
concrete performances of it) and Quine's book Word and Object (as distinct from concrete tokens of it).

It is not plausible to common sense that institutional entities like corporations and universities are denizens of the
world-itself, over and above entities

end p.150

like persons, buildings, land masses, items of office equipment, and the like. Yet, when one considers whether it might
be possible to “reduce” a putative entity like a university to these other kinds of entities—say, by identifying each
university with some set of them (or some “mereological sum” of them), or by systematically paraphrasing statements
that posit universities into statements that do not—there is no plausible reductive account remotely in sight. For, the
project of systematically paraphrasing university-talk into statements that eschew all talk of universities looks
hopeless; and the trouble with attempts to identify a university with some set (or sum) of buildings, persons,
computers, etc. is that there are always numerous equally eligible candidate-sets (or candidate-sums), and there is no
reason to identify the university with any one of these over and against any of the others. Likewise, mutatis mutandis,
for other kinds of institutional entities like corporations and nations, and for non-concrete cultural artifacts like
Beethoven's fifth symphony and Quine's Word and Object.

So the appropriate common-sensical conclusion to draw, about thought and discourse employing these metaphysically
lightweight posits, is that the posits do not pick out items in the world-itself. Common sense does not, however,
thereby embrace the idea that thoughts and statements about such entities don't correspond to the world, don't
depend on how the world is. On the contrary: there are certain ways the world might be such that, if it's one of these
ways, then the statement “General Motors is a multinational corporation” is true; and there are other ways it might
be that would render this statement false. It's just that the correspondence-constituting ways the world might be need
not include such items as CORPORATIONS or NATIONS. Common sense viscerally realizes this—which means that common
sense viscerally appreciates that the truth of the statement about General Motors is a matter of indirect
correspondence.

Of course, one could accept these common-sensical-looking conclusions while remaining a robust ontological realist
about middle-sized dry goods like tables and chairs, about persons and other living organisms, and about the posits of
physics and the special sciences. But there is more to come.

5.2. The Non-Arbitrariness of Composition

Peter van Inwagen (1990) wields an important and powerful form of metaphysical argumentation that has been too
little appreciated in philosophy. He poses what he calls the Special Composition Question (for short, the SCQ): “When
do several objects jointly compose an object?”

Van Inwagen considers several initially plausible candidate-answers to the SCQ. He argues that each has highly
implausible consequences, viz., commitments to putative entities that are not genuine objects at all according to our
 p p p p p y

usual ways of thinking and talking. (For example, the suggestion that contact among a group of objects is what makes
them jointly compose an object entails the grossly counterintuitive result that when two people shake hands, a new

end p.151

compound object comes into existence that ceases to exist when their hands separate.)

Van Inwagen argues, by elimination-via-counterexample of various initially plausible potential answers to the SCQ,
that the only acceptable answer is that several objects compose an object when they jointly constitute a life. On this
basis, he concludes that the right ontology of physical objects includes only two kinds of material beings: (1)
“simples,” whatever these might turn out to be (e.g., electrons and quarks, perhaps); and (2) living organisms.
Discourse that posits other kinds of concrete objects, he says, should be understood by analogy with talk about the
motion of the sun through the sky: useful and informative, but not literally true.

Two important theoretical desiderata are in play, in van Inwagen's discussion of the ontology of material beings: (1)
finding a systematic, general, answer to the SCQ; and (2) adopting an ontology that conforms reasonably well to our
pre-theoretic common-sense beliefs, and our scientifically informed beliefs, about what kinds of physical objects there
are. Van Inwagen argues very persuasively that these desiderata are deeply in tension: they cannot both be satisfied.
But he also assumes, without explicit argument, that insofar as the two desiderata conflict, satisfying (1) is
theoretically more important than satisfying (2).

Now, it might well be asked whether one should attach more theoretical importance to obtaining a general and
systematic answer to the SCQ than to “saving” tables, chairs, and other objects that we pre-theoretically consider
robustly, mind-independently, real. If so, why? We submit that the answer to the first question is affirmative, and
emerges from common sense itself when common sense goes reflective about metaphysics. Here is what common
sense realizes: an adequate metaphysical theory, like an adequate scientific theory, should be systematic and general,
and should keep to a minimum the unexplained facts that it posits. In particular, a good metaphysical or scientific
theory should avoid positing a plethora of quite specific, disconnected, sui generis, compositional facts. Such facts
would be ontological surds; they would be metaphysically queer. Even though explanation presumably must bottom
out somewhere, it should bottom out with the kinds of “unexplained explainers” we expect to find in physics—viz.,
highly general, highly systematic, theoretical laws. It is just not credible—or even intelligible—that explanation would
bottom out with specific compositional facts which themselves are utterly unexplainable and which do not conform to
any systematic general principles. Rather, if one bunch of physical simples compose a genuine physical object, but
another bunch of simples do not compose any genuine object, then there must be some reason why; it couldn't be
that these two facts are themselves at the explanatory bedrock of being.

There cannot, then, be a body of specific compositional facts that are collectively disconnected and unsystematic, and
are individually unexplainable. Such ontological arbitrariness is not possible in the world-itself—the world whose

end p.152

constituents are van Inwagen's concern. In Horgan (1993) this is called the principle of the non-arbitrariness of
composition. This principle is fundamental and highly plausible, and is a very compelling general requirement on
theory construction. It generates the requirement that an adequate metaphysics of concrete particulars be one for
which there is a general and systematic answer to the SCQ. This requirement has very strong weight in metaphysical
theory-construction, enough to trump the desideratum of preserving the posits of common sense and science.
Common sense viscerally appreciates this—even though common sense will also balk if someone boldly and overtly
affirms some austere ontological claim, such as van Inwagen's claim that the only real physical beings are physical
simples and living organisms.

So the upshot so far is that common sense's appreciation of the need to provide a systematic and general answer to
the SCQ, together with the great difficulty of providing such an answer that also pretty much includes all and only the
kinds of physical objects that are posited in ordinary discourse and in scientific discourse, generates a strong
common-sense basis for holding that the right ontology of concrete particulars will have to be one that posits either
many fewer, or else many more, kinds of concrete particulars than those that are usually posited in science and in
common sense. One candidate ontology that provides a systematic and general answer to the SCQ is van Inwagen's
10
own, comprising only physical simples on the one hand, and living organisms on the other.
10 A couple of things can be said about van Inwagen's answer. First, he embraces an austere ontology, as we characterize it early on in this paper.

Second, although it does indeed yield a systematic, general answer to the SCQ, we think it runs afoul of what is said about vagueness in the next

subsection. It is a vague matter which composites of simples constitute a life, and which do not.

5.3. The Benign Logical Incoherence of Vagueness

Elsewhere (e.g., Horgan 1998 and other papers of Horgan's cited therein, Horgan 2002b), it is argued at some length
that vagueness is logically incoherent in a certain specific way; that this kind of logical incoherence is benign in that it
can be quarantined in language and thought in order to avoid problems like becoming committed to rampant
contradictions; but that the logical incoherence inherent to vagueness renders ontological vagueness—i.e., vague
11
OBJECTS and/or vague PROPERTIES—impossible. Here we will summarize the key line of thought very briefly.
11 We owe the following compact way of putting things to a very helpful conversation with Keith DeRose.
 p p p p p y

Essential to vagueness are the following features. First, sufficiently small differences don't make a difference. They
don't make a difference, for instance, with respect to the applicability of a vague category (e.g., “bald,” or “part of
Mount Everest”) to items that differ from one another in sufficiently small ways (e.g.,

end p.153

two people one of whom has one more hair on his head than the other, or two pebbles on the ground one of which is
three inches further from the peak of Mount Everest than the other). Second, sufficiently large differences do make a
difference. Vague categories do discriminate among cases that are sufficiently different from one another—for
instance, between bald folks (who have little or no hair on their heads) and non-bald folks (who have plenty), or
between pebbles that are part of Mount Everest and far-away pebbles that are not. But third, iterated combinations of
sufficiently small differences add up to sufficiently large differences—as illustrated by the kinds of small-difference
iterations that generate instances of the sorites paradox. When common sense reflects clear-headedly on these three
principles, common sense realizes: (i) that all three principles are indeed essential to genuine vagueness; (ii) that the
three principles jointly embody a form of logical incoherence; and hence (iii) that no genuine OBJECT or PROPERTY could
possibly be vague. (Epistemicists reject claim (i) and maintain instead that vague categories always delimit perfectly
sharp, albeit unknowable, boundaries—a view wildly contrary to common sense.) Common sense also appreciates, at
least viscerally, that people nonetheless get by well enough with vagueness in thought-content and in language,
provided that they are suitably judicious in employing vague concepts and terms—which includes systematic
avoidance of sorites-paradoxical reasoning.

But as common sense also realizes, most of the referential commitments of common sense itself—and numerous of
the referential commitments of scientific theory as well—are to vague items. Concreta like tables, chairs, persons,
stars, and cells, for instance, all are vague in certain ways—e.g., with respect to their temporal boundaries, and with
respect to their composition. Categories like baldness, tallness, redness, and numerous scientific categories are vague
with respect to which concreta do and do not fall under them. So common sense appreciates, at least viscerally, that
the right ontology cannot include such items, and thus must be an austere ontology.

Direct-correspondence semantic standards surely cannot tolerate logical incoherence. When one is thinking/talking
under such standards, sorites arguments become sound reductio ad absurdum arguments against the existence of
vague objects or properties. This fact thus reveals that in most contexts of thought and utterance, the contextually
operative semantic standards are indirect-correspondence standards of one sort or another—since most contexts are
ones in which numerous thoughts and statements employ vague concepts and terms in ways that carry referential
commitments.

Although direct-correspondence semantic standards cannot tolerate the kind of logical incoherence endemic to
vagueness, we contend that indirect-correspondence standards can do so—and can operate in such a way that the
incoherence gets “quarantined” rather than leading to malevolent consequences (e.g., rampant commitment to
contradictions). There's a lot that needs to be said about such quarantining, and about the workings of indirect-
correspondence semantic

end p.154

normativity that incorporates the specific kind of incoherence that's essential to vagueness. (For instance, such
normativity cannot be understood as yielding, in combination with how the world is, some full and determinate
distribution of semantic statuses over all truth-evaluable statements and thought-contents; for, any such total
distribution, as long as it respects the principle sufficiently large differences make a difference, would end up positing
sharp status-transitions of one sort or another, regardless of how many semantic statuses are in play.) But pursuing
such matters is a task for another occasion.

For present purposes, the upshot of what has been said about vagueness is this. On one hand, common sense itself
can reflectively appreciate the kind of logical incoherence that's the very essence of vagueness—and thus can also
reflectively appreciate that ontological vagueness is a priori impossible. On the other hand, common-sense's—and
science's—rampant referential commitments to ontologically vague objects and properties are entirely compatible with
austere ontologies that repudiate all vague OBJECTS and PROPERTIES—provided that austere ontology is wedded to the
claim that in most contexts of thought and discourse, including most scientific contexts, the contextually operative
semantic standards are indirect-correspondence standards. AIC realism is just such a package-deal
ontological/semantical position.

6. Quasi-Particularist Semantic Normativity

Various questions/demands are apt to be posed to us about indirect-correspondence semantic standards, and also
about the dynamics of intra-contextual determination of—and trans-contextual variation in—such standards. One kind
of question/demand would be this: “What are the general principles that systematize intra-context semantic
normativity of any given kind, and what are the yet more general principles by which specific normativity is
determined within any given context and is altered across any given change in context?”

Why suppose there are, or need to be, such general principles in order for our position to be viable? One reason that
is apt to lie behind the demand for such principles is the idea that without them, semantic normativity simply would
 p p p p p y

not be learnable and masterable by human beings.

Yet another kind of question/demand might be posed in the following way: “You claim to be offering a picture whereby
indirect-correspondence semantic standards relieve thoughts and statements from being ontologically committed to
objects and properties to which they are referentially committed. Well, in order to make good on such a picture, you
need not only (i) to articulate general principles, but also (ii) to articulate these general principles in an austere
vocabulary that eschews any referential commitments of its own to items other than those posited by an austere
ontology. For, unless and until you can deliver such ontologically austere formulations of the relevant general
principles, you will not have really

end p.155

eliminated ontological commitment to the various putative objects and properties that are eschewed by austere
ontologies.”

Those who pose such questions/demands are apt to think that unless and until we provide the answers being asked
for, our position will not really be a full-fledged position so much as a mere skeleton of a position. They are apt to
think that the skeleton needs some meat on its bones before it will be worthy of serious consideration, and before it
can even be assessed for intelligibility.

Also, those who pose these questions/demands might be skeptical that principles of the kind being asked for even
exist; and on that basis, they might doubt that there really is any conceptually stable, viable, position about truth and
ontology of the kind we have been seeking to delineate here.

Our own view (cf. Potr 2000, Horgan 2002b, Potr and Strahovnik 2004) is that (i) the kinds of demands we have
mentioned probably cannot be met, but (ii) they very probably don't need to be met. We doubt very much that
semantic normativity conforms to fully general principles at all—let alone to fully general principles that could be
stated in an ontologically austere vocabulary. We also doubt very much that semantic normativity needs to conform to
such principles in order to be learnable and masterable by humans.

One powerful-looking line of argument for the claim that semantic normativity does not in fact conform to
exceptionless general principles can be extracted from van Inwagen's investigation of the Special Composition
Question (the SCQ). If there are exceptionless general semantic principles that are conformed to by the kind of
semantic normativity normally at work in common-sense thought and discourse, then these principles presumably
ought to generate some kind of exceptionless general answer to the SCQ—an exceptionless general answer that nicely
systematizes people's various common-sense judgments about whether or not, in various specific cases, a bunch of
things jointly compose a whole thing. But the apparent lesson of van Inwagen's careful investigation of the SCQ is
that common-sense judgments about composition do not conform to any such general principle. (That is why one gets
driven to an austere ontology, insofar as one seeks a systematic general answer to the SCQ.) Rather, common-sense
judgments about object composition are something of an unsystematic hodgepodge—which presumably means that
the semantic-normative standards these common-sense judgments reflect are also an unsystematic hodgepodge
—which presumably means, in turn, that as far as object-composition judgments/statements in ordinary contexts are
concerned, semantic normativity does not conform to exceptionless general principles.

The point can be generalized. For thought and discourse in general, there is just no obvious reason why the semantic
normativity that governs it, in ordinary contexts, needs to be fully systematizable via exceptionless general principles.
On the contrary, we think there is a heavy burden of proof upon those who would insist that it must conform to such
principles; and we would argue that that burden cannot be discharged.

end p.156

We will use the expression “quasi-particularist semantic normativity” for forms of semantic normativity that cannot be
fully systematized by exceptionless general principles. The point of the prefix “quasi” is to allow for the possibility that
to some extent the relevant kind of normativity can be partially systematized by non-exceptionless general principles
—principles that are vague enough to allow for exceptions (whether or not these principles have explicit ceteris
paribus clauses in them). Such partially-systematizing general principles might be at work within contextually specific
forms of semantic normativity, and/or they might partially systematize the dynamics of trans-contextual variation in
semantic normativity. A nice example, concerning trans-contextual variation, is Lewis's rule of accommodation for
presupposition, which he formulates this way: “If at time t something is said that requires presupposition P to be
acceptable, and if P is not presupposed just before t, then—ceteris paribus and within certain limits—presupposition P
12
comes into existence at t.”
12 Lewis (1979) treats principles in a more liberal way, as compared to their treatment in Lewis (1996).

Why think that quasi-particularist semantic normativity is not what prevails, in the case of normal human thought and
discourse? Why think that there must be fully exceptionless general principles at work? One reason that is apt to
come to mind—briefly mentioned already—is the thought that without the latter kinds of principles, the relevant sort
of normativity would be too complex and unsystematic and idiosyncratic to be learnable and masterable by humans.
We will call this the learnability argument.

We think there is strong reason to believe that the learnability argument is seriously mistaken. Its premiss is false:
 p p p p p y

learnable/masterable semantic normativity need not conform to exceptionless general principles. And its conclusion is
false too: the kinds of semantic normativity normally at work in human thought and discourse just don't conform to
such principles; rather, the operative kinds of normativity are more subtle, more complex, and more nuanced than
that.

Why do we think so? Let us briefly sketch our reasons. There has been a persistent pattern of failure in computational
cognitive science to generate plausible models of what Jerry Fodor (1983) calls “central” cognitive processes
—processes like (i) the rational generation of new beliefs on the basis of prior background information plus newly
acquired sensory information, and (ii) rational planning. As Fodor has persuasively argued for quite a long time—e.g.,
in the late parts of Fodor (1983) and more recently in Fodor (2001)—and as was argued at some length (in
elaboration of Fodor's argument) by Horgan and Tienson (1996), there is good reason to believe that the problems
with attempts to computationally model such central cognitive problems are in-principle problems. These problems
involve the fact that the relevant kind of information-processing normally needs to be highly holistic in nature,
potentially drawing upon virtually any item of information the cognitive system might possess, and sometimes (e.g.,

end p.157

in making comparative-simplicity assessments of particular hypotheses) drawing upon highly holistic features of large
bodies of information (e.g., large bodies of information relative to which the given hypotheses are effectively being
assessed for relative simplicity).

Horgan and Tienson argue in their book that the problem lies with the assumption that the relevant sort of holistically
information-sensitive processing is computation over representations—that is, the assumption that such processing
conforms to exceptionless rules of symbol-manipulation, rules expressible as a computer program. They also describe
a non-classical framework for cognitive science, drawing upon certain ideas from connectionism and from the
mathematics that goes naturally with it—viz., dynamical systems theory. (This framework retains the idea of a
“language of thought,” although with syntax conceived somewhat differently than within the computational paradigm.)

Within this framework, central cognitive processes like belief-formation normally will be too complex and subtle
(especially in the ways they accommodate holistic aspects of background information) to conform to exceptionless
general rules. Cognitively competent cognitive-state transitions will be only partially systematizable, via certain
inherently exception-ridden generalizations with built-in ceteris paribus clauses—generalizations we call soft laws.
Such soft laws are not—and cannot be refined into—exceptionless algorithmic rules.

Tying this back to various forms of normativity that humans can and do learn and master (epistemic normativity,
moral normativity, semantic normativity): the very same kinds of reasons that tell against the idea that competent
human cognition is computation over mental representations—is a matter of conformity to exceptionless general rules
of symbol manipulation—are also reasons to deny that the competent human deployment of normativity (moral,
epistemic, semantic, etc.) is a matter of conformity to exceptionless general rules. Thus, the kinds of reasons that
favor a non-computational approach to human central cognitive processing also favor a particularistic or quasi-
particularistic approach to the kinds of normativity to which humans are subject—and in particular, to semantic
13
normativity.
13
For the question of intertwining between epistemic and semantic normative standards compare footnote 4.

Let us elaborate a bit. To a very great extent, the same kinds of holistically information-sensitive cognitive processing
that are needed for intelligent belief-formation and intelligent planning are also needed in the intelligent deployment
of concepts and terms under contextually appropriate semantic standards, and in the intelligent (if implicit)
appreciation of which semantic standards are themselves appropriate in various contexts of thought/language one
might find oneself in. For, to a very great extent, semantic appropriateness itself is apt to be particularistically—or
anyway, quasi-particularistically—dependent on holistic elements of the cognitive agent's total situation.

end p.158

Let us elaborate the point still further, and from a somewhat different perspective. As we see it, the fundamental
claim of particularism and quasi-particularism about a given type of normativity is this: supervenience relations
linking the non-normative to the relevant kind of normativity are not systematizable in terms of compact, general,
exceptionless, cognitively surveyable principles. Now, suppose there are such supervenience principles, in the case of
semantic normativity. Well then, those very principles should make possible the paraphrasability of
thoughts/statements about ontologically dubious entities via thoughts/statements about ontologically respectable
entities. For the general supervenience principles should generate, as specific instantiations, compact, cognitively
surveyable, truth conditions for each thought/statement about ontologically dubious entities; and these very truth
conditions would constitute a paraphrase. (The truth condition for a given statement would be the disjunction of the
potential supervenience bases for its truth. And often enough, in context, it might be clear enough which disjunct is
the potentially relevant one, so that this disjunct alone could serve in context as the paraphrase.) But in general, as
we emphasized early in this paper, acceptable paraphrases for thoughts/statements about ontologically problematic
entities just do not seem to be available—and do not need to be available, in order for such thoughts and statements
to be useful to, and semantically masterable by, humans. The supervenience of semantic normativity upon ways the
world might be need not, and very likely does not, conform to short, sweet, exceptionless, general principles. Rather,
such supervenience is very likely particularistic or quasi-particularistic in nature.
 p p p p p y

It is on these grounds that we repudiate the kinds of demands that we mentioned at the outset of this section, and we
embrace instead a quasi-particularist version of contextual semantics. Do there need to be general and exceptionless
semantical rules that entail, as instantiations of these rules, compactly articulable, cognitively surveyable, and
austerely expressible explicit formulations of the contextually operative truth conditions for any given statement that
has referential commitments to entities that are repudiated by an austere ontology? We say, “No, not at all.” What is
needed, rather, is just this: (i) the class of ways the world might be needs to be reasonably cleanly partitioned
(modulo vagueness) into ways that conform with the given statement under the contextually operative normativity,
and ways that fail to conform with the given statement under the contextually operative normativity; and (ii) a
cognitively competent agent adequately good at judging such conformity or non-conformity, to the extent that it is
accurately indicated by the agent's own available evidence. (Recall the intimate intertwining of epistemic and
semantic normativity.)

So on this approach to contextual semantics, the demands we mentioned at the beginning of this section are
illegitimate, as are the presuppositions that motivate them. We admit that we cannot meet those demands, but we
deny that this fact constitutes good reason for rejecting our view, or for claiming that the view is a mere skeleton
position without adequate meat on its bones, or for denying that

end p.159

we have carved out a genuine and distinctive philosophical position about realism and truth.

7. Conclusion
When common sense goes reflective, powerful considerations emerge from common sense itself in support of an
austere version of metaphysical realism. If such a metaphysical position is wedded to the view that truth is typically
indirect correspondence to the world rather than direct correspondence, the resulting package-deal position—austere
indirect-correspondence realism—turns out to accommodate ordinary beliefs surprisingly well. The key to such
accommodation is this: when the contextually operative standards for semantic correctness require only indirect
correspondence to the world rather than direct correspondence, the referential commitments of one's thought and
discourse are not ontological commitments.

Demands to spell out general normative principles that govern the intra-context workings of semantic normativity,
and general principles that govern the dynamics of trans-contextual variation of semantic standards, rest on dubious
presuppositions about both semantic normativity and human cognition. Central cognitive processes appear to be too
complex and subtle to conform to programmable general rules over mental representations, and semantic normativity
14
appears to be too complex and subtle to conform to exceptionless general rules.
14
Thanks to Robert Barnard, David Chalmers, and Joseph Tolliver for very helpful comments and discussion.

References

Barnard, B. and Horgan, T. Forthcoming. ‘Truth as Mediated Correspondence’. The Monist.

Crimmins, M. 1998. ‘Hesperus and Phosphorus: Sense, Pretense, and Reference’. Philosophical Review. 107: 1048.

Field, H. 2001. ‘Correspondence Truth, Disquotational Truth, and Deflationism’. In Lynch, 2001: 483–503.

Fodor, J. 1983. The Modularity of Mind. MIT Press.

Fodor, J. 2001. The Mind Doesn't Work That Way: The Scope and Limits of Computational Psychology. MIT Press.

Graham, G. and Horgan, T. 1994. ‘Southern Fundamentalism and the End of Philosophy’ (with G. Graham).
Philosophical Issues, 5: 219–47. Reprinted in M. DePaul and W. Ramsey (eds), Rethinking Intuition: The Psychology of
Intuition and Its Role in Philosophy, Rowman and Littlefield, 1998.

Horgan, T. 1991. ‘Metaphysical Realism and Psychologistic Semantics’. Erkenntnis, 34 (1991): 297–322.

end p.160

Horgan, T. 1993. ‘On What There Isn't’. Philosophy and Phenomenological Research, 50 (1993): 693–700.

Horgan, T. 1998. ‘The Transvaluationist Conception of Vagueness’. The Monist, 81: 316–33.

Horgan, T. 2001a. ‘Causal Compatibilism and the Exclusion Problem’. Theoria, 16: 95–116.

Horgan, T. 2001b. ‘Contextual Semantics and Metaphysical Realism: Truth as Indirect Correspondence’. In M. Lynch,
2001: 67–95.

Horgan, T. 2002a. ‘Themes in My Philosophical Work’. Grazer Philosophische Studien, 63: 1–26.

Horgan, T. 2002b. ‘Replies to Papers’. Grazer Philosophische Studien, 63: 303–41.

Horgan, T. and Potr , M. 2000. ‘Blobjectivism and Indirect Correspondence’. Facta Philosophica, 2: 249–70.
 p p p p p y

Horgan, T. and Potr , M. 2002. ‘Addressing Questions for Blobjectivism’. Facta Philosophica, 4: 311–22.

Horgan, T. and Tienson, J. 1996. Connectionism and the Philosophy of Psychology. MIT Press.

Horgan, T. and Timmons, M. 2000. ‘Nondescriptivist Cognitivism: Framework for a New Metaethic’. Philosophical
Papers, 29: 121–53.

Horgan, T. and Timmons, M. 2002. ‘Conceptual Relativity and Metaphysical Realism’. Philosophical Issues, 12: 74–96.

Horgan, T. and Timmons, M. In press. ‘Cognitivist Expressivism’. In M. Timmons and T. Horgan (eds), Metaethics after
Moore: New Essays in Metaethics. Oxford University Press.

Horwich, P. 2001. ‘A Defense of Minimalism’. In Lynch, 2001: 559–77.

Lewis, D. 1979. ‘Scorekeeping in a Language Game’. Journal of Philosophical Logic, 8: 339–59. Reprinted in D. Lewis,
Philosophical Papers, vol. 1. Oxford University Press, 1983.

Lewis, D. 1996. ‘Elusive Knowledge’. Australasian Journal of Philosophy, 74: 549–67.

Lynch, M. (ed.). 2001. The Nature of Truth: Classic and Contemporary Perspectives. MIT Press.

Putnam, H. 1981. Reason, Truth and History. Cambridge University Press.

Potr , M. 2000. ‘Justification Having and Morphological Content’. Acta Analytica, 24: 151–74.

Potr , M. 2003. ‘Blobjectivist Monism’. In A. Baechli and K. Petrus (eds), Monism. Ontos Verlag: 125–55.

Potr . M. and Strahovnik, V. 2004. Practical Contexts. Ontos Verlag.

Timmons, M. 1999. Morality without Foundations: A Defense of Ethical Contextualism. Oxford University Press.

van Inwagen, P. 1990. Material Beings. Cornell University Press.

Walton, K. 1990. Mimesis as Make-Believe. Harvard University Press.

Wright, C. 1987. Realism, Meaning and Truth. Basil Blackwell.

Wright, C. 1992. Truth and Objectivity. Harvard University Press.

end p.161
 p p p p p y

8 Context, Vagueness, and Ontology *

This is a slightly amplified version of comments on an earlier version of Horgan and Potr 's
paper, ‘Abundant Truth in an Austere World,’ given at the St Andrews Conference on Truth and
Realism in June 2004. My thanks to Nancy Bauer for comments, and to Terry for the chance to
comment on his work.

Some think there aren't—or, as it is sometimes said, there aren't REALLY —any numbers. Others say there aren't (let us
leave the capitals by the wayside) any fictional objects. Some say there aren't chairs, or courts, or corporations. Some
people have even been known to deny that there are any people. Those who deny these things have a tough row to
hoe. It is, after all, pretty hard to deny that there are some prime numbers that are greater than seven, that there
are more characters in Uncle Vanya than in Waiting for Godot, that there is a corporation with more employees than
there are people living in any one of the Outer Hebrides, or, for that matter, that there are people living in the
Hebrides. And these things entail that there are numbers, characters, corporations, and people.
1
Horgan and Potr (HP, henceforth) don't think many of the things mentioned in the above paragraph exist.
1
In ‘Abundant Truth in an Austere World,’ in this volume, Ch. 7. Subsequent references are indicated parenthetica ly.

Since HP seem at least sympathetic with the idea that there aren't any persons, so far as I can see they deny the existence of anything of which the

nouns of this paper's first paragraph are true or which its names name.

But they acknowledge that the truisms about primes, plays, people and such are truisms. They propose to reconcile
the belief and the acknowledgment by positing a contextual parameter—a hitherto unnoticed aspect of a
conversation's “score”—which shifts between requiring “strict semantic standards” and relatively lax ones. The
idea—reminiscent, of course, of that behind contextualist accounts of knowledge—is that discourse is normally
governed by lax standards (and so it's normally true, for example, that Muck is one of the Hebrides), but when we
wax philosophical, stricter standards prevail. So everything's hunky dory—HP are right that there aren't any rocks;
common sense is right that there are rocks. I take up this aspect of HP's view in section I.

end p.162

Why would anyone reject common-sense objects, saying that there aren't any primes, plays, or people; chairs, courts,
or corporations? HP have, I think, two reasons. First, they hold that there is “benign logical incoherence” in the
vagueness endemic to our thought and talk. Since it can be “quarantined,” it is harmless within talk and thought. But,
HP say, a chair or a person would be something of which “chair” or “person” was true. And the “logical incoherence” of
these terms makes it impossible that they be true of anything—renders it impossible that there be such properties as
the property of being a chair or of being a person. Secondly, they think that the facts of ontology can't be
“disconnected and unsystematic, . . . individually unexplainable.” (Ch. 7, p. 152) If, for example, a certain collection of
rocks in Central Scotland constitutes a Munroe, but the sum of that collection and a pebble a few inches north of it
constitutes no sort of object whatsoever, there must be some explanation of why this is so. But, they think, to accept
common-sense ontology we must accept a gerrymandered, unsystematizable ontology. I take these arguments up in
section II. Section III draws a few methodological morals.

I
In different contexts, HP tell us, different “semantic standards” are operative. A shift in standards changes how a
sentence is to be interpreted; the shift HP posit is one that can move the interpretation of the sentences “a is F” or
“something is F” so that they can be true even if “a” doesn't refer and the predicate “is F” is true of nothing
whatsoever.

As I understand it, the idea is this. Consider the sentence “Boston Scientific Corporation acquired Advanced Bionics
Corporation.” One “standard” for its truth demands objects a and b to which its terms refer, and a set, the extension
of “acquired,” which contains the pair of a and b. I'll refer to cases in which such a “Tarskian” standard is operative as
cases in which a sentence is to be interpreted referentially.

Now entertain the idea that corporations are fictions. Even if we came to think that there are no corporations, we
would distinguish between a sentence like that just mentioned which (according to The Boston Globe) is correct, and
one like “General Motors acquired Verizon,” which is not. Our practices, of talking about corporations, stock transfers
and the like, establish conditions under which it is correct to say that a corporate acquisition has taken place. HP's
idea is that sometimes we do not require a referential interpretation for a sentence's truth, but only that its use be
correct relative to our practices. They think that we shift back and forth when we use sentences, between requiring
that they be true when interpreted referentially, and requiring only that “ontologically

end p.163

lighter” conditions be in place. When this “lighter” standard is operative, the truth of “Boston Scientific acquired
Advanced Bionics” doesn't require that its terms refer to anything, or that its verb even have an extension. I'll refer to
cases in which this standard is in place as cases in which a sentence is to be interpreted areferentially.
 p p p p p y

It's hard to evaluate such a proposal before hearing more about correctness conditions and how they are determined.
HP don't offer much elaboration, and it's not hard to see why. If we eschew quantifying over corporations and things
to which they are related, it is very hard to say exactly what would make “Boston Scientific acquired Advanced
Bionics” correct. One might try saying that it is correct if there are legally binding arrangements, of the sort we in fact
describe as forming a corporation, that allowed people to transfer property and such under the names “Boston
Scientific” and “Advanced Bionics,” and that a legal transaction of a familiar sort—one we would describe as a
sale—occurred. Of course, that involves all kinds of quantification and predication (“transferring property,” for
example), that isn't kosher on HP's view. And it's not clear that it's essential to Boston Scientific Corporation that it be
named “Boston Scientific.” And it might even be that the legal arrangements which could give birth to a corporation
are too open-ended to allow for finite statement. Of course the same sort of thing is true of claims about plays and
people, courts and chairs.

Let's not worry about the effability of areferential interpretation. A more pressing worry is about the coherence of HP's
picture, especially when it is coupled with the denial that most of what we normally take ourselves to be talking about
exists. Suppose, in order to illustrate the problem, that we accept the idea that standards for interpretation are part of
the “conversational score,” shifting us back and forth between referential and areferential interpretation of our talk.
Suppose further we hold that the only objects are fundamental particles like quarks, and that the only genuine
2
properties and relations are ones which hold amongst these.
2
HP are tempted by the view that even these don't exist, the only thing there is being the single, partless whole which is the universe. Substituting this

view for the view that all there is is quarks and the void, won't help with the problem under discussion.

Is there some sense in which the sentence “quarks exist” is true, but “chairs exist” is false?

One might think the response is “Yes”:

(A) The sentence “Quarks exist” is true when interpreted referentially; “chairs exist” is not true when
interpreted referentially.

The problem with this response is that we have to decide how it is to be understood: are we supposed to interpret it
referentially or areferentially? A referential interpretation doesn't look very promising, since we have to construct
extensions for “sentence,” “is true,” and “interpreted areferentially” from “fundamental stuff.” If we could do that,
though, wouldn't there be more than just quarks and

end p.164

3
their properties and relations? Wouldn't there be sentences and all sorts of other good stuff?
3
Perhaps I should emphasize that the problem is at the level of the concrete, not the abstract. Forget sentence types. Why would one think that sentence

tokens, but not chairs, rocks, or trees, were genuine?

An areferential interpretation doesn't look any more promising. The whole point of invoking the notion of an
areferential interpretation was to be able to say that banal claims such as:

(B) There are chairs.

“chairs” is the word we use for chairs (i.e., “chairs” is true of chairs)

are true. But a sentence of the form “Ns exist” has a true referential interpretation just in case there are some things
of which “N” is true. If we interpret (A) areferentially, the way we understand (A) is a way of understanding on which
it's true that there are some things of which the word “chairs” is true. So, interpreted areferentially, “chairs exist” is
true. So (A), so interpreted, isn't true, as its second half is false. So we still haven't found a way to say that it's true
that there are quarks, but not that there are chairs.

HP's view requires that there be two senses of “referential interpretation.” If one's ontology is sparse—as theirs
is—then one is in no position to say, when one is speaking strictly, that sentences have any interpretations, much less
referential ones. So referential interpretations—when we understand talk about “referential interpretations”
referentially—can't distinguish between sentences about quarks and sentences about chairs. But if we are speaking
with lax standards when we speak of referential interpretations, then sentences like “chairs exist” do have referential
interpretations, and so we still haven't drawn the distinction, between saying that there are quarks and saying that
there are chairs, that we were trying to draw.

Perhaps HP would respond that when they speak of interpreting a sentence strictly—when they speak of what I am
calling referential interpretations—they mean REFERENTIAL INTERPRETATIONS. That is, they mean real, genuine,
referential interpretations, in which real, genuine objects (or sets thereof) are assigned to words. And, they might
insist, if the only things which (“really”) exist are quarks then “quarks exist” has a referential interpretation, but
“chairs exist” doesn't. Such a response, I think, makes no sense, if their view is correct. Notice that we have no
choice, if HP's view is correct, but to interpret the first three sentences of this paragraph by “using lax
standards”—that is, areferentially. If we try to interpret those sentences in any other way, they are at best
false—none of the terms or predicates used in making the response (save “quarks”) refer (interpreted referentially) to
or are true of anything. Given their ontology and their views about interpretation, what they say makes no sense—
unless we interpret it in such a way that, so interpreting, it is true to say that not only are there quarks (not only

end p.165
 p p p p p y

is “quarks” true of objects), but there are chairs, corporations, etc., etc. (the relevant terms are true of objects).

The force of this objection depends on how much of ordinary ontology is being denied when we couple that denial with
the idea of a bifurcation of interpretations. If someone didn't have a beef with chairs, publishers, or novels, but had
ontological worries about fictional characters, I think he could draw the sort of distinction HP want to draw without
running afoul of this sort of objection. One might be able to eschew universals and mathematical objects, say that
ordinary concrete objects are more or less all genuine, and invoke the distinction HP have in mind to explain how it
could be true that there are infinitely many primes greater than 10. But once your ontology gets rid of ordinary
objects, you've pretty much given up the idea that there are REALLY such things as sentences and their interpretations.
4
And then it becomes REALLY hard to draw the sort of distinction HP want to draw.
4 Sorry . . . I couldn't resist.

HP want the idea that there is a contextual referential/areferential interpretation parameter to do explanatory work: it
is supposed to help explain why we find counterintuitive claims like the claim that there aren't chairs. How can it do
this? Well, assuming that something along the lines of Lewis's “scorekeeping” view of contextual sensitivity is correct,
HP offer a hypothesis about how we “process” a conversation. The critical pieces of this hypothesis are that:

(1) “concepts and terms . . . are governed by various implicit contextual parameters” (p. 148); with a shift in
the values of these parameters, the terms' semantic values shift.
(2) competent speakers “typically deal with these parameters so smoothly . . . that [they] don't even notice
them” (p. 148). “We accommodate to score-changes without necessarily realizing that we are doing so, and
without necessarily noticing that contextual parameters are in play at all.” (p. 148)
(3) “Normally when one employs certain categories, and certain terms expressing them, there is a very strong
default presumption in play that the contextually operative semantic standards are ones under which certain
affirmative . . . thoughts/statements employing those categories will be true (i.e., semantically correct)” (pp.
148–9). There's a “default presumption” when we use everyday words like “chair” and “corporation” that the
“conversational score” requires areferential interpretation—and thus humdrum sentences like “there are
chairs” and “there are corporations” are true. HP describe this presumption as a tendency for speakers to not
accommodate remarks made under stricter standards. (p. 149)

HP infer from all this that if one rejects the claim that there are no chairs, one makes an understandable mistake
which “emanate[s] . . . naturally from one's . . . semantic/conceptual competence, including one's competence at
normally handling implicit contextual parameters so smoothly and naturally that one typically doesn't even notice that
they are there” (p. 150).

end p.166

This story just doesn't ring true. The picture seems to be that we are cognizant at some level of shifts in the
contextual score. It is, after all, difficult to know what to make of the idea that we “smoothly deal with” or “handle”
such shifts, unless what is meant is that we typically (a) register—albeit not at a conscious level—such shifts, and (b)
incorporate such recognition into our “processing” of utterances and consequent evaluation thereof as apt or unapt,
true or false, wise or meshuggenah.

But then—given that we normally do manage to cognize a shift in the contextual score, from interpret utterances
referentially to interpret utterances areferentially, we will be aware—aware in a way which we should be able to bring
to consciousness—that in some significant sense it is always an option to interpret an utterance areferentially. For if
we do normally manage to shift the score to interpret areferentially or to interpret referentially when context
demands, we are sometimes primed to evaluate a sentence for truth without requiring that its terms have referents,
its (non-negated) predicates have non-null extensions. If we sometimes do this, surely we can become aware that it's
possible.

If we were sensitive to the possibility of shifting a referential/areferential parameter for interpreting a sentence, we
would be sensitive to the possibility of sentences of the form “there are Fs” being true in a situation in which there
aren't (“REALLY ”) any Fs. Somehow I doubt that speakers are generally aware of any such possibility. When there
really is a shiftable standard governing the use of word, we seem to recognize that this is so when prompted. It
doesn't take very much reflection to convince one's self that there are situations that don't differ, so far as the
topological properties of Farmer Brown's field go, but do differ as to whether an utterance of “Farmer Brown's field is
flat” is apt. But hardly anyone thinks that there are situations that do not differ with respect to the extension of
“chair,” but do differ with respect to the truth of the sentence “there are chairs.”

Frankly, the story HP are telling seems just a tad nuts. They seem to hold that “deep down” we all recognize that
there aren't “REALLY ” any chairs, but due to some processing errors having to do with conversational score we keep
forgetting this. A more plausible thing to say, I think, is simply that talk about chairs is so entrenched that “it couldn't
possibly be false.” That is: (a) no matter what ontological views we might come to accept, we would continue to speak
pretty much the way we currently do about chairs; (b) if we would persist in speaking a certain way even if our
philosophical theories made us think that the terms and quantifications involved in that way of speaking can't be
taken referentially, the truth conditions of that way of speaking are not referential but areferential. In short: we
“respect common sense” simply by tying truth to use; we don't need to posit a dubious contextual parameter.
 p p p p p y

II
Of course, another way to respect common-sense opinion about what there is is to acquiesce to it without reservation.
HP have two reasons for not doing so:

end p.167

(1) If there were such things as chairs and such, they would be vague, in the sense that it would (for example)
5
be indeterminate whether they had certain physical parts. But such matters can't be vague.
5 I take this to be equivalent to what seems like another reason HP have for rejecting ordinary ontology: that chairs and such don't exist

because there isn't anything even close to a single best candidate with which to identify chairs, courts, or corporations; at best, there are a

number of candidates amongst which there's no non-arbitrary way to choose. This claim, I think, is equivalent to the claim that were there

physical objects, it would be indeterminate with what assemblages of physicals to identity them.

(2) Our ordinary ontology is a pretty miscellaneous lot. There are no general principles which would explain
why its posits are objects, while arbitrary fusions are not. But if there are no such principles, there's little
reason to think that ordinary ontology has a fix on what there is.

Why are we supposed to accept (1)? The argument is this: vagueness suffers from a “benign logical incoherence,” by
6
which HP mean that the semantic principles which govern vague expressions are impossible to satisfy.
6
Strictly speaking: the principles are supposed to be impossible to satisfy in a situation in which there are the right sort of objects to construct a sorites

series.

The idea that the incoherence is a matter of the “governing” principles making inconsistent demands on use is found in the response to Williamson

discussed below.

But if there were, say, chairs, our term “chair” would apply to them. So if there were chairs, they would instantiate
the logical incoherence of the word “chair,” and would be literally contradictory.

There's something fishy about the idea that the principles governing the use of words like “heap” are impossible to
satisfy. Suppose there is a collection C of principles which seem to us to be true of the word “heap,” that seem to us to
give or to explicate its meaning. Suppose there is a series of piles of sugar on the floor of a Safeway in Palo Alto which
happen to form a sorites series for the term “heap,” and that when we apply the principles in C to the piles, we derive
a contradiction. What will we do? Will we, looking at the mound of sugar that is slightly taller than John Perry, say
“well, I thought it was a heap, but I guess it's not”? Should we say this?

Of course not! For one thing, among the “principles” which govern the use of “heap” will surely be ones spawned by
application of the term to “good examples” of heaps, as well as ones spawned by the application of “not a heap” to
“good examples” of non-heaps. It would be a fool's errand indeed to try to “formulate” such principles in words—what
we have here is something like what the psychologist calls a prototype and a similarity space—but the rough idea
behind the first sort of “principle” might be something like:

things a lot like [in terms of shape and ratio of average volume of constituents to overall volume] this [point at
a good example] are heaps; the more/less something is like this [in the above respects], the more/less it
deserves to be called a heap; if you increase the ratio of average volume of constituents to overall volume and

end p.168

preserve the size and shape, or increase overall volume while preserving ratio and shape, what you've got is
definitely a heap.

If this “principle” (and cognate ones for non-heaps) come into conflict with other principles like “take away a grain
from a heap; you still have a heap,” we will hedge the latter. And this is what we should do. How could anybody think
that the fact—that our intuitions about heaps, taken as a body, don't sit well together—shows that a pile of sugar the
size of John Perry isn't (“REALLY ”) a heap of sugar?

The meanings of words like “heap” are based in our practices of applying those words to the passing show. At the
center of such practices is the process of “collecting” prototypical examples (and prototypical non-examples) to ground
those applications. If “principles” about the application of such predicates come into conflict with this, it is clear—
definitely clear—that those principles are to be hedged. It may be unclear—it may be indeterminate—how the hedging
is to be done. But that just means that at some point the application of the term may be, well, vague and
indeterminate.

If you think of the semantics of a term as given by a set of principles and you think of these principles as existing
independently of and prior to the use of the term, it perhaps makes sense to think that the semantic rules governing a
term could be inconsistent. If the meaning of a term was dictated by a committee, and the committee had a bad day,
they might have messed up, right? If the rules are etched in the language of thought and evolution took a wrong turn,
we might be badly programmed, eh? But surely the idea—that the norms which govern our talk are independent of, or
prior to, our behavior and our dispositions to evaluate that behavior when it gets us into classificatory trouble—is a
bad idea. I suspect it's one which when stated baldly no one, not even the most gung ho advocate of the language of
thought, will accept. But once we reject this idea, it is very puzzling how a reasonable case could be made, for the
 p p p p p y

idea that the norms governing our language are inconsistent.

HP point to particular principles which they say are “essential” to vague talk, and they say these principles are
impossible to satisfy. In earlier work, Horgan has given a kindred inventory of principles, supposedly central to
vagueness but which make inconsistent demands on speakers. It is plausible that the principles HP point to are truths
about such talk; it does look as if they are inconsistent. Ditto, for Horgan's earlier work. And thus HP might feel that
the above, somewhat abstract, argument is not wholly responsive to what they've said.
7
So let's look at the details. Replying to Timothy Williamson,
7 In ‘Reply to my Critics,’ in a special issue of Grazer Philosophische Studien, 63 (2002), on Horgan's work.

Horgan traces the alleged incoherence of vague concepts to their “boundarylessness”—to the fact that:

end p.169

(3) In a sorites series for a vague concept, there will not be a determinate transition between where the
concept (definitely) applies and where it (definitely) does not apply.

According to Horgan (3) requires, in the case of the concept of baldness, that:

(4) Individuals differing by only a hair always have the same “status”: either both are determinately bald, or
neither is, or it's not determinate whether either is bald;
(5) No matter how we might apply the labels “bald,” “not bald,” “indeterminate” to a series running from the
hairy to the hairless, our labeling won't be correct, since any labeling makes the transition between being bald
and not being bald determinate.

(4) is obviously inconsistent with the truism that some people are bald and some are not. Horgan takes (4) and (5) to
8
inconsistent simpliciter.
8
A quibble: I am not sure I see that these must be inconsistent, It depends on what choices we make about the “logic of vagueness” (can a disjunction

be determinately true without any of its disjuncts being so?) and the treatment of the operator “determinately” (is “determinately” redundant, so that p

and determinately p are equivalent?).

But why think (3) necessitates (4) or (5)? I haven't found an explicit argument for this. I can imagine someone
arguing as follows. Suppose:

(6) X has n hairs, Y has n+1 hairs, X is determinately bald, but it is indeterminate whether Y is bald.

Then it must be determinate where the concept of baldness ceases to apply: it ceases to apply at n+1 hairs. After all,
the guy with n hairs is bald, and the other guy isn't. So (6) can't be true, since the concept and the word expressing
the concept bald are boundaryless. The same considerations rule out it being indeterminate whether X is bald when Y
is not bald; ditto for the idea that X is bald and Y is not. Indeed, this sort of consideration rules out any adjacent pair
with different statuses. So (4) is implied by (3). And (5) follows in much the same way, given that we have to label
the hairiest guy not bald and hairless guy not bald: boundarylessness forbids a sharp transition, so any labeling will
contain at least one mistake.

The critical move in this argument comes right at the beginning, when we infer from (6) that it is determinate at what
point in a sorites series the concept bald ceases to apply. The inference is fallacious. If it's not determinate whether Y
is bald, then it's not determinate whether or not the concept bald ceases to apply at n+1 hairs. To say that X is bald,
but it's not determinate whether Y is bald, is not to announce or imply that there is sharp transition between X and Y.
It's to indicate that there is not such a transition, since—it being indeterminate whether Y is bald—there's just no
saying whether Y is bald.

It is not determinate when a vague concept ceases to apply in a sorites series. For such a concept there's no sharp
transition in its application—that is, there's no point in the series where the concept determinately applies, followed
immediately by a point where it determinately doesn't. But to say that there's no sharp

end p.170

transition in application is not to say that there is no transition whatsoever! In particular, it's not to say that it
mightn't at some point cease to be determinate whether the concept applies.

I say that the following can be true: Y has one more hair than X; X is bald; Y isn't bald, for it's not determinate that Y
is bald. One might respond by asking about my claims, that Y is not bald and that it's not determinate whether “bald”
applies to Y. Are they supposed to be true? If so, it looks like I'm committed after all to a sharp boundary between the
status of X and Y. But if the claims aren't true—well, if they're not true, I shouldn't be making them, should I? This
line of response presupposes that the only thing I could be doing, when I say that Y is not bald or not determinately
bald, is making an assertion. But surely that's not so. I could, instead, be denying that Y is bald, where to deny a claim
is to do something that's appropriate provided the claim is not true—i.e., provided it's false or indeterminate. Denial
so understood is sui generis, in the sense that it can't be identified with the assertion of anything. Surely there is such
an act, and surely we perform it quite often—as when, after reviewing a paradoxical sentence, we say that we are
quite certain that it's not, that it couldn't be true.

Vagueness does indeed appear incoherent, if we suppose that we can only make assertions using vague
 p p p p p y

language—that we can only commit to the truth of claims made with such language. So much the worse, it seems to
9
me, for the idea that all saying is asserting.
9
Some further development of the ideas in these paragraphs is given in my ‘On an Argument of Williamson's,’ Analysis, 60/2: 213–17.

Suppose Horgan insists that there is some sort of a boundary in the case under consideration: after all, the guy with n hairs is bald; the guy with n+1 is

not. We should agree, and assert that X is bald and deny that Y is. But this just isn't the sort of sharp boundary that is inconsistent with vagueness. To

assert that X is bald and deny that Y is bald isn't to say that if we add a hair to X's head, he'll cease to be bald; it is to indicate that if we add a hair, there is

no saying whether or not X is bald.

HP suggest that the “incoherence” of vague language results from three facts about vague predicates: sufficiently
small differences don't make a difference; sufficiently large differences do make a difference; iterated combinations of
sufficiently small differences add up to sufficiently large differences (pp. 153–4). They say that the claims on this list
are “essential” to vagueness, and that they “jointly embody a form of logical incoherence.” This argument is, of
course, related to the argument just discussed. But it merits some additional comments.

If we interpret HP's talk of making a difference in terms of whether vague predicates are true or false of objects, it
appears that the three principles are indeed inconsistent, since the first then seems to entail the truth of sorites
principles such as “if a pile of n+1 grains of sugar is a heap, so is a pile of n grains of sugar.” The power of HP's
argument lies in the fact that (instances) of the principles (ones about small and large differences in number of hairs,
or grains, or hue) do indeed seem to be central truths about predicates such as “bald,” “heap,” or “henna colored.”
What is missing from the argument is some reason to think

end p.171

that the only, or even the best, way to understand the truth about vagueness the first principle captures is in terms of
a requirement that sorites principles for such vague predicates are true.

Why not, following Diana Raffman and Delia Graff, exploit the context sensitivity of predicates like “bald” to explain
the sense in which it is true that sufficiently small differences (in the number of hairs a person has, compared to a
10
bald person's number of hairs) don't make a difference (as to whether the person is bald)?
10 Diana Raffman, ‘Vagueness without Paradox,’ Philosophical Review XX (1994); Delia Graff, “Shifting Sands: An Interest-Relative Theory of Vagueness,”

Philosophical Topics, 28 (2000).

Raffman and Graff's idea, roughly put, is this. We know that what counts as red, or tall, or bald in one situation may
not count as such in another. And when someone counts as bald, someone with about the same number of hairs will
also count as bald, if the two are compared. This means that when we are confronted by people between whom there
is only a small difference in hairiness, that difference won't matter. It also suggests that whenever we consider an
instance of a sorites principle, it will seem true.

Indeed, one might argue along these lines that our considering the principle if someone with 44,661 hairs is bald, so is
someone with 44,662 actually ensures that it is true—relative to the context in which it is considered. But other,
unconsidered instances of the principle will not be true relative to that context; which instances are true and which
are not will shift, from situation to situation. On this view, the truth behind “sufficiently small differences in the
number of hairs a person has (compared to a bald person) don't make a difference as to whether the person is bald,”
is something like sufficiently small differences between the number of hairs a salient bald person has and those had by
11
another person don't make a difference as to whether the other person is bald.
11
This is awfully condensed. It disregards differences between Raffman and Graff, and the details in the text don't correspond exactly to the views of

either. My goal is not to spell out Raffman or Graff's views in gory detail, but simply to say enough to convey that there isn't reason to think that the only

way “small differences don't make a difference” could be essential to vagueness is a way which entails the truth of sorites principles.

I happen to like this account of why we are tempted by sorites principles, but my point in outlining it is not to endorse
it. Rather, I draw attention to it as a plausible way of telling a story on which our (first, relatively unreflective)
intuitions about “bald” turn out to be explicable mistakes, ones driven by (over) applying facets of our competence as
speakers of languages with vague predicates. What I find puzzling about HP's insistence that vague language is
“incoherent” is the fact that they pass directly to this conclusion without first exploring this sort of explanation of our
intuitions about “bald” and baldness. What is puzzling is that this is the sort of explanation they offer of our “intuition”
12
that an utterance of “there are rocks” can only express a banal truth.
12
I don't, of course, mean that the “mechanisms” or aspects of context sensitivity that Raffman and Graff invoke are the same as those to which HP

appeal. I mean, more simply, that each looks for an explanation of our intuitions in contextual factors that we tend not to notice until they are brought to

our attention.

Given the availability of this sort of story in the case of vague talk, and the fact that they think such stories can

end p.172

bear explanatory weight, why do they think it better to convict our language of incoherence than to say that, in
assenting to a sorites principle, we forget that its instances may say different things in different contexts?

We were led to consider the idea that vague talk is incoherent because of its role in one of HP's arguments that
ordinary ontology is mistaken. We haven't yet taken up HP's other argument against ordinary ontology, an argument
that proceeds from the premisses that our judgments about what does and doesn't make for an object are something
 p p p p p y

of a miscellaneous mess. They hold this because they think there is not a “systematic, general” set of principles which
(even roughly) validate common-sense judgments about what sorts of (material) objects there are and are not. They
take the non-existence of such principles as good evidence that everyday objects are phony. This because:

An adequate metaphysical theory . . . should be systematic and general, and should keep to a minimum the
unexplained facts that it posits. . . . if one bunch of physical simples compose a genuine physical object, but
another bunch of simples do not . . . then there must be some reason why.

Suppose we agree with both (I) and (II):

(I) the objects we normally quantify over are, from the point of view of ontological systematics, something of a
miscellaneous mess.
(II) what there is isn't a miscellaneous mess, as there must be general principles which determine exactly when some
objects jointly form another object.
It just doesn't follow that the objects we normally quantify over don't exist. Surely a much more plausible way of
making sense of (I) and (II) is to observe that most of the time we simply ignore—perhaps because of
ignorance—most of the objects over which we could quantify. Certainly there's good reason for ignoring such objects.
Outside of the metaphysics class, there's just no reason to talk or think about, for example, arbitrary fusions—they
serve no explanatory purpose, they don't have causal powers which call them to our attention in the way that the
causal powers of chairs call chairs to our attention, and so on.

Since we have good reason to think that the domains of our quantifiers are exquisitely sensitive to context and
interest, our quantifiers may well usually fail to range over a lot of what there is. So (I) could well be true. And so
could (II), if arbitrary fusions are objects. I don't quite see the argument for thinking that denying the existence of the
chair I'm sitting in is a better way of reconciling (I) and (II) than saying that common sense, having not grown up with
the idea that it just doesn't take much for some things to make up an object, underestimates the range of what there
is.

end p.173

III
In some ways, the most interesting questions raised by HP's paper are methodological. They hold an extremely
counterintuitive position. They posit an unnoticed contextual sensitivity, and use it to explain why we have intuitions
that run counter to their view. I have wondered about the evidence for the posited sensitivity, but I must say that the
whole enterprise of “respecting common sense” by finding some unexpected truth in its utterances about chairs is a
bit puzzling. At the end of the day, nobody is going be assuaged by being told that there is a way of understanding
talk about chairs on which it comes out true . . . but there aren't REALLY any chairs. People think there really are
chairs. (Really—just ask them.) That is the common-sense opinion. If you got a non-philosopher to grasp the
difference between a referential and an areferential interpretation of “there are chairs,” and then asked him what he
meant when he said “there are chairs,” he would, I think, opt for the referential.

If this is right, how is common sense being vindicated or respected by positing a semantic shift it fails to take
cogniscence of? Sometimes the way to show respect for someone you think is wrong is just to say: you're wrong, and
here's why. In the present context, doing that would not involve doing philosophy of language, but giving an
argument in metaphysics, one of the sort considered at the end of the last section.

It is a worthwhile enterprise, in my opinion, to try to get a handle on how contextual and pragmatic factors affect
what we say and how we interpret one another. But I wonder how much explanatory, or even irenic, work the
philosophy of language can do in metaphysics.

end p.174
 p p p p p y

end p.176

9 Must Do Better

Imagine a philosophy conference in Presocratic Greece. The hot question is: what are things made of? Followers of
Thales say that everything is made of water, followers of Anaximenes that everything is made of air, and followers of
Heraclitus that everything is made of fire. Nobody is quite clear what these claims mean, and some question whether
the founders of the respective schools ever made them. But amongst the groupies there is a buzz about all the recent
exciting progress. The mockers and doubters make plenty of noise too. They point out that no resolution of the
dispute between the schools is in sight. They diagnose Thales, Anaximenes and Heraclitus as suffering from a
tendency to over-generalize. We can intelligibly ask what bread is made of, or what houses are made of, but to ask
what things in general are made of is senseless, some suggest, because the question is posed without any conception
of how to verify an answer; language has gone on holiday. Paleo-pragmatists invite everyone to relax, forget their
futile pseudo-inquiries and do something useful instead.

The mockers and doubters had it easy, but we know now that in at least one important respect they were wrong. With
however much confusion, Thales and the rest were asking one of the best questions ever to have been asked, a
question that has painfully led to much of modern science. To have abandoned it two-and-a-half thousand years ago
on grounds of its conceptual incoherence or whatever would have been a feeble and unnecessary surrender to
despair, philistinism, cowardice or indolence. Nevertheless, it is equally clear that the methods of investigation used
by the Presocratics were utterly inadequate to their ambitions. If an intellectual tradition applied just those methods
to those questions for two-and-a-half millennia, which is far from unimaginable, it might well be very little the wiser
at the end. Much of the progress made since the Presocratics consists in the development of good methods for
bringing evidence to bear on questions that, when first asked, appear hopelessly elusive or naive. Typically, of course,
making progress also involves refining and clarifying the initial question: but the relevant refinements and
clarifications cannot all be foreseen at the beginning; they emerge in the process of attempting to answer the original
rough question, and would not emerge otherwise.

end p.177

The Presocratics were forerunners of both modern philosophy and modern natural science; they did not distinguish
natural science from philosophy. For positivists, the moral of the story is that natural science had to be separated
from philosophy, and marked out as the field for observation, measurement and experiment, before it could make
serious progress. There is doubtless something right about that moral, although as it stands it hardly does justice to
the significance of less empirical methods in natural science, such as the use of mathematics and of thought
experiments, for example by Galileo and Einstein. Moreover, the positivist moral misses a deeper methodological
point. The case of the Presocratics shows that one cannot always tell in advance which questions it will be fruitful to
pursue. Even if a community starts with no remotely adequate idea of how to go about answering a question, it does
not follow that the question is meaningless or not worth addressing. That goes for the questions that we now classify
as philosophical as much as it does for those that we now classify as empirical or natural-scientific.

The opponents of systematic philosophical theorizing might reply that they are not judging philosophical questions in
advance; they are judging them after two-and-a-half millennia of futile attempts to answer them. Of course, it is an
important issue how similar our philosophical questions are to those of ancient Greece, or even to those of
Enlightenment Europe. Nevertheless, philosophy has been going too long as an intellectual tradition separate from
natural science (although sometimes interacting with it) for the question ‘How much progress has it made?’ to be
simply dismissed as premature.

We should not be too pessimistic about the answer, at least concerning the broad, heterogeneous intellectual tradition
that we conveniently label ‘analytic philosophy’. In many areas of philosophy, we knew much more in 2004 than was
known in 1964; much more was known in 1964 than in 1924; much more was known in 1924 than was known in
1884. As in natural science, something can be collectively known in a community even if it is occasionally denied by
eccentric members of that community. Although fundamental disagreement is conspicuous in most areas of
philosophy, the best theories in a given area were in most cases far better developed in 2004 than the best theories
in that area were in 1964, and so on. Much of the knowledge is fairly specific in content. For example, we know far
more about possibility and necessity than was known before the development of modern modal logic and associated
work in philosophy. It was widely known in 2004 and was not widely known in 1964 that contingency is not equivalent
to a posteriority, and that claims of contingent or temporary identity involve the rejection of standard logical laws. The
principle that every truth is possibly necessary can now be shown to entail that every truth is necessary by a chain of
elementary inferences in a perspicuous notation unavailable to Hegel. We know much about the costs and benefits of
analysing possibility and necessity in terms of possible worlds, even if we do not yet know whether such an analysis is
correct.

end p.178

What about progress on realism and truth? Far more was known in 2004 about truth than was known in 1964, as a
result of technical work by philosophical and mathematical logicians such as Saul Kripke, Solomon Feferman, Anil
Gupta, Vann McGee, Volker Halbach and many others on how close a predicate in a language can come to satisfying a
 p p p p p y

full disquotational schema for that very language without incurring semantic paradoxes. Their results have significant
and complex implications, not yet fully absorbed, for current debates concerning deflationism and minimalism. One
clear lesson is that claims about truth need to be formulated with extreme precision, not out of kneejerk pedantry but
because in practice correct general claims about truth often turn out to differ so subtly from provably incorrect claims
that arguing in impressionistic terms is a hopelessly unreliable method. Unfortunately, much philosophical discussion
of truth is still conducted in a programmatic, vague and technically uninformed spirit whose products inspire little
confidence.

In 1964, Michael Dummett had just opened his campaign to put the debate between realism and anti-realism, as he
conceived it, at the centre of philosophy. The campaign had a strong methodological component. Intractable
metaphysical disputes (for example, about time) were to be resolved by being reduced to questions in the philosophy
of language about the proper form for a semantic theory of the relevant expressions (for example, tense markers).
The realist's semantic theory would identify the meaning of an expression with its contribution to the truth-conditions
of declarative sentences in which it occurred. The anti-realist's semantic theory would identify the meaning with the
expression's contribution to the assertibility-conditions of those sentences. Instead of shouting slogans at each other,
Dummett's realist and anti-realist would busy themselves in developing systematic compositional semantic theories of
the appropriate type, which could then be judged and compared by something like scientific standards. But that is not
what happened.

True, over recent decades truth-conditional semantics for natural languages has developed out of philosophical logic
and the philosophy of language into a flourishing branch of empirical linguistics. Frege already had the fundamental
conception of compositional truth-conditional semantics, in which expressions refer to items in the mostly
non-linguistic world, the reference of a complex expression is a function of the reference of its constituents, and the
reference of a sentence determines its truth-value. But Frege was more concerned to apply that conception to ideal
artificial languages than to messy natural ones. The systematic application of compositional truth-conditional
semantics to natural languages goes back to Richard Montague (under the influence of Carnap) in its intensional form
and has been mediated in linguistics by Barbara Partee and others. In its extensional form, it goes back to Donald
Davidson (under the influence of Tarski) and has been mediated in linguistics by Jim Higginbotham and others.
Needless to say, that crude schema does no justice to the richness of recent work and the variety of contributors to it
(in both departments of philosophy and

end p.179

departments of linguistics), which one can check by looking at any decent handbook of contemporary semantic theory
as a branch of linguistics. Surprisingly, however, most participants in the Dummett-inspired debates between realism
and anti-realism have shown little interest in the success of truth-conditional semantics, judged as a branch of
empirical linguistics. Instead, they have tended to concentrate on Dummett's demand for ‘non-circular’ explanations of
what understanding a sentence with a given truth-condition ‘consists in’, when the speaker cannot verify or falsify
that condition. That demand is motivated more by preconceived philosophical reductionism than by the actual needs of
empirical linguistics. Thus the construction and assessment of specific truth-conditional semantic theories has almost
disappeared from sight in the debate on realism and anti-realism.

As for assertibility-conditional semantics, it began with one more or less working paradigm: Heyting's intuitionistic
account of the compositional semantics of mathematical language in terms of the condition for something to be a proof
of a given sentence. The obvious and crucial challenge was to generalize that account to empirical language: as a first
step, to develop a working assertibility-conditional semantics for a toy model of some small fragment of empirical
language. But that challenge was shirked. Anti-realists preferred to polish their formulations of the grand programme
rather than getting down to the hard and perhaps disappointing task of trying to carry it out in practice. The
suggestion that the programme's almost total lack of empirical success in the semantics of natural languages might
constitute some evidence that it is mistaken in principle would be dismissed as crass.

Some participants in the debate denied any need for anti-realists to develop their own semantic theories of a
distinctive form. For, it was proposed, anti-realists could take over truth-conditional semantic theories by interpreting
‘true’ to mean assertible, verifiable or true on some epistemic conception of truth. But that proposal is quite contrary
to Dummett's original arguments. For they require the key semantic concept in the anti-realistic semantics, the
concept in terms of which the recursive compositional clauses for atomic expressions are stated, to be decidable, in
the sense that the speaker is always in a position to know whether it applies in a given case. That is what allows
anti-realists to claim that, unlike realists, they can give a non-circular account of what understanding a sentence
consists in: a disposition to assert it when and only when its assertibility-condition obtains. But it is supposed to be
common ground between realists and anti-realists that truth is not always decidable. A speaker may understand a
sentence without being in a position either to recognize it as true or to recognize it as not true. I can understand the
sentence ‘There was once life on Mars’, even though I have neither warrant to assert ‘There was once life on Mars’ nor
warrant to assert ‘There was never life on Mars’. The point is particularly clear in the intuitionistic semantics for
mathematical language. The key concept in the compositional semantics is the concept p is a proof of s, which is
decidable on

end p.180

the intuitionistic view because to understand a sentence is to associate it with an effective procedure for recognizing
 p p p p p y

whether any given putative proof is a proof (in some canonical sense) of it. By contrast, what serves as the
intuitionistic concept of truth is not the dyadic concept p is a proof of s nor even the monadic concept s has been
proved but the monadic concept s has a proof or s is provable. According to intuitionists, we understand many
mathematical sentences (such as ‘There are seven consecutive 7s in the decimal expansion of ’) without having a
procedure for recognizing whether they are provable. We understand them because we can recognize of any given
putative proof, once presented to us, whether it is indeed a proof of them. Nor can we replace ‘true’ in a truth-
conditional semantics by ‘has been proved’ (treated as decidable), because that would reduce the semantic clause for
negation (that the negation of a sentence s is true if and only if s is not true) to the claim that the negation of s has
been proved if and only if s has not been proved, which is uncontroversially false whenever s has not yet been
decided.

Dummett's requirement that assertibility be decidable forces assertibility-conditional semantics to take a radically
different form from that of truth-conditional semantics. Anti-realists have simply failed to develop natural language
semantics in that form, or even to provide serious evidence that they could so develop it if they wanted to. They
proceed as if Imre Lakatos had never developed the concept of a degenerating research programme.

Dummett's posing of the issue between realism and anti-realism provides a case study of an occasion when the
philosophical community was offered a new way of gaining theoretical control over notoriously elusive issues, through
the development of systematic semantic theories. The community spurned the opportunity, if that is what it was.
Those who discussed realism and anti-realism on Dummett's terms tended to concentrate on the most programmatic
issues, which they debated with no more clarity or conclusiveness than was to be found in the traditional metaphysical
reasoning that Dummett intended to supersede. The actual success or lack of it in applying the rival semantic
programmes to specific fragments of natural language was largely ignored. Far from serving as a beacon for a new
methodology, the debate between realism and anti-realism has become notorious in the rest of philosophy for its
obscurity, convolution and lack of progress.

Of course, one may reject Dummett's attempted reduction of issues in metaphysics to issues in the philosophy of
language. I have argued elsewhere that not all philosophical questions are really questions about language or
1
thought.
1 T. Williamson, ‘Past the Linguistic Turn?’, in B. Leiter (ed.), The Future for Philosophy (Oxford: Oxford Univers ty Press, forthcoming).

That a question is non-semantic does not, however, imply that semantics imposes no useful constraints on the process
of answering it. To reach philosophical conclusions one must reason, usually in areas where it is very hard to
distinguish valid

end p.181

from invalid reasoning. To make that distinction reliably, one must often attend carefully to the semantic form of the
premisses, the conclusion and the intermediate steps. That requires implicit semantic beliefs about the crucial words
and constructions. Sometimes, those beliefs must be tested by explicit semantic theorizing. Philosophers who refuse to
bother about semantics, on the grounds that they want to study the non-linguistic world, not our talk about that
world, resemble astronomers who refuse to bother about the theory of telescopes, on the grounds that they want to
study the stars, not our observation of them. Such an attitude may be good enough for amateurs; applied to more
advanced inquiries, it produces crude errors. Those metaphysicians who ignore language in order not to project it on
to the world are the very ones most likely to fall into just that fallacy, because the validity of their reasoning depends
on unexamined assumptions about the structure of the language in which they reason.

Explicit compositional semantic theories for reasonable fragments of particular natural languages also have the great
methodological advantage of being comparatively easy to test in comparatively uncontentious ways, because they
make specific predictions about the truth-conditions (or assertibility-conditions) of infinitely many ordinary
unphilosophical sentences. The attempt to provide a semantic theory that coheres with a given metaphysical claim can
therefore constitute a searching test of the latter claim, even though semantics and metaphysics have different
objects.

Discipline from semantics is only one kind of philosophical discipline. It is insufficient by itself for the conduct of a
philosophical inquiry, and may sometimes fail to be useful, when the semantic forms of the relevant linguistic
constructions are simple and obvious. But when philosophy is not disciplined by semantics, it must be disciplined by
something else: syntax, logic, common sense, imaginary examples, the findings of other disciplines (mathematics,
physics, biology, psychology, history, etc.) or the aesthetic evaluation of theories (elegance, simplicity, etc.). Indeed,
philosophy subject to only one of those disciplines is liable to become severely distorted: several are needed
simultaneously. To be ‘disciplined’ by X here is not simply to pay lip-service to X; it is to make a systematic conscious
effort to conform to the deliverances of X, where such conformity is at least somewhat easier to recognize than is the
answer to the original philosophical question. Of course, each form of philosophical discipline is itself contested by
some philosophers. But that is no reason to produce work that is not properly disciplined by anything. It may be a
reason to welcome methodological diversity in philosophy: if different groups in philosophy give different relative
weights to various sources of discipline, we can compare the long-run results of the rival ways of working. Tightly
constrained work has the merit that even those who reject the constraints can agree that it demonstrates their
consequences.

Much contemporary analytic philosophy—not least on realism and truth—seems to be written in the tacit hope of
 p p p p p y

discursively muddling through,

end p.182

uncontrolled by any clear methodological constraints. That may be enough for easy questions, if there are any in
philosophy; it is manifestly inadequate for resolving the hard questions with which most philosophers like to engage.
All too often it produces only eddies in academic fashion, without any advance in our understanding of the subject
matter. Although we can make progress in philosophy, we cannot expect to do so when we are not working at the
highest available level of intellectual discipline. That level is not achieved by effortless superiority. It requires a
conscious collective effort.

We who classify ourselves as ‘analytic philosophers’ tend to fall into the assumption that our allegiance automatically
confers on us methodological virtue. According to the crude stereotypes, analytic philosophers use arguments while
‘continental’ philosophers do not. But within the analytic tradition many philosophers use arguments only to the
extent that most ‘continental’ philosophers do: some kind of inferential movement is observable, but it lacks the clear
articulation into premisses and conclusion and the explicitness about the form of the inference that much good
philosophy achieves. Again according to the stereotypes, analytic philosophers write clearly while ‘continental’
philosophers do not. But much work within the analytic tradition is obscure even when it is written in everyday words,
short sentences and a relaxed, open-air spirit, because the structure of its claims is fudged where it really matters.

If the high standards that make philosophy worth doing are often absent even in analytic philosophy, that is not
because they are a natural endowment found only in a brilliant elite. Even if Frege's exceptional clarity and rigour
required innate genius—although they undoubtedly also owed something to the German mathematical tradition within
which he was educated—after his example they can now be effectively taught. Some graduate schools communicate
something like his standards, others notably fail to do so.

Of course, we are often unable to answer an important philosophical question by rigorous argument, or even to
formulate the question clearly. High standards then demand not that we should ignore the question, otherwise little
progress would be made, but that we should be open and explicit about the unclarity of the question and the
inconclusiveness of our attempts to answer it, and our dissatisfaction with both should motivate attempts to improve
our methods. Moreover, it must be sensible for the bulk of our research effort to be concentrated in areas where our
current methods make progress more likely.

We may hope that in the long term philosophy will develop new and more decisive methods to answer its questions,
as unimaginable to us as our methods were to the Presocratics. Indeed, the development of such methods is one of
the central challenges facing systematic philosophy. Paul Grice once wrote: ‘By and large the greatest philosophers
have been the greatest, and the most self-conscious, methodologists; indeed, I am tempted to regard this fact as
primarily

end p.183

2
accounting for their greatness as philosophers.’
2 ‘Reply to Richards,’ in R. Grandy and R. Warner (eds), Philosophical Grounds of Rationality: Intentions, Categories, Ends (Oxford: Clarendon Press,

1986), at p. 66.

Nevertheless, we must assume, in the short term philosophy will have to make do with currently available methods.
But that is no reason to continue doing it in a methodologically unreflective way. A profession of very variable
standards can help the higher to spread at the expense of the lower, by conscious collective attention to best practice.

One might think that methodological consciousness-raising is unnecessary, because on any particular issue good
arguments will tend to drive out bad in the long run. But that is over-optimistic. Very often—especially in debates
between realists and anti-realists—a philosopher profoundly wants one answer rather than another to a philosophical
question to be right, and is therefore predisposed to accept arguments that go in the preferred direction and reject
contrary ones. Where the level of obscurity is high, as it often is in current debates about realism and truth, wishful
thinking may be more powerful than the ability to distinguish good arguments from bad, to the point that convergence
in the evaluation of arguments never occurs.

Consider a dispute between rival theories in natural science. Each theory has its committed defenders, who have
invested much time, energy and emotion in its survival. The theories are not empirically equivalent, but making an
empirical determination between them requires experimental skills of a high order. We may predict that if the
standards of accuracy and conscientiousness in the community are high enough, truth will eventually triumph. But if
the community is slightly more tolerant of sloppiness and rhetorical obfuscation, then each school may be able to
survive indefinitely, claiming empirical vindication and still verbally acknowledging the value of rigour, by protecting
samples from impurities a little less adequately, describing experimental results a little more tendentiously, giving a
little more credit to ad hoc hypotheses, dismissing opposing arguments as question-begging a little more quickly and
so on. Each tradition maintains recruitment by its dominance and prestige in some departments or regions. A small
difference in how carefully standards are applied can make the large difference between eventual convergence and
ultimate divergence.

It seems likely that some parts of contemporary analytic philosophy just pass the methodological threshold for some
cumulative progress to occur, however slowly, while others fall short of the threshold. A reasonable fear is that
 p p p p p y

debates over realism and anti-realism constitute a part that falls short. That is not to condemn every piece of work in
the area individually—which would surely be unfair—but to say that collectively the community of participants has not
held itself responsible to high enough methodological standards. Perhaps these debates raise even more difficult
issues than are encountered elsewhere in philosophy: if so, all the more reason to apply the very highest standards
available. As already noted, that appears not to have happened.

end p.184

How can we do better? We can make a useful start by getting the simple things right. Much even of analytic
philosophy moves too fast in its haste to reach the sexy bits. Details are not given the care they deserve: crucial
claims are vaguely stated; significantly different formulations are treated as though they were equivalent; examples
are under-described; arguments are gestured at rather than properly made; their form is left unexplained; and so on.
A few resultant errors easily multiply to send inquiry in completely the wrong direction. Shoddy work is sometimes
masked by pretentiousness, allusiveness, gnomic concision or winning informality. But often there is no special
disguise: producers and consumers have simply not taken enough trouble to check the details. We need the
unglamorous virtue of patience to read and write philosophy that is as perspicuously structured as the difficulty of the
subject requires, and the austerity to be dissatisfied with appealing prose that does not meet those standards. The
fear of boring oneself or one's readers is a great enemy of truth. Pedantry is a fault on the right side.

Precision is often regarded as a hyper-cautious characteristic. It is importantly the opposite. Vague statements are the
hardest to convict of error. Obscurity is the oracle's self-defence. To be precise is to make it as easy as possible for
others to prove one wrong. That is what requires courage. But the community can lower the cost of precision by
keeping in mind that precise errors often do more than vague truths for scientific progress.

Would it be a good bargain to sacrifice depth for rigour? That bargain is not on offer in philosophy, any more than it is
in mathematics. No doubt, if we aim to be rigorous, we cannot expect to sound like Heraclitus, or even Kant: we have
to sacrifice the stereotype of depth. Still, it is rigour, not its absence, that prevents one from sliding over the deepest
difficulties, in an agonized rhetoric of profundity. Rigour and depth both matter: but while the conscious and
deliberate pursuit of rigour is a good way of achieving it, the conscious and deliberate pursuit of depth (as of
happiness) is far more likely to be self-defeating. Better to concentrate on trying to say something true and leave
depth to look after itself.

Nor are rigour and precision enemies of the imagination, any more than they are in mathematics. Rather, they
increase the demands on the imagination, not least by forcing one to imagine examples with exactly the right
structure to challenge a generalization; cloudiness will not suffice. They make imagination consequential in a way in
which it is not in their absence.

Beyond rigour and precision, mathematics has less obvious values to teach. In particular, a mathematical training
makes one appreciate the importance of the aesthetics of definitions. Experience shows that a mathematician or
logician with no ability to discriminate between fruitful and unfruitful definitions is unlikely to achieve much in
research. Such discriminations involve a sort of aesthetic judgement. The ugly, convoluted, ramshackle definitions of
concepts and theses that philosophers seem to feel no shame in producing—not least in debates between realism and
anti-realism—are of just the kind to strike a mathematician

end p.185

as pointless and sterile. Of course, it is notoriously hard to explain why aesthetic criteria are a good methodological
guide, but it would be dangerously naive to abandon them for that reason.

In addition to the humdrum methodological virtues, we need far more reflectiveness about how philosophical debates
are to be subjected to enough constraints to be worth conducting. For example, Dummett's anti-realism about the past
involved, remarkably, the abandonment of two of the main constraints on much philosophical activity. In rejecting
instances of the law of excluded middle concerning past times, such as ‘Either a mammoth stood on this spot a
hundred thousand years ago or no mammoth stood on this spot a hundred thousand years ago’, the anti-realist
rejected both common sense and classical logic. Neither constraint is methodologically sacrosanct; both can intelligibly
be challenged, even together. But when participants in a debate are allowed to throw out both simultaneously,
methodological alarm bells should ring: it is at least not obvious that enough constraints are left to frame a fruitful
debate. Yet such qualms surfaced remarkably little.

Part of the problem is that it is often left unclear just how extensively a constraint is being challenged. A philosopher
treats the law of excluded middle as if it carried no authority whatsoever but implicitly relies on other logical
principles (perhaps in the meta-language): exactly which principles of logic are supposed to carry authority? A
philosopher treats some common-sense judgement as if it carried no authority whatsoever but implicitly relies on
other judgements that are found pre-philosophically obvious: exactly which such judgements are supposed to carry
authority?

When law and order break down, the result is not freedom or anarchy but the capricious tyranny of petty feuding
warlords. Similarly, the unclarity of constraints in philosophy leads to authoritarianism. Whether an argument is
widely accepted depends not on publicly accessible criteria that we can all apply for ourselves but on the say-so of
charismatic authority figures. Pupils cannot become autonomous from their teachers because they cannot securely
 p p p p p y

learn the standards by which their teachers judge. A modicum of wilful unpredictability in the application of standards
is a good policy for a professor who does not want his students to gain too much independence. Although intellectual
deference is not always a bad thing, the debate on realism and anti-realism has seen far too much of it. We can
reduce it by articulating and clarifying the constraints.

Philosophy can never be reduced to mathematics. But we can often produce mathematical models of fragments of
philosophy and, when we can, we should. No doubt the models usually involve wild idealizations. It is still progress if
we can agree what consequences an idea has in one very simple case. Many ideas in philosophy do not withstand even
that very elementary scrutiny, because the attempt to construct a non-trivial model reveals a hidden structural
incoherence in the idea itself. By the same token, an idea that does not collapse in a toy model

end p.186

has at least something going for it. Once we have an unrealistic model, we can start worrying how to construct less
unrealistic models.

Philosophers who reject the constraints mentioned above can say what constraints they would regard as appropriate.
Of course, those who deny that philosophy is a theoretical discipline at all may reject the very idea of such
constraints. But surely the best way to test the theoretical ambitions of philosophy is to go ahead and try to realize
them in as disciplined a way as possible. If the anti-theorists can argue convincingly that the long-run results do not
constitute progress, that is a far stronger case than is an a priori argument that no such activity could constitute
progress. On the other hand, if they cannot argue convincingly that the long-run results do not constitute progress,
how is their opposition to philosophical theory any better than obscurantism?

Unless names are invidiously named, sermons like this one tend to cause less offence than they should, because
everyone imagines that they are aimed at other people. Those who applaud a methodological platitude usually assume
that they comply with it. I intend no such comfortable reading. To one degree or another, we all fall short not just of
the ideal but of the desirable and quite easily possible. Certainly this paper exhibits hardly any of the virtues that it
recommends, although with luck it may still help a bit to propagate those virtues (do as I say, not as I do). Philosophy
has never been done for an extended period according to standards as high as those that are now already available, if
only the profession will take them seriously to heart. None of us knows how far we can get by applying them
systematically enough for long enough. We can find out only by trying.

In making these comments, it is hard not to feel like the headmaster of a minor public school at speech day, telling
everyone to pull their socks up after a particularly bad term. It is therefore appropriate to end with a misquotation
from Winston Churchill. This is not the end of philosophy. It is not even the beginning of the end. But it is, perhaps,
the end of the beginning.

end p.187
 p p p p p y

10 A World without Isms: Life after Realism, Fictionalism, Non-Cognitivism, Relativism,

Reductionism, Revisionism, and so on *
I would like to thank Simon Blackburn and Paul Boghossian for helpful conversations about these
matters. I am indebted, also, to participants at the St Andrew's conference at which the first draft
of this paper was presented—especially to my commentator, Marian David.

We philosophers tend to be most comfortable with fairly obvious states of medium-sized objects—the cat being on the
mat and such like. To the extent that facts are dissimilar to these, they can seem more-or-less weird and in need of
some sort of demystifying account. Consider, for example, the domains of ethics, arithmetic, modality, and theoretical
physics. In such areas the facts at issue—e.g., whether religion is bad, whether Goldbach's conjecture is true, whether
tables are necessarily tables, whether space-time has 11 dimensions—are ontologically and epistemologically unlike
humdrum physical facts. They lack a certain ‘solidity’; and none of the standard five senses puts us in touch with
them. Thus it becomes puzzling how such facts could really be ‘out there’ and how, even if they are, their identities
could ever be known. And these puzzles appear to call for one or another type of philosophical theory.

One general form of theoretical response is to suppose that the discourse in question does not really require the sort
of weird fact that we might naively take it to require. And one particular way of implementing this strategy is to deny
that propositions in the domain are ever true, thereby relieving the pressure to recognize facts corresponding to
them. Perhaps we have beliefs, make assertions, disagree with each other, argue, give grounds, and so on. But, on
the present view, there is no reality behind all that thought and talk; nothing to make some of these claims right and
others wrong; hence nothing weird.

The trouble with this idea is that the utility of our notion of truth presupposes that just about all propositions be truth-
apt—indeed, that each one be materially equivalent to our attributions of truth to it. The notion is needed to
formulate, amongst other things, laws of logic—e.g., “If a conditional and its antecedent are

end p.188

both true, then so is its consequent”—laws that apply within all the discourses at issue; and it can do this job
1
(together with its other expressive jobs) if, and only if, we accept instances of, “The proposition that p is true p”.
1 See my Truth (2nd edn), Oxford University Press, 1998: 84.

So we are not at liberty to believe and assert, for example, that torture is wrong, and at the same time deny that
2
there is any truth of the matter (or any fact of the matter).
2
One might be tempted by the idea that, whereas all propositions are candidates for being true, not all true propositions are backed by REAL facts. For

critical discussion see Truth: 11–21.

This objection invites a slightly different way of implementing the general weirdness-avoidance strategy. Maybe all
propositions are indeed truth-apt. Still, one might deny that the sentences in some problematic domain articulate
propositions, that they express genuine beliefs, and that they are used to make genuine assertions. Certainly it may
look as if they do these things; for our pronouncements are categorical—they are syntactically just like those that
concern obvious characteristics of medium-sized objects. But it might be maintained that this syntactic appearance is
deceptive; that the sentences in question don't in fact function to communicate descriptions of the world, but serve
some other purpose. The best-known example of this position is emotivism in ethics (a.k.a. non-cognitivism)
—whereby, for example, “torture is wrong” is uttered in order to express a negative feeling rather than a belief.

But, just as the first shot at a ‘weirdness-avoidance theory’ did violence to our notions of ‘truth’ and ‘fact,’ the present
attempt clashes with our ordinary notions of ‘assertion,’ ‘belief,’ and ‘proposition.’ Whatever else we may normally be
doing when we assent to “torture is wrong” we are surely asserting that torture is wrong; and whatever other state of
mind that assertion may typically reveal, it surely, in addition, tends to manifest the conviction that torture is wrong.
As Crispin Wright has observed, all it takes for the sentences within a domain of discourse to express propositions is
that they superficially seem to do so—for example, that they occur as premisses and conclusions of arguments. In
3
such a case, “that” - clauses may be formed with them—a species of singular term. Thus we have propositions.
3
See Wright's Truth and Objectivity, Harvard University Press, 1992: 35–6.

So it cannot reasonably be denied that the sentences within the apparently assertoric domains of discourse with which
we are concerned are genuinely assertoric, express genuine propositions, and are, therefore, genuinely true or false.

A third way of trying to avoid the unpalatable ontological commitments that may seem to be implicit in a given
linguistic practice is, not to deny that its sentences express propositions that are true or false, but to suppose that
they don't express the sorts of proposition that we might naively take them to express and therefore don't really
require for their truth the sorts of weird fact that we naively take them to require. For example, consider relativism.
The relativistic account of a domain of discourse is that the sentences, “p”, that we normally assert within it

end p.189

are elliptical; they don't fully articulate the propositions that they express; what is really meant is always something
 p p p p p y

like, “Relative to such-and-such point of view, p”; the only facts around are the ones that are articulated by these
longer sentences; and those facts would not be especially weird. Thus, just as there are no facts of the form ‘x is to
4
the right of y’, so a relativist about normative matters (like Hartry Field
4 See Field's “Disquotational Truth and Factually Defective Discourse,” Philosophical Review, 103 (1994).

) will say that there are no facts about what ought to be done, but only about what ought to be done from a specified
perspective—i.e., given certain basic norms, or according to a certain community. Similarly, it might be claimed, about
arithmetic, that the only facts in the vicinity concern what obtains relative to certain arithmetical assumptions.

Against the particular form of relativism whereby “p” is regarded as short for “p according to person(s) S” (or short for
“S believes that p”, or for “S's principles entail that p”), one can complain that, since it is thereby acknowledged that
there is such a thing as the proposition that p, then the original, unrelativized assertions will be true or false after
all—which is just what the doctrine was at pains to avoid. In response, our relativist might feel forced to maintain that
“p” remains elliptical even when it occurs within the sentences that further articulate what the original “p” is short for.
But that would land him in a vicious regress, whereby “p” abbreviates “S believes that p” which abbreviates “S
believes that S believes that p”, and so on. And the same regress would threaten relativists (e.g. about arithmetic) for
whom “p” is used to mean “Such-and-such principles entail that p”. Thus it would seem that the most natural ways of
5
formulating philosophical relativisms leave us with the same weird facts that we had before.
5 This is not an objection to all relativistic theories, since not all such accounts take the form, just assumed, in which “p” is held to be short for something

that makes reference to the proposition that p. (Consider, for example, the normal expansion of “x is to the right of y”.) See below, however, for a more

general objection to relativism—one that counts against all non-revisionary relativism accounts.

A proposal that is often taken to differ from relativism, but which is barely distinguishable from it, is fictionalism.
When we say “Holmes lived on Baker Street” we tend to mean, not that there really was such a guy who lived on that
street, but rather that according to the well-known detective stories this was so. Similarly, a fictionalist about possible
worlds says that “Possibly q” is used to mean, not that there really is a possible world in which “q” is true, but rather
that from within the possible worlds account of modality one of them involves the truth of “q”. But—to repeat the
just-mentioned complaint—it is then being conceded that there is such a thing as the proposition, true or false, that
there exists a possible world in which “q” is true. Weird facts have not been dispensed with.

In response to this point, a relativist or fictionalist might well observe that the facts he is concerned to avoid are
merely those that would involve the possession of weird properties or the existence of weird objects; there may be
nothing

end p.190

distasteful about negative facts—e.g., that torture is not wrong (and not right either), and that there aren't any
numbers, so it is not the case that 1+2=3. Thus the relativist/fictionalist regarding a problematic domain can happily
acknowledge that its original sentences express propositions, as long as he adds that all the positive ones are false,
and so don't call for the sort of weird fact that ‘surely’ cannot obtain. This is the sort of approach—a so-called error
theory—recommended by John Mackie for ethical discourse and by Paul Boghossian and David Velleman for colour
6
discourse.
6 See Mackie's Ethics: Inventing Right and Wrong, Penguin, 1977. And see Boghossian and Velleman's “Colour as a Secondary Quality,” Mind, 98/389

(January 1989): 81–103, and their “Physicalist Theories of Color,” Philosophical Review, 100/1 (January 1991): 67–106 .

Does it imply—as one might think—that such practices must be discontinued or modified? Not according to most of its
proponents. They tend to suppose, rather—just like most other forms of relativism and fictionalism—that our patterns
of discourse can proceed just as they did before we came to appreciate their literal erroneousness. The crucial proviso
is that the communicative intension behind any given statement, “p”, no longer be what it was, no longer to manifest
the belief that p.

I will be arguing that this recommended separation between what we say and mean is not possible. But first it's worth
mentioning a different way of motivating that feature of the relativist/factionalist/error-theoretic stance—a way that
involves regarding it as a mitigating response, not to a definite renunciation, on metaphysical grounds, of positive
“p”-facts, but rather to the epistemological difficulties that would arise if, despite their weirdness, they really did exist.
The idea is that the metaphysical oddity of such facts would make them impossible to discover; however, our usual
pronouncements within the discourse—as long as we mean by them what the relativist/fictionalist/error-theorist
recommends—will nonetheless be justifiable and useful. This is Hartry Field's view of arithmetic and Gideon Rosen's
7
view of possible-world talk.
7
See Field's Science Without Numbers, Blackwell, 1980; and Rosen's “Modal Fictionalism”, Mind, 99 (1990): 327–54.

They don't deny that there may, undetectably, happen to exist numbers and possible worlds with various
characteristics; but they deny, given the elliptical readings we give (or should give) to “3 is a prime number”, “There's
a world in which cats bark”, and so on, that the inaccessibility of such facts has any import for the conduct of these
practices.

A similar move—the suggestion that what we say does not straightforwardly capture what we really think—is
sometimes made in order to ameliorate the effects of skepticism with respect to the unobservable entities postulated
in theoretical science. Bas van Fraassen's constructive empiricism is the view that when a physicist assents to a
 p p p p p y

sentence, “p”, concerning quarks and the like, what he believes and means to assert is not the proposition that p, but
rather the weaker proposition that some empirically-adequate total theory of the world entails that

end p.191

8
p.
8
See van Fraassen's The Scientific Image, Oxford University Press, 1980.

This position is similar to most forms of relativism, fictionalism, and error theory, in proposing that what we are
committing ourselves to in our pronouncements is not what one would naively suppose.

These ‘non-literalist’ proposals are typically advanced in a conservative spirit. They are supposed to protect and
9
preserve a given practice as it is from the metaphysical and/or epistemological difficulties that put it in jeopardy.
9 Revisionary forms of relativism, error theory, et al. (which take these doctrines to dictate modifications of our present practice) will be considered

towards the end of this paper.

But one might wonder whether the recommended prophylactic—the suggested divergence between what we think and
what we say—is even so much as possible, let alone desirable. Is it conceivable that our linguistic practice within some
domain proceed exactly as if we were believing the propositions that are fully formulated by the sentences to which
we assent, even though we are not really doing so?

One might be tempted to reply, “Why not? We often don't bother to fully articulate what we are thinking. Discourse
about fiction—e.g., Sherlock Holmes—is a case in point.” However, this answer overlooks a crucial difference between
genuine ellipsis and what is recommended by the relativist, fictionalist, error-theorist, and constructive-empiricist.
Although we affirm, “Holmes lived in turn-of-the-century London”, we wouldn't scour Somerset House for records of
his birth; nor do we feel entitled to infer that our great-grandparents might have run into him. Thus genuine ellipsis
has tell-tale signs: namely, we don't engage in the trains of thought characteristic of believing the proposition fully
expressed by such utterances. But the fictionalist thesis (and the same goes for the other non-literalist positions) is
that our assertoric practice—including our reasoning and explanation giving—need not be altered in any way by the
proposed suspension of belief. Thus the normal marks of ellipsis will not be present.

It is worth stressing that the ‘non-literalist’ thesis is not fundamentally about communication. It is about what is going
on when a person thinks, in the privacy of his own study, about numbers, or modality, or fundamental particles; when
he reasons to himself about these matters, searches for explanations, and so on. The suggestion is that our current
discourse—both internal and externalized—is consistent with there being a systematic divergence between the literal
meanings of the sentences that we ‘think to ourselves’ and the propositions (if any) that we are thereby believing, or
conjecturing, or entertaining. In particular, when we are assenting to “p”, it could well be, they say, that what we are
believing is not the proposition that p, but rather some qualified, hedged version of it. But one might wonder if there
is any plausible account of what it is to have a belief that could be consistent with this suggestion.

Consider the familiar and attractive idea—ventured by Jerry Fodor and Stephen Schiffer—that to believe that p is to
have in one's so-called ‘belief box’ a sentence that expresses the proposition that p. To say that certain sentences

end p.192

are in a person's belief box is to say, picturesquely, that they have a particular status within his cognitive economy, a
status characterized, roughly speaking, by the following principles: (a) that sensory experiences cause associated
sentences to have that status (i.e. to go into the box); (b) that sentences in the box bring it about, in virtue of
inferential rules, that certain other sentences are also there; (c) that some such sentences are relied upon in practical
reasoning—i.e., they engender, in conjunction with the person's desires and via principles of deliberation, certain
decisions and actions; and (d) that when (and only when) a sentence is in the box there is a disposition to utter it.

The fictionalist (et al.) would have to suppose, in effect, that there is another box for sentences—let's call it the fiction
box—and that someone can choose, when convinced by skeptical arguments directed at a certain discourse, to shift its
sentences from his belief box to his fiction box. He will then qualify as ‘accepting’ the propositions they express but,
allegedly, no longer as believing them.

But the problem is that, if there is to be no revision of the practice, any such sentence will now have to go into the
fiction box in whatever circumstances it would have previously gone into the belief box. In addition, the consequences
of a sentence being in the fiction box vis-à-vis inference, utterance, and action will have to be the same as what the
consequences would have been of its being in the belief box. Therefore, even though someone's saying “p” will,
allegedly, no longer express his belief that p, but merely his acceptance of that proposition, this won't have any casual
significance; for the psychological import of the fiction box is just the same as that of the belief box. So there would
appear to be no difference at all between these supposedly different boxes—no difference at all between believing that
10
p and the attitude of accepting that p, which is being urged as a less committal alternative.
10
Anyone adopting one of these non-literalist positions will distinguish himself by saying, “For such-and-such reasons I don't really believe the

propositions (if any) that are fully expressed by my D-sentences. So the utterances, inferences, actions, etc., that standardly indicate such beliefs should

not be taken to do so in my case.” But, in light of the above argument, such declarations should be seen as mistakes fostered by confusion about the

nature of belief.

Note that since one can have beliefs without having the concept of belief, it would be quite implausible to maintain that the constituting conditions for

having a given belief include not denying that one has it.
 p p p p p y

Thus it is an illusion that we might persevere in deploying sentences exactly as if we believed what they literally
express, but without really doing so. In their popular non-revisionary forms, neither relativism, nor fictionalism, nor
11
the error theory, nor constructive empiricism, appears to be a coherent option.
11 This skepticism about there being a genuine distinction between belief and mere acceptance is elaborated in my “On the Nature and Norms of

Theoretical Commitment,” Philosophy of Science, 58/1 (1991), pp. 1–14 (reprinted in my From a Deflationary Point of View, Oxford University

Press, 2004). See also Simon Blackburn's “Realism: Deconstructing the Debate,” Ratio, XV/2 (2002): 111–33.

Does our self-styled anti-realist—still determined to avoid queer facts—have any further room to maneuver? He may
well think that he does. For a common

end p.193

move has been to assume a distinction between, on the one hand, so-called robust facts—facts that are REAL (with
capital letters)—and, on the other hand, merely deflationary facts, facts to which we are committed merely by virtue
of making assertions and accepting the trivial equivalence of “p” and “It's a fact that p.” These deflationary facts are
certainly taken to be real in the ordinary sense of that word (since everything that exists is real, in that sense), but
not REAL (with capital letters), not robust. The point of this distinction is supposed to be that it's not so unpleasant,
metaphysically speaking, to have to swallow weird facts, as long as they are merely deflationary. It's only weird robust
facts that are hard to stomach. Thus a subtle expressivist may say: “OK. I agree that ethical claims fully express
beliefs (as well as desires) and are true or false; but the facts in virtue of which some of them are true are not robust;
so we can acknowledge them without metaphysical over-indulgence.” Similarly, a so-called non-factualist about
12
meaning, such as Kripke's Wittgenstein,
12 See Saul Kripke's Wittgenstein on Rules and Private Language, Blackwell, 1982.

will agree that our practice of meaning-attribution is coherent and useful, and will acknowledge that its
pronouncements (e.g., “Jones means PLUS by ‘+’ ”) may be legitimately re-phrased as claims about the truth or the
facts (e.g., “It's a fact that Jones means PLUS”), but will go on to say that such facts are not robust or really REAL; they
13
are merely deflationary, hence innocuous.
13 Although it is convenient to present realist/anti-realist debates as disputes about the existence of certain sorts of fact, the central issues do not

presuppose a serious ontology of facts. For, in the first place, even if facts are always eschewed, similar puzzles and disputes may be stimulated by the

sense that a total theory involving normative claims, modal claims, etc., would be unattractively heterogeneous and would be impossible to verify in

certain parts. And, in the second place, more-or-less the same debates arise with respect to ontological categories that are not going to be generally

renounced—e g. objects. It may be felt that the existence of numbers, or possible words, or 11-dimensional superstrings, etc., would be intolerably weird.

And this worry may lead to the thought that the terms that seem to refer to them don't really do so, or that their referents are no more than useful

fictions, or that we are dealing with ‘real objects’ in some less than full-blooded sense, etc.

So far so good, perhaps. But we are owed an account of the robust/merely-deflationary distinction. And no satisfactory
way of drawing it has yet been established. Not that there is any shortage of competing candidates. The domains
whose truth-making facts are robustly REAL have been variously defined as those that:
(a) obey the law of excluded middle;
(b) induce a causal theory of reference;
(c) exhibit epistemological convergence;
(d) yield consensus amongst non-defective enquirers with the same evidence;
(e) are such that any justified belief that p is explained by the fact that p;
(f) satisfy a correspondence account of truth;
(g) are best treated by a truth-conditional theory of meaning;

end p.194

(h) contain undiscoverable facts;

(i) have many of the above characteristics.
And there are a fair number of further suggestions as well. The problem is to decide which, if any, of these criteria is
the correct one—which of them merits association with peculiar ontological weight.

Notice that although proponents of the REAL/un REAL distinction will tend to place humdrum physical facts on the REAL

side of it, this is not simply a matter of definition. It is rather a matter of motivation. For, although our sense of there
being such a distinction is provoked by worrying contrasts between ordinary physical facts and others, once the idea of
the robustly REAL takes hold it would then be conceivable for a philosopher to persuade himself that the right account
of that esteemed metaphysical status also excludes physical facts—indeed it could turn out to exclude all facts. Thus
Richard Rorty and John McDowell might be interpreted as maintaining that the domains of the REAL are those
possessing a certain extreme objectivity, those for which a correspondence theory of truth is correct, those whose
14
concepts are not to be explicated in terms of human practices and interests:—i.e. no domains at all!
14
See Rorty's “Is Truth a Goal of Enquiry? Davidson vs. Wright,” Philosophical Quarterly, 45 (1995): 281–300 , and his “Pragmatism,

Davidson, and Truth,” in E. Lepore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Blackwell, 1986. And see

McDowell's “Non-Cognitivism and Rule-Following,” in C. Leich and S. Holtzmann (eds), Wittgenstein: To Follow a Rule, Routledge & Kegan Paul, 1981. I

am not entirely confident that this is the right way of interpreting Rorty and McDowell. It may be that one or both of them should be read as claiming, not

that our notion of REALITY clearly excludes all facts, but rather—as I myself w ll go on to argue—that it is misconceived and radically defective.
 p p p p p y

However, one might well be even more skeptical than this about the robustly REAL. For there would appear to be no
clear justification for preferring any specific test of REALITY (including any of the ones that nothing is able to pass). So
it would seem that the real trouble with the notion isn't its failure to apply to anything, but its failure to provide a way
of deciding what it does and does not apply to. In which case, arguably, there are no determinate facts of the matter
here. Indeed one might well go further—one might suspect that ‘robust REALITY ’ is such a bad notion that it should
15
simply be thrown away.
15 Of course we are at liberty to define the jargon term “robustly
REAL ” in any way we want—e g., by picking one or another of the above-mentioned

criteria. But, in the first place, any such resort to stipulation concedes that there is no prior notion of REALITY worth trying to capture. And second, the

aptness of the stipulated terminology—its suggestion of two levels of reality—would call for some sort of defense.

The most sophisticated and promising attempt (that I have seen) to address these concerns head-on is Kit Fine's 2001
16
paper, “The Question of Realism.”
16 Published by Philosopher's Imprint.

Fine does not offer a definition of the intended sense of “REAL”. Indeed he says that we should no more expect such an
account than we should expect a definition of “good”. What we can and should do instead, he says, is specify

end p.195

constraints on the REAL that are rich enough to motivate a procedure that will allow us to figure out, for various
domains, whether or not its facts are REAL. And he proceeds to implement this strategy in a strikingly ingenious
17
way.
17
The following exposition of Fine's proposal takes a certain liberty with his terminology for the sake of conformity with my own. I say that the target of

his proposal is the concept of REALITY. But he himself describes it as focused on FACTUALITY (in a non-deflationary metaphysical sense of the word), and he

reserves the term “REAL” for what is both FACTUAL and irreducible.

Unfortunately there's no space here for most of the interesting details. But the core of his account is a derivation of
his procedure from the following assumptions:
0. Every fact is either basic (i.e. fundamental) or else grounded in (i.e. constituted by) more basic facts.
1. If a fact is REAL , then the basic facts that constitute it must also be REAL.

2. All the components of a basic and REAL fact are REAL.

3. Facts of assertion, belief, etc.—like the fact that John believes torture is wrong—are REAL.

4. If something REAL and basic is a component of some fact, then the fundamental constituting grounds of that fact
must also contain it.
5. If something is a component of a fact, then it is a component of assertions, beliefs, etc. with respect to that fact.
From these supposedly intuitive constraints on robust REALITY , Fine infers that, in order to discover whether some
basic property, f-ness, is REAL, a good strategy is to investigate the fundamental constituting grounds of assertions,
beliefs, and other attitude-facts regarding f-ness. For, if we find, for example, that wrongness is contained in the basic
grounds of the fact (agreed to be REAL) that S believes that torture is wrong, then—since all the components of the
basic grounds of REAL facts are REAL—wrongness must be REAL . But if we find that the basic grounds of that belief do
not contain wrongness (and if we assume that wrongness is basic) then—since any basic REAL component of a fact
must recur in the grounds of that fact—wrongness cannot be REAL .

Note the similarity between Fine's conclusion and a criterion of REALITY that is often bandied about: namely, that the
mark of something's being robustly REAL is that our thoughts about it are in some way causally related to, or grounded
in, the thing itself. On such a view, for example, if our ethical beliefs are constituted from our desires and, therefore,
are not in any way dependent on the properties of rightness and wrongness, then ethical facts could not qualify as
robustly REAL.

But it would not do justice to Fine's analysis simply to assimilate it to that familiar idea. For his important innovation
is to appreciate that it is not

end p.196

enough to state one's allegiance to some criterion of robust REALITY . Alternative intuitions on this point are two-a-
penny, as we've seen. What is needed, in addition, is some independent motivation for adopting one criterion as
opposed to the others—some objective reason for thinking that a peculiar ontological weight attaches to the facts that
satisfy it.

However it's one thing to recognize this need, and another to meet it satisfactorily. The trouble with Fine's account, it
seems to me, is that its proposed intuitive constraints on the robustly REAL are not compelling.

Consider principle 4. The intuition that a basic and REAL component of a fact must recur somewhere in what
constitutes that fact, is no more powerful than the parallel idea that a basic and unREAL component must similarly
recur. But the latter idea—combined with principles (1) and (2)—would imply, contrary to principle (3), that attitudes
with respect to unREAL facts are themselves unREAL.

Conversely, given—as implied by the possibility of anti-REALISM—that there may exist grounds, not containing a certain
basic un REAL element, that constitute facts that do contain it, one would expect the same sort of thing to happen with
REAL elements. In particular, a belief-fact containing a certain basic element (whether REAL or unREAL) is likely to be
 p p p p p y

constituted, in part, by the fact that some term (of the believer) refers to that element. This fact of reference is likely
to be grounded in some fact about the usage of that term. And that fact of usage may well not take the form of a
relation between the term and the element to which it refers—even when that element is REAL. It could be, for
example, that the word, “quark”, refers to ‘quark-hood’ merely in virtue of our reliance on certain principles,
18
“T(quark)”, involving that word.
18
For a fuller appraisal of Fine's theory, see my “The Quest for REALITY”, Dialectica, forthcoming.

Thus Fine's attempt to show that we do have a useful conception of the robustly REAL—something rich enough to
rationalize a specific method of detecting its presence—is not convincing. One cannot help but suspect that his
principles were selected, not entirely on the basis of intuitive plausibility, but also with an eye on their ability to
deliver that particular method. But, bereft of any genuinely motivating rationale, the method's intrinsic defects
become even more telling. For what it amounts to, as we have seen, is a version of the familiar idea that f-ness is REAL

when thoughts about it are grounded in facts involving f-ness. More specifically:

The entity designated by a given term

is robustly REAL iff some non-semantic
(naturalistic) relation between them is what
engenders that designation.

end p.197

19
Thus REALITY becomes the preserve of those concepts to which a relational meta-semantics applies.
19 Examples of this sort of relational meta-semantics are Jerry Fodor's informational theory (see his Psychosemantics, Cambridge, Mass., MIT Press,

1987) and Ruth Millikan's “teleological theory” (see her Language, Thought and Other Biological Categories, Cambridge, Mass., MIT Press, 1984).

But, in the first place, such a criterion, unmotivated, amounts to the un-illuminating stipulation that nothing beyond
the naturalistic is to qualify as robustly REAL. And, in the second place, even within the naturalistic domain there are
phenomena which someone might well wish to include within robust REALITY , but for which no relational account of
meaning-constitution is especially plausible. As suggested above, one might well think that the theoretical predicates
of physics derive their meanings (hence designate the properties they do) by virtue of our acceptance of theory
formulations containing them, rather than by virtue of standing in causal relations to the properties they designate.

The failure of Fine's strategy—our best hope for vindicating the notion of robust REALITY —must fuel the suspicion that
there is something irremediably wrong with this idea. For one thing, if the theory of REALITY that provides its content is
too weak to allow any determinate facts as to its extension, then the notion is useless. But, even worse, we may well
come to suspect that this theory is ill-motivated and irrational—in which case, the notion it engenders is defective and
should be abandoned. There will be no point in trying to strengthen it in such a way as to resolve the indeterminacy
problem.

Indeed, it seems to me that we can identify two clear mistakes underlying the tendency to invoke a REAL/un REAL

distinction. In the first place, it has often been thought that the correct theoretical account of meaning will imply that
some of the declarative sentences we naively deploy are in fact MEANINGLESS (in the technical sense of “meaning” that
is proposed by the theory) and therefore do not concern REALITY . It may be conceded that there is an ordinary,
pre-theoretic notion of meaning in which such sentences do have meaning, and therefore purport to describe facts (in
some “weak” sense); but it will be natural to give a special priority to those facts that are articulated by the
20
MEANINGFUL sentences—natural to define REALITY as comprising just those facts.
20
See, for example, Wittgenstein's Tractatus, the positivists, and Michael Dummett's critique of classical logic.

But the mistake here is that (following the later Wittgenstein) we can devise a perfectly good theory of meaning—a
use theory—that accounts, in a unified way, for the meanings of all our expressions. Therefore, there is in fact no
difference between the sentences we naively take to be meaningful and those that our best theory tells us are
MEANINGFUL. So there is no basis for drawing a distinction between REALITY and mere reality.

A second motivation for taking seriously the notion of robust REALITY is its role in the treatment of metaphysical
queasiness—in enabling us to consign weird-seeming facts to a relatively superficial level, and to thereby preserve our
intuition

end p.198

that they could not be ‘out there’ in the way that physical things are. In other words: given the queerness of certain
kinds of fact—i.e., the difficulty of seeing how such things could really exist (as physical facts presumably do)—it
appears reasonable to infer (to find it intuitive) that there exists a special kind of metaphysical ‘solidity’ possessed by
certain facts, notably, physical facts, but not others.

However, there is an alternative way of responding to the sense of weirdness provoked by numbers, oughts, possibilia,
and so on. Instead of explaining it as the result of our imagining certain facts as REAL which are not REAL, we can
explain it—in much less extravagant terms—as the result of imagining these facts as physical when they are not.

Elaborating this alternative, the basic thought is that we tend to over-generalize. Most of the facts that we
encounter—most of the facts that we care about—involve material objects and their empirically detectable properties.
 p p p p p y

As a consequence, these facts become our paradigms. So much so, that the very idea of a fact comes to seem
conceptually bound up with the characteristics of physical facts. Thus the acknowledgment of genuine ethical facts,
arithmetical facts, modal facts, and so on, becomes paradoxical: on the one hand they are unlike physical facts; but,
on the other hand—in order to qualify as facts at all—they must resemble physical facts. Hence the aura of
21
weirdness.
21
Sometimes the impression of weirdness manifests itself merely as a gut feeling; sometimes it is cryptically expressed in the thought, “How could there

possibly be such things?!”; and sometimes the puzzlement is elaborated in the form of an articulated paradox: e.g. (1) No fact can be such that an

awareness of it is intrinsically motivating; (2) The awareness that torture is wrong is intrinsically motivating; (3) Torture is wrong it is a fact that

torture is wrong.

And for similar reasons their epistemological status can seem problematic. For knowledge of the physical requires
certain causal or counterfactual relations between the facts and our beliefs about them—relations that cannot hold in
a variety of other domains. Thus generalization from the physical case—over-generalization again—will lead to
skepticism with respect to those areas.

But if this diagnosis of our metaphysical and epistemological troubles is right—and it seems quite plausible to me—the
proper response is, not to produce a two-level theory of reality, but merely to guard against the initial scientistic
tendency to over-generalize. We should take on board (again following the later Wittgenstein) that the terms of a
normal language vary enormously with respect to how they are used and what they are for. In other words, colour
concepts, size concepts, logical concepts, normative concepts, temporal concepts, modal concepts, numerical concepts,
etc., have very different conceptual roles—including different epistemological characteristics and explanatory
potentials. The point is not to exaggerate this diversity; nor is it to disdain economy and simplicity as virtues, even in
philosophy. It is to face the fact that this diversity is really there, that its presence is wholly unsurprising, and that it
cannot be magicked away. The simplicity that we have abundant empirical reason to expect in fundamental physics,
for example, is not to be expected here. Once this

end p.199

point has properly sunk in, there will seem to be nothing especially peculiar about the variety of facts that are
articulated by means of all these concepts—no reason to feel that the world would be a nicer place if everything were
22
physical. On the contrary, vivent les differences!
22
A further defect in our weak ‘theory of REALITY’ is elaborated in “The Quest for REALITY” namely, its inability to provide an explanation, in terms of the

character of our concept of REALITY, of why we should have any concern at all for which facts are REAL —why we should think of such facts as having

‘greater ontological weight’.

If these considerations are correct then our intuitions about robust REALITY —our concept-constituting claims about
it—are irrational. They derive from an irrational view of meaning, and they are sustained by an irrational view of how
best to explain the phenomena of weird-seeming facts. So the notion they engender should be given up.

We are now well placed to address the pros and cons of a further theoretical approach—namely reductionism. Unlike
other “isms” in the area, this sort of theory cannot be condemned out of hand. Clearly there are cases in which
explanatory progress is made by reducing a certain body of facts to some better understood body of facts. Life, as it
turns out, doesn't involve any sort of peculiar vital fluid; it boils down to DNA. Of course, our present concern is not
with science but with philosophy—not with empirical reduction but with a priori reduction. Still, there are many
reasonable attempts at the a priori form too: e.g. conceptual analyses of “brother” as “male sibling”, of “bachelor” as
“unmarried man”, of “chair” as “movable seat with a back”, and so on. There are even cases that are not merely right
(at least, to a first approximation), but philosophically interesting as well—e.g., accounts of meaning in terms of use,
of causation in terms of law-governed association, and of knowledge as “justified, true belief.” But note that such
analyses are motivated by nothing beyond the innocent desire to articulate the character of a concept; they are
obliged to be faithful to the distinctive complexities of its deployment; their adequacy hinges on no more than that.
Thus they stand in contrast with attempts that are constrained by a further agenda: namely, an effort to avoid the
sort of weirdness on which we have been focused—the sense of weirdness that stems simply from radical differences
in how different words are used. In such cases it has been found tempting to insist that, appearances to the contrary,
the facts in question are really nothing more than obvious physical facts: arithmetic is really about marks on paper;
history is really about current records; ethics is really about expressions of desire-satisfaction; and so on. Contrived to
23
the point of absurdity! But what can be expected of proposals so woefully misbegotten?
23
The highly idiosyncratic character of a species of fact can pull us in conflicting directions when we consider the prospect of their physicalistic reduction.

On the one hand, it motivates a reduction; for the existence of such facts would be shown, thereby, not to be weird after all. But, on the other hand, it

can make any reductionist proposal seem implausible; for it can be hard to see how such an account could do justice to the unique character of the facts

at issue.

A good example is provided by mental facts, facts of conscious experience. Considerations of weirdness can make their reduction to physical states seem

at once necessary and impossible! In this case, what we should say (as suggested in the text) is that the pressure towards reduction (the hope for such a

theory) stems from illegitimate overgeneralization, and so the pressure against it—at least when the issue is conceptual reduction—is quite right.

But when it comes to the prospects for an empirical reduction—in the spirit of water reducing to H2O—matters are somewhat reversed. For such an

account is well motivated, not by the desire to avoid weirdness, but by the desire to avoid an implausible causal over-determination. (See Donald

Davidson's “Mental Events” (1970), reprinted in his Essays on Actions and Events, Oxford University Press, 1980, and David Papineau's Thinking about

Consciousness, Oxford University Press, 2002.) To see this, consider the mental cause of some item of behavior. This mental cause either (1) precludes
 p p p p p y

there being a complete physical cause of that behaviour; or (2) exists in addition to the physical cause; or (3) is constituted by it. But option (1) conflicts

with ‘the causal autonomy of the physical’. And option (2) involves a remarkable causal over-determination. So the only reasonable option is (3) namely,

reductionism. This line of thought gives us good reason to expect an empirical reduction of the mental to the physical. And it is not undermined by the

above-mentioned anti-reductionist considerations—those that stem from the idiosyncratic character of mental facts—for they have force only when the

issue is conceptual analysis.

end p 200

And something similar can be said about revisionist theories—for instance, intuitionistic logic, amoralist ethics,
anti-essentialist modality, and finitist arithmetic. Like reductionisms, revisionist proposals cannot be dismissed per se.
We often have excellent grounds for revising our views. Indeed it may well be that no claim within any domain is
exempt; perhaps fallibilism rules. So there is nothing in general wrong with recommending modifications—even
revolutionary modifications—in a given discourse. But there is something badly wrong with modifications that are
inspired merely by over-stretched analogies and ontological panic. And this is what is going on when the queerness of
unverifiable facts provokes abandonment of the law of excluded middle, when the queerness of normative facts
provokes a rejection of unrelativized normative discourse, when the queerness of essentialist facts provokes
avoidance (à la Quine) of ‘the third grade of modal involvement’, or when the queerness of infinitary facts provokes a
distortion of standard arithmetic. Thus, although revisionist suggestions, when properly motivated, are what
intellectual progress is all about, the above philosophical moves—which respond to an irrational metaphysical
anxiety—must be seen as perversely regressive.

Suppose our critique of all the various ‘weirdness avoidance strategies’ is right. Where does this leave us? Are we
realists or anti-realists? Or does some other ‘ism’ apply? The moral of the story is that all these ‘isms’ are discredited.
They are alternative theoretical responses to conceptual tensions that are better dealt with another way—a more
therapeutic way. Thus we should not be asking, for example, whether the normative expressivism of Simon Blackburn
and Allan Gibbard is a form of anti-realism. Those authors may have sometimes written as if it importantly is—as if
24
one of its main virtues is weirdness-avoidance.
24
See Blackburn's Spreading the Word, Oxford University Press, 1984, and Gibbard's Wise Choices, Apt Feelings, Harvard University Press, 1990.

But

end p 201

from the present perspective that is a faux pas. The real and considerable interest of their view lies in its emphasis on
the distinctive conceptual roles of normative concepts—specifically, the basic tie between the appearances of such
concepts in someone's belief box and their appearances in his desire box. But this idiosyncrasy need not and should
25
not engender any ontological or epistemic puzzlement.
25 For further discussion, see my ‘The Frege Geach Point’, Philosophical Issues, 15 (2006): 78–93 .

Of course it was wishful thinking for me to have suggested in my title that this could spell the end of all “isms.” For
26
there was bound to be one for the very point of view that I'm recommending.
26
It might be suggested (indeed it has been, by Marian David) that insofar as I am allowing that there are moral facts, arithmetical facts, etc. (in a

deflationary sense of “fact”) then I am, after all, advocating realism with respect to all these domains. And I don't deny that one is free to define the

jargon term “realism” in such a way that it applies to my position. But I myself think that if there really has to be an “ism” term for this point of view, then

“realism” is not a good choice. For it has been ruined by having been given too many different meanings by different philosophers in different contexts. The

result is that, now, whenever anyone uses it, they have to add, “and I mean it in the following sense: ‘blah, blah, blah’.” But isn't the whole point of naming

a position to avoid having to spell it out on every occasion?

I wouldn't mind using Arthur Fine's turn of phrase, “the natural ontological attitude,” for the point of view developed here; but there is some uncertainty

about what he himself means to associate with the expression. (See Fine's ‘The Natural Ontological Attitude,’ in J. Leplin (ed.), Scientific Realism,

University of California Press, 1984.) For, having located his position, “NOA”, as occupying a certain space between realism and anti-realism, he proceeds

to characterize rea ism in various radically non-equivalent ways: (1) as the view (pace van Fraassen) that theoretical beliefs are perfectly legitimate; (2)

as the view that modern physics (e.g. Quantum Mechanics and General Relativity) aims to describe a hidden reality and not mere devices for the

organization of experimental results; and (3) as the inclination to supplement theoretical assertions with table thumpings and exclamations of “And that's

how things really are!” My own take on these three ideas is that “realism(1)” is correct (for t is no more than a rejection of skepticism); that “realism(2)”

is also correct (for it is no more than a rejection of instrumentalism); and that “realism(3)” is correct if it is regarded as a rhetorical reaffirmation of (1)

and/or (2), but incorrect if it is construed as gesturing towards “special ontological weight.” To the extent that the non-realism in Fine's NOA is restricted

to the last of these points, I am in full agreement with it.

A “bloodless quietism” is how Crispin Wright has labeled it—“the bland perspective of a variety of assertoric ‘language
games’, each governed by its own internal standards of acceptability, each sustaining a metaphysically emasculated
27
notion of truth, each unqualified for anything of more interest or importance.”
27 See Wright's Truth and Objectivity: 76.

Well, Crispin, I'm sorry for being so anemic, boring, and effeminate, but I confess to thinking that if the “anything of
more interest or importance” is the project of describing REALITY (with capital letters), then we should indeed give it
28
up—or better, give up the idea that there was ever any well-founded project there in the first place.
28 Let me stress that in recommending that we adopt a stance of “bloodless quietism” towards the debates surrounding realism that have been under

discussion here, I am not implicitly suggesting that we adopt it towards all philosophical disputes and theories. I admit to being in some sympathy with
 p p p p p y

Wittgenstein's general anti-theoretical meta-philosophy, and I think that its successful application to realism debates should encourage us to take it

seriously throughout the subject. But in this paper I haven't even attempted to articulate that perspective properly (which would be no easy task), let

alone to argue in favor of it.

end p.202
 p p p p p y

11 Horwich's World

Paul Horwich calls his world “a world without isms.” You probably noticed, though, that it isn't really a world without
isms. As Horwich himself acknowledges, there is room for a number of isms in his world, provided they are “properly
motivated.” There is room for revisionism as well as for two non-revisionist isms: namely reductionism and bloodless
quietism. The isms Horwich does not want to make room for in his world include non-cognitivism, fictionalism, some
forms of subjectivist relativism, and realism. Horwich rejects realism explicitly, together with its opposite number,
anti-realism. These isms are “discredited,” he says.

My concern here is with realism. By my lights, Horwich has no business rejecting realism: this ism is already part of
Horwich's world, even if he doesn't like to admit it. As I see it, Horwich's explicit rejection of realism is merely verbal.
The “bloodless quietism” Horwich is content, even happy, to embrace is simply a form of realism under another name;
in particular, it is non-reductive realism under another name. Let me put this slightly more precisely. Horwich
advocates bloodless quietism, but not across the board. He holds that bloodless quietism is appropriate for domains for
which revisionism and reductionism are not appropriate—he usually mentions the domains of ethical facts, modal
facts, and arithmetical facts in this context. What I maintain is this: Horwich is committed to non-reductive realism
with respect to these domains, i.e. with respect to all, but only, those domains for which he takes bloodless quietism
to be appropriate.

Actually, I think this point is close to trivial. There isn't much to argue about here. It is mostly a matter of pointing
out the obvious. So, to add some zest to my comments, I will argue later on that one of the considerations Horwich
advances in favor of his bloodless quietism, i.e. in favor of his non-reductive realism, is vulnerable to the charge of
being overly naive.

In a moment, I will focus on Horwich's vivent les differences passage. But first I need to remind you of what has been
going on in Horwich's paper up to that passage. Horwich has pointed out that certain kinds of facts—e.g. ethical facts,
modal facts, and arithmetical facts—seem weird or queer to many philosophers. Because of this, he says, some
philosophers (metaphysicians) attempt

end p 203

in various ways to avoid commitment to such facts. Horwich criticizes these attempts: according to him, they lead
nowhere. He then offers a diagnosis of our metaphysical troubles together with a remedy.

Horwich's diagnosis is this: most of the facts we encounter involve (mid-sized) material objects and their empirically
detectable properties. We tend to over-generalize. Our tendency to over-generalize leads us to regard the familiar
(mid-sized) material object facts as the paradigmatic ones, with the result that other facts—facts that don't easily fit
the paradigm—come to seem weird or queer (cf. p. 199). That's the diagnosis.

Horwich's remedy is sketched in his vive les differences passage. The remedy has two parts (by the way, though
Horwich writes as if there were a natural connection between the two parts, it's not obvious that there is one). The
first part is very simple: we (philosophers) ought to guard against our tendency to over-generalize. The second part is
slightly more involved. This is the Wittgensteinian part of Horwich's remedy. It goes like this. Horwich reminds us that
the terms of our language “vary enormously with respect to how they are used and what they are for” and that we
have various different concepts with “very different conceptual roles” (p. 199). Call this the “variety of concepts”
point. Horwich then says (p. 200):

Once this point [the variety of concepts point] has properly sunk in, there will seem to be nothing especially
peculiar about the variety of facts that are articulated by means of all these concepts—no reason to feel that
the world would be a nicer place if everything were physical. On the contrary, vivent les differences!

In this passage Horwich commits himself to there being a variety of facts that are articulated by means of the different
concepts he has in mind; that is, by means of ethical concepts, modal concepts, and numerical concepts. This looks
like realism to me. Horwich is saying that there are ethical facts, modal facts, and numerical facts. That's realism with
respect to these facts. More specifically, it looks like non-reductive realism. For he is saying “on the contrary,” i.e. not
everything is physical. The emphasis on les differences signals a non-reductivist stance with respect to the facts in
question.

Horwich's non-reductivist stance with respect to the relevant facts comes out more clearly in the subsequent passage.
There Horwich talks about reductionism. He allows for empirical (i.e. scientific) reductions. He also allows for some
forms of a priori (i.e. philosophical) reductions, as long as they are “motivated by nothing beyond the innocent desire
to articulate the character of a concept” (p. 200)—this form of philosophical reduction is roughly the same as what
used to be called “conceptual analysis.” What Horwich does not allow for is attempts at reduction that are motivated
by the desire to avoid the sort of weirdness we (philosophers) experience when considering ethical, modal, and
arithmetical facts. Attempts to reduce such facts to physical facts are, Horwich exclaims, “contrived to the point of
absurdity” (p. 200).

end p 204

So, here it is. I don't see how Horwich can say, as he does, that realism is one of the isms discredited by his remedy.
 p p p p p y

On the contrary, his remedy simply consists in embracing non-reductivist realism with respect to facts of various sorts
that other philosophers take to be weird. By the way, Horwich is a realist in any case. For, as we have seen, he
accepts non-revisionist reductions of the scientific kind. This is a form of realism too: reductive realism. However, the
sort of facts for which scientific reductions are available are not the sort of facts that are at the focus of Horwich's
concerns (ethical, modal, and arithmetical facts), so I won't press this point.

Two supplementary remarks.

First, one might point out that realism with respect to Fs requires a bit more than a mere affirmation of the existence
of Fs, that it requires in addition some claim to the effect that the Fs are not mind-dependent or evidence-dependent
or something like that. Does Horwich make any such claim with respect to the sorts of facts under consideration? Yes,
namely when he discusses, and rejects, fictionalism and subjectivist relativism—two views on which the relevant facts
would come out as mind-dependent.

Second, where I detect Horwich's non-reductive realism, Horwich—towards the end of his paper—talks about bloodless
quietism. Maybe this indicates that Horwich would be inclined to respond to my point along the following lines—I
quote Horwich's imaginary response: “Okay, call it realism, if you want; but it's not full-fledged REALISM, with capital
letters: it's merely realism, with small letters.” I am by no means sure that Horwich would be inclined to respond in
this manner. In fact, there are strong indications to the contrary in his discussion of Kit Fine's view. On the other
hand, there are also indications, at the end of his paper, that he might be tempted to respond in this manner after all
(cf. p. 202). Anyway, if he did respond in this manner, I would, in turn, respond like this: I have no idea what the
difference between capital letters and small letters has to do with realism. You have said that there are physical facts,
and ethical facts, and modal facts, and arithmetical facts. You have said that these facts are not subjective, and you
have said that they are different facts; in particular, you have said that the last three sorts of facts are different from
physical facts. You are a non-reductive realist with respect to these facts.

So far I have pointed out that, despite his claim to the contrary, Horwich advocates non-reductive realism with respect
to various facts that are often regarded as weird by philosophers. Now I want to argue that some of Horwich's grounds
for advocating non-reductive realism with respect to these facts—his Wittgensteinian-therapeutic grounds—are a bit
meek. I think Horwich is vulnerable here to the charge of giving in to some sort of Wittgensteinian naïveté—of moving
too hastily from a point about language and concepts to a point about facts.

Let's take another look at the vivent les differences passage, cited above. There Horwich begins with his point about
the variety of our concepts. He reminds us

end p 205

that we have various different concepts with various conceptual roles. He then moves rather briskly from the variety
of concepts to the variety of facts that are articulated by all these concepts: he suggests that, once the point about
the variety of concepts has properly sunk in, we shouldn't be surprised about the variety of facts that are articulated
by these different concepts. I don't agree. I think we should be quite a bit surprised if the variety of facts articulated
by our concepts matched the variety of our concepts.

I can see how having lots of different concepts might naturally lead to there being different conceptual articulations of
the facts, whatever the facts are. In other words, I can see how one can move fairly briskly from the point about the
variety of our concepts to the following point:

We have a variety of different conceptual articulations of the facts, whatever the facts are.

So far, so good. What I don't quite see is how one can move at all briskly from here to this further point:

There is a variety of different facts that are being articulated by our different concepts.

This step strikes me as dubious and rather contrary to reasonable expectations based on ordinary experience.

As I move and look around I am presented with different views. Some of these different views are indeed views of
different things or situations. But by no means all of them. Very often different views are not views of different things
or situations but merely different perspectives of the same things or situations. Indeed, I am quite familiar with there
being different perspectives from which the same things or situations look rather different. I would be rather surprised
if it turned out that different views typically or reliably presented me with different objects or situations.

The same goes, I think, when it comes to my different conceptual articulations of the facts. On the face of it, the
reasonable thing to expect is that quite a few of my different conceptual articulations of the facts are merely different
conceptual articulations of the same facts—different conceptual perspectives of the same facts. Contrary to what
Horwich suggests, I should be surprised if I could move swiftly from the variety of my concepts to a variety of different
facts articulated by my concepts: I should be surprised if turned out that my different conceptual articulations were
typically conceptual articulations of different facts. Horwich's Wittgensteinian point about the variety of concepts
doesn't support non-reductionism; it doesn't support his vivent les differences attitude with respect to facts.

Of course, it doesn't support reductionism either. On the face of it, I should be equally surprised if it turned out that,
typically, my different conceptual articulations were merely different conceptual perspectives of the same facts. On
the

end p 206
 p p p p p y

face of it, the reasonable thing to expect is some mixture: that quite a few of my different conceptual articulations
articulate the same facts and quite a few articulate different facts. In other words, on the face of it—i.e., merely from
considering the variety of my concepts and independently of any more theoretical grounds—I should expect
reductionism to hold for some domains and non-reductionism to hold for some other domains.

I am not saying that there are no good grounds for rejecting reductionism with respect to ethical, modal, and
arithmetical facts—for present purposes, I remain neutral about this. I am saying, though, that these grounds have to
be found elsewhere: Horwich's Wittgensteinian point about the variety of our concepts won't provide them. Say we
have “no reason to feel that the world would be a nicer place if everything were physical,” as Horwich puts it in the
passage cited above. When he adds “On the contrary, vivent les differences!” and invokes the variety of facts that are
articulated by means of our concepts, he goes beyond what Wittgenstein can provide. The Wittgensteinian point
cannot take us from: “We have no reason to feel that the world would be a nicer place if everything were physical” to:
“We have a reason to feel that the world would be a nicer place if some things were non-physical.”

end p 207
 p p p p p y

12 Intuitions and Truth

Philosophy has perennially relied on what seems intuitively right to the reflective mind. It did so already with
paradoxes such as the liar or the sorites, and with hypothetical examples like the Ring of Gyges, or the knife withheld
from its rightful owner gone mad, or the statue and the lump. And it does so today with trolley cars, split brains,
Matrix scenarios, fake barns, and Twin Earth. It is such reflection that drives philosophical controversies concerning
reference, or personal identity, or propositional knowledge, or essentialism, or externalism and internalism, whether
in philosophy of mind, or in ethics, or in metaphysics or epistemology.

What then is intuition? What accounts for its probative force? How is intuition related to truth? These are questions to
be addressed in what follows.

I
Traditionally, intuition is understood on a perceptual model. It is through the mind's eye that we gain insight. In
perception one's eyes may come to rest on an object seen in good light. A sensory experience then mediates between
object seen and perceptual belief, as in the following example:
• One's facing a red, round apple with open eyes in good light.
• One's visual experience as of seeing a red, round object.
• One's belief that one sees a red, round object.
Here the second item rests causally on the first, as does the third on the second, on which it is also rationally based.

Two problems arise for the perceptual model. First, nothing like sensory experience seems to mediate analogously
between facts known intuitively and beliefs through which they are known. Consider for example:
• The fact that 1 + 1 = 2.
• Belief that 1 + 1 = 2.

end p 208

Here no sensory experience mediates between fact and belief, nor does anything like sensory experience play that
role. We are aware of nothing like that sandwiched between them, on which the belief might be rationally based.

Moreover, many truths known intuitively lie outside the causal order, unable to cause experience-like intuitions, even
if there were such intuitions. Nor can such truths be tracked, not if tracking requires sensitivity. What are we to make
of the claim that if it were not so that 1+1=2, one would not believe it to be so? Hard to say, but that is what tracking
it with “sensitivity” would require.

Even if there are no experience-like intuitive seemings, intuitions do not collapse into beliefs; they remain distinctive
conscious states in their own right. Thus consider paradoxes like the liar, or the sorites. Each proposition in a
paradoxical cluster exerts a powerful intuitive attraction, although it is equally intuitively plausible that they cannot
all be true together. Even when one eventually settles on a solution, moreover, the pull of the rejected proposition is
not removed but overcome.

Paradoxes also establish that a powerful intuition can even be false. A proposition might remain plausible upon
reflection even when its rejection is required by one's solution to a paradox.

What then might intuitions be, if they are to be conscious states with probative force despite being fallible, while
distinct from beliefs?

II
When we deliberate on a choice or ponder a question, we might give weight pro or con to a consideration, being
attracted either to assent: consciously to choose or to affirm; or to dissent: consciously to opt against, or to deny.
Even when a reason for something is defeated by some contrary reason, it might still exert attraction. Forces vie in
our minds when we deliberate on a choice or ponder a question, introspectable forces with psychological reality. There
is something it is like to host a conflict among them. I take these to be facts familiar to us all.

In perception we encounter such forces, as when Mueller-Lyer lines seem incongruent while background knowledge
suggests they are congruent. Until we measure, and even afterwards, we are attracted to think the lines incongruent.
Here we find three distinct elements:
(a) the belief as to the congruence of the lines
(b) the attraction to judge them incongruent
(c) the visual experience that explains the attraction.
The pull of a paradox-constituting proposition can survive our disbelieving it, as required by our solution to the
paradox. At least initially, the proposition remains intuitively plausible even so. What conscious, introspectibly
accessible state might constitute such intuition?

end p 209

Why not the attraction itself? The pull itself may play the role of intermediary, when it outpulls any conflicting
 p p p p p y

reasons. Which attractions count thus as intuitive? Many are the mechanisms of attraction, many the ways in which a
proposition can appear true. Which count as intuitive? Do such intuitions have probative force? If so, how is it
explained?

We begin with a negative characterization of intuition as attraction that does not derive from perception,
introspection, testimony, memory, or reasoning, whether inductive or deductive. On this broad conception, there
might be radically different sorts of intuition, which could vary in probative force. What would determine such
probative force? Plausibly, this would depend on the truth-reliability of the specific source of attraction.

Within the class of simple arithmetical propositions, or the class of simple geometrical propositions, or their union,
etc., one is able to discriminate the true instances from the untrue. Are we thus able to discriminate, among the
simple enough propositions, those that are true, as such? Can we discriminate them thus, even if we do so by means
of other features, intrinsic or relational?

Take the enormous diversity of shapes, including those of letters or phrases of a written language; also colors,
sounds, and other sensible characteristics; also the numerals, or the numbers themselves. For each such dimension,
each of us is able systematically to discriminate truths from falsehoods within some range of the simple enough. And
there seems no way to capture our truth-tropic abilities without appeal specifically to the truth of the propositions
distinguished as true.

For each such dimension, we enjoy some ability systematically to discriminate the true as such directly upon
understanding, upon grasping the objects of discrimination. Exercises of this ability are apt candidates for the
traditional title of a priori intuition.

Intuitive belief may accordingly have a basis in the subject's understanding of its specific content. But the set of truths
one accepts this way is not the same for everyone; much less is it the same for every conceivable subject. There is
something further about a subject, beyond being a subject, and about a truth with a certain understood content,
beyond its being a truth with that content, which presumably accounts for why that subject accepts that truth merely
upon understanding it: that is to say, while relying neither on the support of any reasons, whether perceptual,
1
introspective, deductive, or inductive, nor on memory or testimony.
1 Here, I am abstracting away from cases of conflicts of reasons, as in paradoxes, where the subject's assignment of positive weight to the proposition as

a candidate for belief is outweighed. So I am abstracting away from the complications that arise in such cases when other things are not equal.

An arithmetical truth known intuitively to Ramanujan may not be known to lesser mortals, even if well understood.

If this is how we think of the intuitively believed, and the intuitively justified and known, there is no apparent reason
to restrict the scope of intuition to necessary truths only. Descartes, rationalist par excellence, did not so restrict it. No

end p 210

cogito thought, for example, is necessarily true. Indeed, Descartes's paradigm of something clearly and distinctly
perceived—a felt pain, or, better, that one then feels such pain—is no necessary truth. In these cases, cogito and
self-presenting cases, the truth of what one believes, upon considering it with full understanding, follows necessarily
2
given only the conditions of reference and truth for the content understood.
2
That these are all cases naturally of a sort w ll be more clear if we conceive of our knowledge of the felt pain as one whose content is that one feels

thus, where one ostends the very experience of pain that one then undergoes.

And that helps explain our reliability when we intuitively believe such truths.

Something similar may seem plausible for necessary truths generally, simple arithmetical and geometrical truths for a
start. Consider such truths, and also the class of “cogito” contents, as those whose conditions of truth or reference
guarantee that if believed they must be true. Not every such content is justifiably believed through mere
understanding, not nearly. If I in some superstitious way immediately accept some complex necessary truth, or some
complex “cogito” content, based on no rationale, my belief will not thereby attain justification, despite being
unavoidably true. After all, the “superstitious” way in which I accept that content could as easily or more easily have
3
given me a tendency to accept falsehoods.
3
Thus, a cogito content might be truth-functionally complex enough that its cogito character is not open to view, as with “I think,” but is rather hidden as

when “I think” is relevantly embedded in some intricate molecular compound.

Accordingly, the fact that a belief is intuitive will not suffice to make it epistemically justified, if the way in which the
belief is intuitive is one that could as easily make false beliefs intuitive.

The present approach makes room for distinctions among (a) rational intuitions, which all rational beings would have,
and (b) human intuitions, proprietary of humans. Now we face a choice. Should we allow also intuitions distinctive of a
given culture? No such extension is in line with traditional thought, according to which intuition could hardly derive
from enculturation with mores or myths. An individual's intuitive beliefs are those assented to directly or immediately
upon consideration (or at least those that exert a direct or immediate pull when thus considered). Such “directness or
immediacy” requires that the attraction not derive from other familiar sources: that it not derive just from
enculturation, for example, nor from perception, introspection, testimony, or inferential reasoning, singly or in
combination, not even through the channel of memory. More strictly, it is required that the attraction not derive from
such sources except insofar as the very understanding that exerts the attraction does derive from them. So, the
 p p p p p y

further sources are not required in addition to the understanding. Compatibly with this, it is of course possible for an
intuition to have any of various sources.

end p 211

It may be that any rational being who considered a given proposition would be directly attracted to it, making it an
object of rational intuition. Examples might be the simplest necessary truths of arithmetic, geometry, logic, and the
like. Basic principles of folk physics or psychology wired into us would be distinctive of some subset of possible rational
beings, presumably those with sensory equipment like ours, or even with faces and limbs like ours, and so on. These
basic principles—those distinctive of taking experience at face value, for example, or knowing the mind of another
directly from its facial and behavioral manifestations—would not be acquired by the individual through any normal
perception, introspection, testimony, or inferential reasoning, singly or in combination, not even through the channel
4
of memory. Should such basic principles accepted directly upon consideration then be counted as intuitive?
4 We have focused on intuition as a conscious phenomenon that requires consciously considering a proposition, thus putting the question of its truth on

the scales of judgment. This is in line with a traditional understanding of intuition as a conscious phenomenon. We can naturally abstract from this,

however, to include also implicit beliefs and how they come to be held, and it seems wise to so abstract in due course, though here we stick to conscious

intuition for simplicity.

If so, what gives epistemic efficacy to such intuitive acceptance? We have focused, for one thing, on why the subject is
so constituted and situated as to feel an immediate pull. Various sources are possible, and the epistemic efficacy of an
5
intuition might depend substantially on its source. Intuitions true by sheer luck provide no epistemic justification.
5 However, I do not myself draw from this any crucial advantage for “evolutionary epistemology.” It seems to me conceivable that Swampman, or some

variant, could eventually attain intuitive justification. But this might be because by that time the source of his relevant constitution lies in his developmental

history and not just in his serendipitous origin. Several other issues would need addressing in a satisfactory treatment of Swampman phenomena. Among

others, we would need to consider the bearing of a firmly established virtue, no matter how serendipitously acquired, on the epistemic status of its yield.

And we would need to address also how having a more comprehensive, evolution-involving explanation for the established virtues might bear on the

epistemic status of one's beliefs that they issue.

III
We have reached a conception of intuition as a state distinct from and prior to both belief and knowledge. Intuitive
belief is based on intuition but goes beyond it, and in turn constitutes intuitive knowledge only if all goes well.

Here we focus on propositional intuition, which has the following features:

(a) It is a conscious state.
(b) It has propositional content.
(c) It is distinct from belief. One can have an intuition that p without believing that p, as in a paradox.

end p 212

(d) Its content can be false; there can be false intuitions.

(e) It does not derive just from enculturation, nor from perception, introspection, testimony, or inferential reasoning,
singly or in combination, not even through the channel of memory.
(f) It can serve as a basis for belief, helping thus to provide epistemic justification for the supported belief.
An intuition is hence a conscious state with propositional content that can serve as a justifying basis for belief while
distinct from belief, not derived from certain sources, and possibly false.

A sensory experiential state can fit this profile impressively well. No wonder it is tempting to think of intuition as a
6
kind of perception, often called “insight.”
6
According to the OED, the Latin intuitus corresponds to: the action of looking upon or into; contemplation; inspection.

Should we yield to this temptation? At most with serious qualifications, or so I will argue.

Compare first of all introspection, and how it differs from perception. Perception that p involves an experiential state
as if p. While distinct from the fact that p, this experiential state rationally grounds and epistemically justifies S's
belief in that fact. Introspection, by contrast, involves no such distinct state. Introspectable states are “self-
presenting.” What rationally grounds introspective belief is normally the very fact believed or the state constitutive of
that fact.

Of course, a sensory experiential state need not lead to outright belief. In the Mueller-Lyer case it looks as if the lines
are incongruent but we don't believe our eyes. Eventually we may not even be attracted to think them incongruent.
Before we know of the illusion, however, when we have measured the lines just once quickly, we might well be
attracted to believe them incongruent though also pulled the other way by the result of our measurement. Here we
have three elements:
(a) Whether the lines are in fact incongruent.
(b) Whether we experience as if they are incongruent.
(c) Whether we are attracted to assent to (a) in virtue of (b).

Introspection is different. Take a visual experience of an image that encompasses just three fat black dots in a linear
 p p p p p y

horizontal array against a pure white background. Here one is strongly, overwhelmingly, attracted to think that the
dots in one's image number exactly three. What explains this attraction? We need not postulate some further
conscious state, beyond one's visually experiencing as if one sees three such dots, to explain this attraction. It's
enough that one does visually experience that way. If a proposition attributes to me a current state of consciousness,
therefore, where assent is the most attractive option, what normally explains the attraction is that very state. Thus, it
is simply the truth of the

end p 213

understood simple enough content that explains why one is drawn to assent and why one does assent.

We are contrasting two models of intuition: perception versus introspection. When, in visual perception, a fact attracts
our assent, what does so more directly is some corresponding visual experience, the undergoing of which is distinct
from the fact visually perceived. In visual introspection, by contrast, what attracts our assent is the very fact
introspected, or the corresponding state. Unlike visual perception, then, visual introspection has a propositional
content whose very truth is what explains why it draws our assent. It accords with this that there are no introspective
hallucinations.

Finally, intuition and introspection are also similar in that neither depends on any epistemic channel, on the operation
of anything like a sensory organ in its proper medium.

That all being so, the introspective model seems more properly applicable to intuition. Indeed, pure introspection
seems now a special case of intuition, introspective intuition. Thus, Descartes's own paradigm of clear and distinct
7
perception is the perception of a felt pain, when held distinct from any hypothesis as to its origin.
7 Compare his cogito thoughts, such as sum itself, “I exist.” These involve a sort of introspection even more similar to the intuition of a simple abstract

necessary truth. My existing, for example, though contingent, need have no conscious qualitative content, nor is it a state properly so called, something

that can figure as a node in the causal order. Further discussion of these may be found in my ‘Reliability and the Apriori,’ in Imagination, Conceivability and

Possibility (Oxford University Press, 2002), edited by Tamar Gendler and John Hawthorne.

We are thus led to the following account:

S intuits that p if and only if S's attraction to assent to p is explained by three things in concert: (a) that p
8
is simple enough,
8 Actually we must allow that even a qu te complex content might be known intuitively to be true, so long as there is a feature of it, some formal

property, in virtue of which it is true, one simple enough to be discerned intuitively as truth-ensuring. Thus one might know intuitively that a

proposition of the form not-(p p) is true, despite the considerable complexity of p .

(b) that S understands it well enough, and (c) that it is true.

Even if that is on the right lines, it is only a template to be applied case by case. Compare a correlate of this account
for the case of perception: here what constitutes proper endowment and situation will vary from modality to modality.
For each modality, nevertheless, some similar principle is at work. In each case we assume that if the proposition is
simple enough and true, and we have the competence and situation characteristic of that modality—say, good eyes
and unobstructed view in good light—then we will discern the truth reliably enough. It is presumably some such
knowledge that would enable us to predict and explain the answers of others on relevant multiple-choice tests.

Perceptually derived information is often channeled through memory or testimony, emerging as later knowledge.
Similarly for deductive or inductive

end p 214

reasoning, which also requires more direct inputs, though reasoning is of course more than a channel. Let us restrict
ourselves in what follows to direct input, the more likely home of intuition. So, again, we leave aside mediators such
as memory, testimony, and inference.

What someone else introspectively believes can be ascertained given their other conscious states at the time, which
we can often know through our grasp of their psychological make-up and situation. If we see their eyes trained on a
clearly visible scene, to which we have access from a similar angle, we can tell what things look like to them. Based
on this, we can tell further what they will believe about how things seem visually. And here's the part of special
interest to us: in thus predicting and explaining a neighbor's introspective beliefs, we appeal to something about the
relevant set of truths, perhaps to their simplicity. Much in her manifold of present consciousness is unknowable on
inspection because it is too complex.

The same goes for intuition, where again we can predict and understand a lot concerning the beliefs and verbal
behavior of someone else by appeal to three facts about a given question: first, whether they understand the
question; second, whether it is simple enough; and, third, which answer is true. It is in some such way that we are
able to predict and explain how people would answer multiple-choice questions of simple arithmetic, geometry, and
logic.

IV
Our tripartite proposal has its shortcomings. First, truth has a special role among the factors cited. One is intuitively
 p p p p p y

attracted to a propositional content because of its truth. That is what one's belief seems based on, moreover, once the
proposition is understood clearly enough, provided it is also simple enough. If I ask myself why I believe that 1+1=2,
the most natural answer is because it is true or because 1+1 is equal to 2. I don't adduce, as my reason for believing
it, the fact that it is simple. Its simplicity is rather an enabling condition, one that makes it possible for me to respond
to the relevant truth with proper assent upon consciously considering it with adequate understanding. The account as
it stands does not distinguish among these different roles.

That brings us to a second, more important, shortcoming: namely, that many utterly simple, well-understood
propositions do not elicit our intuitive response when true. Compare for example the proposition that you have a
headache with the proposition that I have one. The truth of the latter draws my intuitively positive response, once I
understand it, since it is simple enough to allow that. Yet the former is similarly simple, and equally well understood,
but draws no such intuitive response. Our special, intuitive access is clearly restricted beyond truth and simplicity, to
certain kinds of propositions. Which kinds? Kinds that turn out to be traditionally familiar: truths about one's own
conscious states of the moment, and truths about the subject matter of certain abstract fields such as mathematics
and logic.

end p 215

Is there any simple way to characterize the scope of the intuitive? If we put aside the introspectively intuitive—the
category of simple truths, well enough understood and grasped, about one's own current states of
consciousness—what is left over? It is not enough to say: purely abstract truths. Plenty of these are simple and
grasped with good understanding without eliciting belief directly and reliably. Thus consider whether the stars are
even in number. Not something we can know directly, despite its simplicity and abstractness, and despite our
understanding it perfectly well.

More promising is the restriction to truths of logic or mathematics. But this leaves out enormously many truths that
seem equally intuitive. Take any claim, about two shapes, that one is different from the other (say, the shapes of two
English words). Or take the statement, about any shape and any color, that the two are distinct. And so on.

More promising is the appeal to necessity. And here we are back with a more traditional approach, leading to a revised
account, as follows:

S intuits that p if and only if S's attraction to assent to p is explained by four things in concert: (a) that p is
simple enough, (b) that S understands it well enough, (c) that p is either a modally strong proposition
(necessarily true or necessarily false), or else a self-presenting proposition (attributing to S some current state
of consciousness), and (d) that p is true.

V
How well does our account of intuition fit the profile laid down in section III? Quite well on the whole, or so I will
argue. On our account, intuition is a conscious state of felt attraction explained through the content's being (a) either
modally strong or self-presenting, (b) understood well enough by the subject, (c) simple enough, and (d) true; and
such a conscious state can serve as a justifying rational basis for belief, ceteris paribus.

It is instructive to consider how we can tell as extensively as we can what someone else believes in given
circumstances. Based on our own perception of her situation, and the placement and orientation of her sense organs,
we can tell a lot about what she is likely to believe. Thus if I draw your attention to a square plainly visible on a
surface and ask you whether it is a triangle, or a circle, or a square, etc., I can easily predict your answer. Moreover, I
can explain why you answer as you do, assuming your sincerity, good eyesight, and understanding of the question, by
appeal to the sheer truth of the matter, to the fact that the figure is a square. If it had been a circle, or a triangle,
etc., I would have been able to predict and explain similarly mutatis mutandis. But not if it had been a chiliagon!

Somehow I am able to know ahead of time the proper bounds of this sort of explanation. I know that it works with
pentagons, and maybe as far as octagons or thereabouts; after that, most people need to stop and count. Moreover,
there is

end p 216

an immense amount that each of us knows through such direct perception of the facts, and that each of us knows that
nearly everyone else knows similarly when appropriately situated. A lot of what we know in a given situation we know
others share with us when thus situated. What accounts for this? Do we depend on some assumption of simplicity? Do
we say that others will know the things we know about the perceptible situation so long as they are properly endowed
and situated, so that their relevant competence enables them to tell the relevant truths?

No such appeal to truth is required, it may be argued, since in any particular case where we see a square figure from
a similar angle as our subject, we know through a kind of simulation what the other person can tell directly. We put
ourselves in her shoes; indeed we are already in similarly placed shoes, so we know by simulation how she is likely to
react. We extrapolate from how we would react in such a situation.

That does seem right about prediction. But how do we replace the appeal to truth in our explanation of why the
subject reacts as she does, with assent to the proposition that the figure seen is a square? We could hardly appeal to
 p p p p p y

the fact that she believes as we do when similarly situated. Even if this last enables us to predict what she believes
and says, it seems useless in explaining why she says it. She need not be trying to agree with us.

We might try a minimalist tack, saying that the subject believes the figure to be a square because the figure is a
square. No appeal to truth per se is required.

Consider, however, our ability to predict and explain what will show up on a calculator screen when certain buttons
are pushed. Or take how we can predict what an idiot-savant will say about the number of toothpicks strewn on a
surface. Here we do not predict through simulation, and it is hard to see how to avoid using the concept of truth for
such predictions and explanations.

Consider, further, a subject's plethora of beliefs concerning simple modally strong (necessarily true or necessarily
false) or self-presenting truths. Is there nothing in common to these various things that she believes? Is it just an
accident that these things are nearly all true? After all, there are equally many, comparably simple false things,
equally well understood. Just take the negations of all the simple, true things. She does not believe any of these, or
nearly any. Yet she believes all their negations, or nearly all. Can this possibly be just an accident? That seems
incredible. What recourse might we have other than an appeal to the truth specifically of the simple things that she
believes, in the fields of propositions specified?

It is also plausible that in every such case she not only gets it right; she also gets it right with proper epistemic
justification. Consider what we go on in predicting as we do. Evidently we attribute to the subject an ability to discern
what is true in a certain field with high reliability, so that her relevant beliefs will be explained in part at least through
the combined effect of three things: first, her competence in that field and her situation vis-à-vis the given
proposition; second, the simplicity of the proposition; and, third, also its truth. So, we predict that if asked whether

end p 217

1 + 1 = 2, the subject will answer in the affirmative. In so predicting, we presuppose the subject's competence in
arithmetic generally, and her understanding of this proposition in particular. However, you can of course understand
some frightfully complex propositions well beyond your ken, so your general competence plus understanding will not
alone enable us to predict your answer. In predicting the answer we presumably rely on the simplicity of that
particular proposition, that 1+1=2. But any simple enough, understood proposition will have a similarly simple
understood negation. Of these two, we systematically believe the true, which is why the truth itself figures in our
predictions and explanations.

VI
There is an important further reason for thinking that rational intuitions must be true. Consider a belief derived as a
conclusion from certain reasoning. If that reasoning is fallacious, we would presumably deny that it provides
justification for its conclusion. Perhaps it can still help give subjective justification for believing that conclusion, if for
example the subject has taken care and knows himself to be normally good at such reasoning. Nevertheless, in some
more objective sense we would deny that he is really justified in so concluding. One cannot attain epistemic
justification through fallacious reasoning.

When we work our way back through the reasoning, suppose we find a fallacious step of affirming the consequent.
Something of the following form at that point seemed intuitively right to our subject: that, necessarily, if q q) then
p. What of his justification for so assenting, for taking that step of immediate inference? Is he there justified through
immediate intuitive appeal? Well, he may be subjectively justified, in which case he would presumably retain such
justification for assenting to the eventual conclusion of that reasoning. As we have seen, nevertheless, the reasoner
must also lack some more objective justification for assenting to his conclusion. It is this more objective justification
that he also must lack for assenting to that fallacious step (or to its corresponding conditional). Fallacious intuition
cannot secure that stronger sort of justification. Only true logical intuitions could provide such justification, or so we
may reasonably conclude, since anything less would leave the same problem.

Although that is a reasonable conclusion, it is not unavoidable. Here is an alternative: namely, that the defective
intuition involved in affirming the consequent is fallacious because the subject proceeds in some avoidably defective
way. It's a fault in the subject that he has that intuition, an individual flaw, or defect. It is not just an inaccuracy. By
contrast, the false intuitions involved in deep paradoxes are not so clearly faults, individual flaws, or defects. For
example, it may be that they derive from our basic make-up, shared among humans generally, a make-up that serves
us well in an environment such as ours on the surface of our planet.

end p 218

Compare the subject who sees lines in a Mueller-Lyer pattern for the first time, with no prior knowledge of that sort of
illusion, while in fact the lines are incongruent; or compare the subject who takes a wall to be red while it is indeed
red, though it would have looked red even had it been white, because the light is red, which she has no reason to
suspect. In these cases we take her to be epistemically justified in believing as she does. Yet, the explanation for why
she so believes is not that the content of her belief is true, even though the content is in fact true. In each case she
might very easily have believed as she does, even if her belief had been false, so long as the situation had remained
misleading in the ways specified.
 p p p p p y

Here in any case is for us the more important point. Let us now suppose the belief to be false, while the situation
remains misleading in the ways specified, with the misleading arrows at the ends of the lines, or with the misleading
light shining on the surface. In that case, the subject's perceptual belief would still be epistemically justified, yet this
can no longer be explained through the truth of its content. At one level the explanation here appeals to the fact that
the subject still undergoes a sensory experience sharing its content with the belief based on it as its reason. And this
supposedly gives him epistemic justification. Can this be just a brute fact? Preferably we should try to explain it.

Consider the defective appearance involved in one's attraction to think the lines incongruent, or in one's attraction to
think the wall red. Does the subject proceed in a remediably, avoidably defective way? Well, yes, in a way; yet
remediation is not really an option on her available basis for belief. Is it a fault in the subject that she is then pulled
so powerfully to think the wall red? Is it an individual flaw, or defect? Surely not. As with the false intuitions involved
in deep paradoxes, appearances thus induced are not faults, nor individual flaws, or defects. They seem to derive
rather from our basic make-up, shared among humans generally, a make-up that serves us well on the surface of our
planet.

That is not to say that the subject will be justified in taking her appearances at face value. Moreover, even if she is so
justified, this can change with further experience: for example, once she measures the lines, or examines the light.
What spontaneously draws her assent, moreover, once she has that further experience, can also change, at least in
strength. Getting wise to the effect of the angles at the ends of the lines might reduce the spontaneous attraction to
assent. Possibly one's visual module will not be denied, however, but will have its way in and through a stubborn
visual appearance of incongruence, even if now only by bucking firm belief to the contrary.

The important question for us is how to understand the epistemic justification of belief that the lines are incongruent,
or belief that the wall is red, in the circumstances respectively specified. Might we appeal to the ability or competence
or faculty or cognitive virtue manifest in the formation of that belief? We explain the formation of true, apt beliefs in
whole fields of true propositions by appeal to such abilities seated in a subject. But these abilities need not be thought
infallible, which they would have to be if we defined them as abilities always to discern the

end p 219

true from the false in those fields. Our abilities might be highly reliable without being infallible. Our perceptual
abilities seem clearly fallible, given that circumstances can so often be misleading while about this we have no clue,
and cannot be expected to have a clue. This remains true even when the ability depends, for its proper, successful
operation, not just on pertinent circumstances, but also on the subject's mental shape at the time.

A belief formed by such an ability might still be considered epistemically justified, when the subject is in bad
circumstances or in bad shape. This suggests that our evaluation of the act (the assent, or in another context the
choice) may indirectly involve an evaluation of the agent, and of her character so manifest, abstracting from the low
quality of her internal or external situation. In the case of vision, and more specifically of congruence vision, or color
vision, it seems possible, indeed likely, that the virtues manifest in the relevant inclinations to believe, and in the
assents and beliefs themselves, should be viewed not as infallible, but as fallibly dependent for their success on
conditions of the external situation and on further features of the subject's mind. Suppose that, through no fault
attributable to the subject, things go wrong like that, the effect being a false belief. Even so, the belief might be
epistemically justified, in the following sense: that it is well-formed, rationally well-motivated, deriving as it does from
the exercise of a virtue; that is to say, from a feature of the subject's mind that guides him reliably well in normal
9
conditions.
9 However, it might be best to distinguish human justification from the superhuman justification of a Ramanujan. Human justification would derive from

human facu ties, shared by normal people.

Applying this to the case of intuition, we have a way to distinguish the blunder of the thinker who affirms the
consequent from the non-blunder of the subject with a paradox-enmeshed belief. The fallacious step, we can now say,
is unjustified, in that it depends on the shape of the subject's mind at the time in a way that was easily enough
remediable, and not epistemically excusable. It depended on the subject's inattentiveness, since no normal thinker
who is sufficiently attentive will commit that kind of blunder. By contrast, paradoxes reveal something much more like
deep-seated perceptual illusions to which the visual system itself is inevitably prone. Thus, perhaps the same rational
virtue that accounts for our prowess at telling the true from the false in the field of simple enough a priori
propositions, is inevitably prone to errors that manifest themselves in deep paradoxes. If so, if our intuitions in the
paradoxical cluster do all derive from that same rational faculty, then they can all enjoy prima facie epistemic
justification, even if one of them at least must be false. Familiar facts about deep paradoxes render that account
plausible. Intuitions constitutive of such a paradox can hardly be explained through inattentive blundering, or through
any other of the factors to which we appeal in explaining why we reason fallaciously.

Finally, the paradoxes seem in this way more persuasive indictments of our reasoning competence than any of the
sorts of mistakes that psychological inquiry

end p 220

has more recently revealed. After all, these latter sorts of mistakes are rather like individual blunders due to
inattention. These we are able to correct and avoid through better concentration, which distinguishes them from how
 p p p p p y

we go astray in traditional paradoxes. Even if the difference is a matter of degree, moreover, it is a big difference
nonetheless.

So we see how paradoxes make the infallibility of intuition questionable. And the infallibility of introspection seems no
less questionable. I may look within and think there are nine dots in my visual field where in actual fact there are
10
only eight. This would seem still to count as introspection despite the fact that it goes astray.
10
Suppose, for another interesting case, that we focus on the length of the lines in a Mueller-Lyer image. I mean the relative length of the lines as they

appear in one's visual image. Are they congruent? Are they incongruent? Neither answer seems obvious, yet there plausibly is a correct answer. If so, we

have here a way in which one might go introspectively wrong. Or take a victim shown a red-hot, smoking iron, soon after which a very cold iron is pressed

to his bare back. He screams in apparent pain. But it is not immediately obvious what to say about this case.

VII
Compare justification of a moral, prudential, or even legal sort, “practical” justification for short. It is easy to think of
practically justified choices that turn out to be unfortunate nonetheless. This is analogous to epistemically justified
beliefs that turn out to be false. How deep does this analogy run? How might a choice be unfortunate despite being
practically justified? If external circumstances beyond the subject's ken or control turn out to be unpropitious, though
the subject had no reason to suspect this, and no obligation to find out, then even if the choice has very bad
consequences, it might still manifest the best possible rational procedure. So it might be a permissible choice, despite
being unfortunate. And that is one way to be practically justified, by choosing thus permissibly.

That, however, is a weak form of normativity, mere permissibility. A choice is presumably permissible if there are no
preponderant reasons against it, and in favor of some incompatible choice. If I have no more and no less reason for
bringing p about than for bringing not-p about, then I am free to bring about either one at my option. Again, this
is a weak sort of normativity, and one with no clear analogue on the side of epistemology. Consider the analogous
epistemic principle:

If I have no more and no less reason for believing p than for believing not-p , then I am free to believe
either one at my option.

No, this is clearly false. If I have no more and no less reason for believing that the number of stars is even than for
believing that it is not, I am not epistemically free to believe either way at my option. On the contrary what I must do
in that case is to suspend judgment.

end p 221

Contrast the notion of being justified in choosing or in believing a certain way, where this now means that the attitude
or act in question is “rationally well motivated.” If an attitude is rationally well motivated then it is permissible, but
the converse is not true. If I have no more or less reason for putting my right shoe on first in the morning than for
doing the opposite, it does not follow that my option to do so is rationally well motivated. In general, then, the two
concepts—the permissible and the rationally well motivated—are distinct, even if they necessarily coincide for
epistemic normativity.

Consider again a case where the internal conditions of choice are bad, though beyond one's ken or control. These can
then be excusing conditions for a bad choice, about on a par with excusing external conditions beyond one's ken or
control. One's choice is permissible, now in a different way. Now it is permissible because the agent's choice-
determining was flawed in a way that cannot be charged against him. He is blameless. He is not to be held
responsible.

That does have its correlate for beliefs, but one of scant importance for epistemology. The fact that a belief is
excusable in this sense, through brainwashing, or drunkenness, or some other disabling internal condition, does not
make the belief justified in any sense of much importance to epistemology. This is not what we care about in
epistemology, nor does it seem even to qualify as a source of properly epistemic justification. Bearing in mind this
distinction—between mere excusability and epistemic justification proper—compare again two forms of epistemic
justification, the perceptual and the introspective. The former seems clearly to permit fallibility and falsehood. How
about the latter?

Eventually we want to understand rational intuition and intuitive justification more generally. Are they fallible?
Descartes did not think so. And there are reasons to agree, some already noted in our discussion of fallacious
reasoning. On the other side, consider the paradoxes. These plausibly give rise to examples of false justified
intuitions. Only certain special intuitions are protected against fallibility. Similarly, we can feel introspectively
confident that we host a certain attitude, with prima facie justification, even if our intuitive justification is outweighed
by evidence from our behavior that shows our confidence to be misplaced.

VIII
Justified introspective and rational intuitions can thus be false, as can perceptual intellectual seemings and even
justified perceptual beliefs. These latter can be explained by appeal to unfavorable external conditions. But how are
we to explain the former?

In what way does our fallacious reasoner fall short of justification? In what way might this be different from the
 p p p p p y

paradox-enmeshed intuition?

If we focus on rationally well-motivated choice and belief, then we focus on reason, and what is due to its operation
and what not. Compare misremembering.

end p 222

How are we to understand this? We might define retentive memory of p so that the following is a necessary
condition:

At t, the subject believes p and at t , later than t, the subject believes p because she believed p at t.

If so, then no instance of the following will count as misremembering, because none will count as a case of memory at
all (assuming no sort of memory except the retentive is in play):

At t, S believes p , and at t (t t), S believes p because she believed p at t, where p is very similar to
p .

Thus, for example, if at t you believe that your friend's phone number is 352–2792 because at t you believed it to be
352–3792, this will not count as misremembering, since it will not count as remembering at all.

By contrast, common sense allows the fallibility of memory. When our new belief about the phone number differs from
the earlier belief by only one of seven digits, this would be considered a case of misremembering.

We can thus see a way to think of rational and introspective intuition so that these are fallible, which suggests the
following modified account:

S rationally intuits that p if and only if S's attraction to assent to p is explained by a competence (an
epistemic ability or virtue) on the part of S to discriminate the true from the false reliably (enough) in some
subfield of modally strong propositions that S understands well enough, with no reliance on introspection,
perception, memory, testimony, or inference (no further reliance, anyhow, than any required for so much as
understanding the given proposition).
S introspectively intuits that p if and only if S's attraction to assent to p is explained by a competence (an
epistemic ability or virtue) on the part of S to discriminate the true from the false reliably (enough) in some
subfield of self-presenting propositions that S understands well enough, with no reliance on ratiocination,
perception, memory, testimony, or inference (no further reliance, anyhow, than any required for so much as
11
understanding the given proposition).
11
Note that I speak here of propositions, though more properly it should be propositional contents, in order to take proper account of cogito

effects, where indexicals and demonstratives must be given their due.

IX
We have arrived at a general account of intuition as a source of epistemic justification, a conceptual specification of
what it is, one that plausibly accommodates

end p 223

probative force for such intuitions. What remains to be seen is how far such intuition can take us in philosophy. To
offer positive support here incurs the appearance of vicious circularity. What might be done more effectively is to face
the objections of those who reject philosophical intuition as useless.
12
Of such objections, two recently stand out. One argues that there is a special problem of “calibration” for intuition.
12
See Robert Cummins's ‘Reflection on Reflective Equilibrium,’ in M. DePaul and W. Ramsey (eds), Rethinking Intuition (Rowman & Littlefield, 1998).

This would of course require more extensive discussion, but at first blush it is hard to see why intuition might have a
special problem of calibration, one not shared by perception, for example, or by introspection, or memory. But once it
would have to be shared if it exists at all, it is hard to credit this as a really serious problem, given how limited (i.e.
nil) seem the prospects of calibrating, say, memory without any use of memory.

The other main objection derives from alleged disagreements in philosophical intuitions, ones due in large measure to
cultural or socioeconomic or other situational differences. This sort of objection is particularly important and
13
persuasive, so I devote a separate paper to discussing it in detail.
13
That paper w ll appear in Stich and His Critics, edited by Michael Bishop and Dominic Murphy, with replies by Stich (Blackwell, 2006); a preliminary

version is at http://homepage.mac.com/ernestsosa/Menu2.html .

14
Objections and Replies
14
Here I am indebted to Michael Lynch for posing these objections in his probing commentary on my
paper at the St Andrews Truth and Realism conference.
 p p p p p y

Objection 1.

When we know or believe something intuitively we often do not introspect an intuitive attraction in addition to the
intuitive belief.

Response

A resultant attraction of sufficient intensity may just be a belief. On this view, all beliefs are resultant attractions, but
not all resultant attractions are beliefs.

Objection 2.

Why restrict all intuitions to modally-strong contents?

Response

I do not so restrict them. I am simply distinguishing a certain sort of intuition that way, and this could even just be
stipulative. An intuition is, stipulatively, of this sort only if it derives from a competence. But this competence need
not be restricted to delivering modally strong propositional contents, though it might be. Moreover, it is even left open
that more than one competence be a reliable source of the intuitions that enable us to discern modally strong truths.

end p 224

Objection 3.

Intuition cannot be calibrated without relying on intuition, and this is a worse problem for intuition than for sense
perception or other alleged faculties, since intuition seems in general less reliable than those other faculties.

Response

This depends on the identity of the competence that yields our rationally intutitive attractions generally. If it is the
same competence as is involved in our grasp of simple a priori truths generally, then it may well turn out that this
competence is more reliable, not less, than our other alleged faculties. So, differential reliability seems an implausible
basis on which to demand more of intuition. Moreover, the fact still remains that memory cannot be calibrated without
relying on memory. And so on, for the other faculties.

Objection 4.

Why introduce a mysterious competence beyond our conceptual competence? Why not say that our ability intuitively
to discern modally strong truths is simply our conceptual ability?

Response

This final point seems to me potentially the most important, so I will discuss it more fully. I assume the following: (a)
that there is an immense set of modally strong beliefs that we accept directly upon understanding them, and with no
further reliance on perception, testimony, inference, etc.; and (b) that these beliefs constitute knowledge, and indeed
a particularly high grade of knowledge. That being so, we might well ask ourselves whether we are right in all those
cases simply by accident. A bald affirmative response would seem incoherent. We might try consciously suspending
judgment on the matter, but this also seems incoherent, though not quite as bad. Our only coherent option, as I see
it, if we insist on (a) and (b), requires us to assent also to the following: (c) that it is not just by accident or luck that
we are right in that immense set of intuitive modally strong beliefs. Given (c), moreover, the question naturally arises
as to what is involved in such avoidance of accident or luck. It is here that my suggested postulation of a competence
would fit. And it seems compatible with our commitment to there being such a competence that we be unable to detail
its modus operandi from the armchair. Note that the way we proceed here follows almost to the letter Descartes's
procedure when he first postulates the clearness and distinctness of a perception as the source of its certainty, and
postulates also the (infallible) reliability of such clear and distinct perception. This he does even before he has
discerned through rational theology the divine details of its modus operandi.

Shall we say that it is just our conceptual ability that is at work? Would this give us the desired further explanation
from the armchair? Presumably our conceptual ability is exercised in every propositional attitude that we adopt, with
regard to any propositional content. For, in adopting any such attitude we must at a minimum understand the
propositional content involved, and it is our “conceptual ability” that is manifest in such understanding. Moreover, our
purely conceptual

end p 225

abilities seem also exhausted in our understanding of such propositional contents. Now, there are some propositional
contents that we do immediately accept when we consider them consciously with understanding, and there are many
that we do not accept thus immediately. On this much we all seem to agree, so where do we disagree? One further
thing that I affirm is that when we rational humans exercise our conceptual competence in consciously understanding
 p p p p p y

certain simple modally strong contents, we reliably discern the true from the false. What guarantees that the things
we must accept directly upon consciously understanding them would always or generally be true? Well, perhaps this
also is axiomatic in our theory of understanding.

However, here again it is hard to discern any real disagreement. I hold that we enjoy a competence, an ability to
discern truths reliably in a certain range of the modally strong. On this there is no disagreement. Now it is suggested
that this ability does not go beyond our conceptual ability to understand those contents. But this is not incompatible
with anything I have claimed. For all I have said, it may well be true that not only have we a competence or ability to
discern modally strong contents in a certain range (presumably ones that are simple enough, for one thing), but also
we could not possibly understand those contents if we did not have that ability, since the immediate acceptance of
those contents is constitutive of that understanding. As far as I can see, we are all in agreement on the issue where
we all make a claim. Then my ostensible opponents make a further claim, one concerning how understanding of
propositional contents is constituted. But this further claim is not one that I deny. Still, it may be suggested that I fall
short in one important respect, since I have no explanation for my postulated competence. My ostensible opponents, it
may be suggested, are by contrast able to explain that competence and to defend its attribution to us rational
humans, in a way that is denied to me. This may perhaps be true in the end, but I don't yet see it; initially, at least, I
find it less than crystal clear.

Compare the following. Suppose that a piece of snow constitutes a snowball when it is round. And suppose humans
have a competence or ability to make round pieces of snow and to avoid making unround ones. Given our ontological
theory about the constitution of snowballs, humans could not possibly exercise that ability without making snowballs.
And snowballs are essentially such that their constituent pieces of snow are round. So we know that in making a round
piece of snow we are creating something, a snowball, such that (a) the piece of snow exclusively constitutes that
snowball, and (b) the snowball is such that its exclusively constitutive piece of snow must be round. All of that seems
plausible enough, but how plausible would it be to suggest that we now have an explanation for our ability to make
round pieces of snow and avoid making unround ones?

end p 226
 p p p p p y

13 Trusting Intuitions *
Thanks to Ruth Millikan, JC Beall, Sam Wheeler, Tom Bontly, Carrie Jenkins, Daniel Nolan, Patrick
Greenough, Joel Pust, Jonathan Weinberg, Ernest Sosa, and Michael Huemer for helpful comments
and discussion.

Forget that blind ambition, learn to trust your intuition—plowing straight ahead, come what may.
Jimmy Buffett, Cowboy in the Jungle

One often hears it said that debates between realists and antirealists come down to a “clash of intuitions.” After all,
such debates often reduce to arguments over primitive logical principles like bivalence or non-contradiction; or over
subjunctive conditionals about what would or wouldn't be the case if there were no believers; or over how seriously
we must take the idea that truth involves a correspondence with the world. In each case, a remark that it is all just a
“clash of intuitions” is intended to signal that the debate is intractable, and perhaps even spurious. No more progress
can be made.

In general, appeals to intuition have been taking a beating recently. Critics like Robert Cummins, Steve Stich, and
1
Jonathan Weinberg have argued that intuition is no more an objective guide to truth than crystal-ball gazing.
1
See Cummins (1998) and Weinberg, Nichols, and Stich (2003).

If so, then the importance of debates that rely heavily on intuitions—such as debates over truth or realism—is called
into question. So it is natural for anyone engaged in the debates represented in this volume to wonder what can be
said on behalf of intuition. Sosa, in his penetrating chapter, ‘Intuitions and Truth,’ argues that there is quite a bit to
say. I am not so sure. But I do think that intuition, and our trust in it, are here to stay.

I'll organize my remarks around two questions about intuitions that Sosa considers: what are they, and are they
trustworthy sources of evidence?

I What are Intuitions?

As Sosa notes, a traditional answer to this question, and one still popular, is that intuitions are experiential states or
“intellectual seemings” (see Bealer, 1998;

end p 227

Pust, 2000). Sosa rejects this view. As he puts it, “there is no sensory experience that mediates” between the fact
that two and two make four and my belief that this is so; “nor does anything like sensory experience play that role”.
(p. 209, this volume) When I introspect I don't find any distinct intellectual seeming. I just find my belief that two and
two make four. Accordingly, he argues, it doesn't seem plausible to take intuitions as perception-like experiences.

Another view Sosa rejects is that intuitions are beliefs (see, e.g., van Inwagen 1997). I can have an intuition that the
naive comprehension axiom of set theory is true even while knowing it is not. So intuitions aren't beliefs, although
they can lead me to form beliefs. To intuit that p, he concludes, is a distinct propositional attitude from belief.

Sosa's own view is that to intuit that p is to be consciously attracted to assenting to p when (1) you understand p well
enough; (2) p is modally strong or self-presenting; (3) your attraction to judge that p does not derive from any of the
usual sources of evidence: introspection, perception, memory, testimony, or inference; and (4) your attraction to
judging that p is virtuously based (that is, based on a reliable ability to discriminate truths from falsehoods).

There is a lot to like about this view. It does not equate intuitions with seemings. It avoids objections to an earlier
2
account of Sosa's.
2 See Sosa (1998) and criticisms by Bealer (1998) and Pust (2000). Sosa's earlier minimal account, roughly, held that to intuit that p was to be inclined to

believe that p. The account has been criticized, among other things, for ignoring the supposed intuition that intuitions are conscious states.

It allows that intuitions are fallible, which is good, for they are (as our intuitions about the naive comprehension
3
axiom, or the paradoxes, show).
3 For an excellent discussion about what the fallibility of intuitions should teach us about their role in philosophical theory, see Weatherson (2003).

Thus the view can accommodate the fact that I can continue to find some proposition intuitively attractive while
knowing better than to believe it. Yet it explains how intuitions can have probative force. As Sosa correctly
emphasizes, intuitions aim at the truth, even if they sometimes fail to hit their target. But, assuming they are
virtuously based in reliable cognitive abilities, they still can be said to enjoy prima facie epistemic justification. Finally,
the account is agreeably externalist. I don't have to have any beliefs, justified or not, about my intuitions in order for
them to have probative force. It is enough that they are virtuously grounded. We might sum this by saying that, for
Sosa, intuitions are inherently trustworthy.

Despite its attractions, there are reasons to wonder whether Sosa's analysis accounts for all our intuitions about
intuitions.

As I noted above, one argument Sosa raises against what he calls the perceptual model of intuition is that
 p p p p p y

introspection doesn't reveal any separate mental state of intellectual appearance “in between” the proposition I intuit
and my belief in that proposition. My first worry is that this argument seems just as cogent against Sosa's own
suggestion that my intuitive beliefs are based on attractions. Consider

end p 228

Sosa's example: my intuition that two and two are four. When I look inward I don't find any conscious attraction to
believe this proposition, pulling me, as it were, towards its truth. Rather, what I find is simply that I believe that two
and two are four. Consequently, we might say that not everything I intuitively believe I first intuit in Sosa's sense.

One might reply that everything that one believes one must first be inclined or attracted to believe. But that seems
false. Compare: some actions I was inclined to do before I did them; others not. With some actions, such as catching a
baseball that is thrown towards my face, there is no time for forming the inclination. I just act. Similarly, there are
doubtless beliefs I form on the basis of inclinations (intuitive or otherwise) but there are many beliefs which I just
4
have spontaneously. No mental attraction, conscious or otherwise, seems needed to explain this fact.
4 For similar reasons, it would be wrong to say that believing is simply the limit case of being attracted to believe. Not all the things I believe, and believe

intuitively, are necessarily things I want to believe. I may in fact not want to believe them (when, for example, they contradict a favored theory).

Sometimes I intuitively believe without that be ief being formed on the basis of a conscious attraction to believe.

My second worry has to do with the scope of Sosa's analysis. According to his account, everything I intuit is either
self-presenting or (if true) necessarily true. Indeed, most philosophical intuitions fall into one or both of these camps.
But do all? Let's consider some examples, starting with:

(1) Causes necessitate their effects.

When I think about causation, I find myself attracted to this proposition. It seems true, and importantly, a priori true.
Moreover, my finding it plausible doesn't seem to be the result of any conscious process of inference from more basic
premisses. Neither do I detect “necessary connections” perceptually. But as Hume argued, even if it is true, it isn't a
logical or analytic truth. We can imagine worlds where effects are only constantly conjoined with their causes. (This
world may be one.)

Of course, one might argue that it is necessarily true anyway, in some deeper sense of “metaphysical” necessity, and
therefore apt for intuition in Sosa's sense. (Perhaps it is synthetic a priori.) But that would take significant argument.
A better reply would be to say that whether or not (1) is necessarily true or even true at all, it certainly presents itself
5
as necessarily true to the reflective mind.
5
This is similar to Bealer's presentation in his 1993; see also Pust (2000).

And that, perhaps, is all Sosa needs.

But is it? Consider:

(2) Physical objects continue to exist when we don't perceive them.

Again, this seems true. Again, it also seems a priori; it is not my perceptual experience on which I base my belief in it,
since clearly my experience is consistent

end p 229

with its being false. Further—and although Russell would famously disagree—I don't think I believe it on the basis of
any sort of inference to the best explanation, at least, no conscious inference of that sort. It just seems true to me.
But not necessarily true. I can easily imagine worlds that appear to be like ours but where objects wink out of
existence when we don't perceive them.

Similar questions can be raised for (more) standard examples of the contingent a priori, such as Kripke's:

(3) The standard meter-stick in Paris is one meter long.

Or perhaps less controversially:

(4) I exist.

Take this last claim. How do I know that it is true? It just seems immediately obvious: it is intuitive. I don't seem to
infer it; or need to remember it, and so on. Neither is it “self-presenting” in Sosa's sense: it is not a truth about my
6
current states of consciousness.
6
See his paper in this volume (Ch. 12). However, he suggests (see footnote 7) that “I exist” is what he calls a “cogito thought”, knowable by

introspection in a way that is similar to a simple abstract necessary truth.

Yet it clearly is contingent, and presents itself to me as such.

Finally, take even general moral intuitions such as:

7
(5) It is prima facie wrong to break a promise.
7 Of course, there are also particular moral intuitions, such as: “it is wrong to break my promise to this particular person.”

This too seems obviously true to me. But does it seem necessarily true? Is it? Are there possible worlds where
 p p p p p y

promise-breaking is not wrong, other things being equal? At the very least the case is unclear.

In sum, it seems to me that not all of my non-self-presenting intuitions are necessarily true or even present
themselves as such. Doubtless, one can quarrel with my individual examples. And one might say that they aren't all
really intuitions, even if they have all been called intuitions by philosophers. I'll have more to say on this in a
moment, but here let me just make the following point: if, as Sosa and others believe, intuitions are a particular kind
of propositional attitude, then, there seems to me to be a presumption against restricting the scope of the intuitive. If
intuiting is a distinct kind of attitude, why can't we, given the right circumstances, take up that attitude towards
almost any proposition, in the way that, given the right circumstances, we can find ourselves hoping or fearing, or
believing almost any proposition? Without argument, it is difficult to see how intuition would be restricted in a more
comprehensive way than other attitudes.

end p 230

But let us put that aside and turn to a third worry about Sosa's account. Sosa's account is in danger of violating a
good commandment to live by, philosophically speaking. Namely, thou shalt not posit mysterious faculties without
necessity.

According to Sosa, I intuit that p when I am attracted to judging that p because, in part, I understand that p and my
attraction to p stems from a reliable competence to discriminate the truth for propositions of p's type. Just what sort of
competence or faculty might this be? Well, we know what sort of competence it is not: it isn't perceptual or inferential
or a matter of memory.
8
Why not say that the sort of competence that is in play here is conceptual competence?
8 See, for example, Graham and Horgan (1998).

That is, why not say that what explains our attraction to judge that p is our ability to apply concepts? To say as much
is to claim, in the canonical Gettier case for example, that it strikes us as obvious, given the description of the case,
that the subject doesn't know because of our grasp of the concept of knowledge. In such cases, the thought goes, we
intuit that p because of our inherent grasp of a relevant concept.

Interestingly, Sosa does not say this; although he comes close at points (he emphasizes the role of our understanding
of an intuited proposition, for instance). Why not? Well, one speculation is that he wants to leave room for the idea
that there are intuitions that aren't grounded in our conceptual competencies. And that seems reasonable. At the very
least, it is not consciously obvious to many people who might intuit any of propositions (2)–(5) that their finding them
attractive is the result of their grasp of the relevant proposition's component concepts. Further, some believe that
there are necessary truths that are not conceptual truths. Might not I find some of these intuitive? Consider, for
example:

(6) Anything that has determinate size also has determinate shape.

This seems true and necessarily true; but is it a conceptual truth? Likewise for (1) above. At the very least it is hard
to say. And that it isn't easy to say is not too surprising, given Quine's well-known misgivings about drawing a sharp
line between those propositions true in virtue of their concepts, and those that are not.

Yet if my intuiting that p isn't a matter of my being attracted to p on the basis of my conceptual competence, or my
perceptual, inferential competencies and so on, then what virtue or competence does it derive from? Worry: we are
drifting towards treating philosophical intuitions as being the product of a mysteriously general and reliable cognitive
competence. We are coming dangerously close to saying: intuitions are what I get from Intuition—the Natural Light.
9
But that light is neither natural nor illuminating; there darkness lies.
9 Here are two quick reasons for thinking so. First, when it comes to our perceptual faculties, or memory, or inference, we have good scientific evidence

that our brain realizes such faculties. This is a marked contrast with the case of intuition. As I ask below, how, except by using intuition, do we know that

we have a competency or reliable faculty of intuition? Moreover, our above reflections indicate that what philosophers call intuitions come in a variety of

forms. How likely is it that just one competency underlies them all?

end p 231

II The Marks of Intuition

Part of what is at issue here is what counts as an intuition. This is a rather tricky question. For one, in answering it,
we presumably rely on intuitions; and as I'll remark below, that may seem troubling all by itself. For another, since
what we are interested in is what counts as a philosophical intuition, we are also presumably appealing, at least in
part, to how philosophers use the word “intuition” when supporting or criticizing a theory.

This may suggest to some that the issue is merely semantic. It isn't. Knowing whether something counts as an X is
important precisely when something hangs on debates involving Xs. And of course there is an important debate going
on about intuition, namely a debate over the validity of a certain sort of philosophical methodology. What is as stake
is whether appeals to intuition—in counterexamples, or supporting a theory—are trustworthy. And in order to settle
that debate, it would be helpful to know what counts as an intuition.

I think we can get reasonably clear about what counts as an intuition without having to give anything like a
High-Church Conceptual Analysis. Which is a good thing, for “intuition” is not the sort of term that is likely to be
 p p p p p y

susceptible to such analyses. One method—there are others, obviously—is to appeal to common marks displayed by
the things that we typically call intuitions. These marks or platitudes can serve to jointly constrain our understanding
10
of intuition sufficiently to determine what, in at least a wide range of cases, counts as an intuition.
10 Recent employments of the same strategy include Wright (1992) and Lynch (2001; 2004) on truth, and Greenough (2003) on vagueness.

Abstracting from Sosa's discussion and our above criticism, it seems that we should be able to grant that at least two
important marks of intuitions are these.

Believability: Intuitions are propositions that the subject either believes or is inclined to believe.
Source-opacity: The subject does not know the causal processes that make the relevant proposition believable
for her; in particular, its believability does not appear to be the result of the usual sources of evidence, e.g.
perception, memory, inference and the like.

Harking back to an earlier account of Sosa's (Sosa, 1998) we can call the conception that results from taking these
two principles as definitive marks of what

end p 232

counts as an intuition the minimalist conception of intuition. It has several distinguishing features.

First, we don't need to appeal to a propositional attitude other than belief. Believability incorporates Sosa's point that
I don't actually believe everything I intuit. Nonetheless, everything I intuit is at least believable. Thus intuition is
11
explained in terms of belief.
11 I was helped to see this point by JC Beall.

We can then define the mental act of intuiting as well: I intuit when I find some proposition believable where its
believability is source-opaque. More generally, intuition is source-opaque believability.

Second, the account is liberal: it doesn't restrict intuitions to those propositions that are necessary or self-presenting;
hence it accords with our own intuitions about intuitions adduced above. But it is not too liberal: many propositions
that I find believable are not intuitions. For example, I find it very believable that George Bush is not extremely
intellectually gifted. But I know why I find this believable: from observing his speeches and press conferences.
Likewise, I find it believable that there is a book on my desk, or that I had cereal for breakfast this morning. In both
cases, I understand why I find these propositions believable—I know the sources of their believability: perception and
memory respectfully.

Third, source-opacity leaves the nature of the source of any particular intuition open. Source-opacity requires only
that the subject be unaware of the source of the intuition's believability; it remains mute about the actual nature of
the source. That source might in fact be one of the usual suspects after all. Perceptual mechanisms I am unaware of,
for example, might cause me to find certain propositions believable (as blindsight patients attest). Or the cause may
be something else altogether: enculturation, neural hardwiring or the Natural Light. Further theoretical investigation
may reveal the source: we may discover the mechanism that makes the proposition so believable. If so, then we face
a semantic decision. If, for example, I continue to lack introspective access to why I find a particular proposition
believable, even while having theoretical knowledge of that fact, I may find it easiest to say that I have come to know
the cause of my intuition. But it is possible that I become so introspectively familiar with why I find certain
propositions believable that what I once regarded as an intuition I now regard as a perception, or a memory or an
unfortunate prejudice. The minimalist account does not announce in advance which, if either, of these verdicts may be
appropriate; it depends on the particular case.

Fourth, the account allows, plausibly, for subject-relativity. Introspective awareness and self-knowledge are not
uniform. What is source-opaque believable for one person may not be for another. You may intuit what I do not.

end p 233

Fifth, the account is at least consistent with naturalism; it requires the positing of no new mysterious faculties.

And sixth: together, the two marks gel with a pre-theoretical intuition many people, including philosophers, have
about intuitions. Namely, my intuitions are those propositions that I find believable without knowing why. We just find
them believable as it were. That is similar to what our first two marks tell us about intuitions.

The minimalist conception of intuition may seem too minimal. Our two marks of intuition don't seem to distinguish
between the detective's intuition that the butler did it and Kripke's intuition about the necessity of origins. This is not
a vice but a virtue. We should take seriously the fact that we use the same word to cover both sorts of cases. And the
liberalness of the account doesn't imply that we can't distinguish between different types of intuition. Some intuitions
may have probative force and others not, for example. Moreover, we may, if we wish, distinguish semantic intuitions
from pure intuitions, where a semantic intuition is an intuition about concepts or the meanings of words.
Philosophizing may reveal that some of our intuitions are semantic intuitions in this sense. Yet we may still call them
“intuitions” since, first, their believability is not obviously due to any of the usual sources of evidence; and second, my
conceptual competency is not best understood as a distinct, monolithic causal source of evidence but as a
manifestation of a complex set of abilities.

Moreover, we may wish to say more about philosophical intuition, since what is at stake in the debate over intuitions
 p p p p p y

is their use in philosophical argument. In particular, the intuitions that concern philosophers are those that are
epistemically relevant, either positively or negatively, to some philosophical theory. Consider (1)–(5). Two facts are
12
salient about how such intuitions are used by philosophers.
12 See Williamson (2004) for similar remarks on the role intuitions play in philosophical theory.

First, philosophers who believe them use them to refute or defend some particular philosophical thesis. Yet, second,
philosophers who don't believe them still typically grant that they are believable, or intuitively appealing. They thus
try to explain the proposition's believability away, by arguing, for example, that their believability is the product of
bad theory. A rabid phenomenalist, for example, might argue that while, strictly speaking, it is false that unperceived
objects exist, it is understandable that people find the proposition so intuitive, since (they might say), talk of
unperceived objects can be translated into subjunctive conditionals concerning what people would perceive in certain
conditions. Likewise with the Humean who denies necessary causal connections, or the error theorist who denies
moral intuitions.

end p 234

Accordingly, we may add to our two marks above:

Theoretical Relevancy: Philosophical intuitions are of positive or negative epistemic theoretical relevance.

Intuitions are those propositions we find believable without directly knowing why. And philosophical intuitions are
those propositions we just find believable which play a particular role in philosophical inquiry. If something like this is
correct, then philosophical intuitions don't form a special objective kind. Their uniqueness is not the source of the
game but a product of its being played.

III Why Trust Intuition?

If intuitions are understood in the minimal way, what makes us think that intuition is reliable? An advantage of Sosa's
position is that an answer to this question drops right out of his analysis of what an intuition is. On his account, to
have an intuition just is to be attracted to assenting to a proposition on the basis of a reliable cognitive ability. This
may seem at first to be helpful in answering the skeptic. Intuition can certainly be a reliable basis for belief even if we
can't show how this is so. Consequently, in so far as intuition really is the result of a reliable faculty, its outputs have
probative force.

The minimalist about intuition can avail herself of a slightly modified version of this strategy. On the minimalist
conception, intuitions aren't the products of reliable cognitive faculties by definition. But the account leaves it open
that they might nonetheless still be the product of such faculties—even the product of more mundane reliable
faculties, like memory or perception. Consequently, the minimalist can join Sosa in saying that intuitive beliefs are
justified just when they are the result of a reliable cognitive competence or faculty.

This may seem cold comfort, however, on either account. Suppose we grant that intuitions are as Sosa says they are,
for example. The skeptic worried about the reliability of intuition will simply re-phrase her question from “Why think
intuition is reliable” to “Why should we think we actually have inherently reliable intuition?” Likewise, one can ask the
minimalist why we should think that what we find intuitively believable but source-opaque is ever produced from a
reliable faculty as opposed to, say, prejudice.

We don't need any fancy-schmancy scientific study, à la Weinberg, Stich, and so on to cause these doubts. We all
know that sometimes we accept propositions as obvious for less than virtuous reasons. Philosophers are not immune
from this certainly. We can be prejudicial; we sometimes find a proposition believable only because we are in the grips
of theory, for example. And if this sometimes happens, how do we know it is not happening now, or always? So, again,
even if intuitions are inherently trustworthy, what evidence do I have that I have any?

end p 235

Presumably, my evidence will ultimately involve, and perhaps immediately involve, other propositions I find believable
without known appeal to the usual sorts of evidence—that is, other intuitions. Intuition, in the parlance, is
epistemically circular.

To this, one might be inclined to say, “So what? So we can't demonstrate our reliance on intuition without relying on
intuition. This is true for all our basic sources of evidence, most famously, as Alston (1993) has argued, for sense-
perception. And it isn't as if we've given up on sense-perception; so we shouldn't be worried about intuition either.”

The problem with this initially sensible-sounding answer is that it seems too open to being turned around: if sense-
perception really is no better off epistemically than intuition, then sense-perception is in big trouble. Yet surely
philosophical intuition isn't as reliable as sense-perception! Two thousand years of philosophical debate show that.
Accordingly, doubts about intuition, and a desire for a non-circular demonstration of its reliability carry more force
13
here than when it comes to sense-perception.
13
See Pust's illuminating discussion of epistemic circularity and intuitions in his 2000:119 ff. Pust holds that intuitions aren't epistemically worse off with

regard to circularity than sense-perception or any other doxastic faculty. One may wonder, as he admits, whether this is supposed to be comforting or

damning. Moreover, there does seem to be this relevant difference: intuition has been challenged as a source of evidence; the challenges that face

perception are made merely for the sake of argument.

 p p p p p y

Where does that leave us? Back where we started in one sense. But I don't think this is worrying, and neither should
you. As Sosa wisely says at one point, and as my argument just above implicitly demonstrates, “It is hard to avoid
appeal to direct intuition sooner or later”. Indeed, more than hard, practically impossible. When engaging in inquiry,
no matter how carefully, something is always being taken for granted as immediately obvious. As Sosa puts it, “there
is an immense amount that each of us knows through direct perception of the facts, and that each of us knows that
nearly everyone else knows similarly when appropriately situated”.

If so, then it seems to me that even if I can't demonstrate that what we take as obviously believable in this way is
reliably based, I may still be acting responsibly, from the rational point of view in forming beliefs on that basis. The
thought is that some intuitive beliefs can be trustworthy, and importantly, I can know them to be, even though—as
far as I or anyone else knows—I haven't done anything positive to earn them that status. The epistemic status I have
14
in mind here is now commonly called entitlement.
14
See Crispin Wright (1985, 2004); Hazlett (forthcoming). Compare Burge (1993).

At a minimum, we might say that I am entitled to intuitively believe that p only if: (a) I find p believable; (b) its
believability is not knowably derived from the usual sources of evidence; and (c) its believability has no strong
defeaters, where any reason adequate to strongly defeat my belief that p

end p 236

15
is a reason to believe that not-p.
15 Thus one may be entitled to intuitively believe that p even if there is no more or less evidence to believe p than not-p.

What more there is to say, in particular about what makes it sufficient for me to be entitled to intuitively believe, is
complicated by the fact that intuitive entitlement, like intuitions themselves, will doubtless come in more than one
form. I am content here to simply suggest, without elaboration, the following two possibilities.

First, one might claim that we are entitled to believe some intuitive propositions just in virtue of our understanding
them. Barring defeaters, we are entitled to believe intuitions of this sort simply because of our cognitive grasp of their
16
content. Their positive epistemic status, in other words, is due to their being “low-grade” or epistemically analytic.
16 See Boghossian (1997) for a defense of the epistemic view of analyticity.

Under the assumption that our conceptual abilities are not properly sorted with the usual sources of evidence, or even
as a source of evidence at all, properly so-called, their believability remains source-opaque. Thus, on the minimalist
conception, they are intuitions—but intuitions that wear their epistemic status on their sleeve, so to speak.

Second, some intuitions might be epistemic hinges. Crispin Wright (1985, 2004), drawing on Wittgenstein's (1969)
work, has pointed out that we might be entitled to some beliefs simply because they act as the hinges on which the
rest of our knowledge turns. Consider (2), namely that physical objects continue to exist when unperceived. Such a
belief not only seems to “transcend” evidence in a peculiar way, but if we did not believe it we wouldn't be in a
position to reasonably believe large swaths of other propositions we take ourselves to know. For, if it is false, then
arguably our general standards of evidence might be called into question. It is at least arguable that we are entitled to
believe many philosophically important intuitions in precisely this sense. And, if so, then we may come to know that
we are so entitled, and thereby provide a defense of intuition.

Admittedly, epistemic entitlement might not rise to the level of true justification—at least in the high-octane sense
usually demanded by the skeptic. But given our worries about circularity, it may have to be good enough. If it is, then
we needn't worry about our appeals to intuition. Even at our most skeptical, we can't question everything at once. We
need to start somewhere, to keep something fixed, in order to get our doubts up and running. If what is fixed is what
in other moods we call “intuition” then Sosa is right: intuitions are here to stay.

Or so it seems to me.

References

Alston, W. P. 1993. The Reliability of Sense Perception. Ithaca: Cornell University Press.

Bealer, G. 1993. ‘The Incoherence of Empiricism’. In S. Wagner and R. Warner (eds), Naturalism: A Critical Appraisal.
Notre Dame: University of Notre Dame Press, 163–96.

end p 237

Bealer, G. 1998. ‘Intuition and the Autonomy of Philosophy’. In DePaul and Ramsey, 201–41.

Boghossian, P. 1997. ‘Analyticity’. In B. Hale and C. Wright (eds), A Companion to the Philosophy of Language.
Oxford: Blackwell: 331–68.

Burge, T. 1993. ‘Content Preservation’. Philosophical Review: 102: 457–88.

Cummins, R. 1998. ‘Reflection on Reflective Equilibrium’. In DePaul and Ramsey: 113–28.

DePaul, M. and Ramsey, W. 1998. Rethinking Intuition. Lanham, MD: Rowman & Littlefield.

Graham, G. and Horgan, T. 1998. ‘Southern Fundamentalism and the End of Philosophy’. In DePaul and Ramsey:
 p p p p p y

271–92.

Greenough, P. 2003. ‘Vagueness: A Minimal Theory’. Mind: 112/446.

Hazlett, A. forthcoming. ‘How to Defeat Belief in the External World’. Pacific Philosophical Quarterly.

Lynch, M. P. 2001. ‘A Functionalist Theory of Truth’. In The Nature of Truth. Cambridge: MIT Press: 723–750.

Lynch, M. P. 2004. ‘Truth and Multiple Realizability’. Australasian Journal of Philosophy, 82: 384–408.

Pust, J. 2000. Intuitions as Evidence. New York: Garland.

Sosa, E. 1998. ‘Minimal Intuition’. In DePaul and Ramsey: 257–70.

Sosa, E. 2006. ‘Intuitions and Truth’. This volume, Ch. 12.

van Inwagen, P. 1997. ‘Materialism and the Psychological-Continuity Account of Personal Identity’. In J. Tomberlin
(ed), Philosophical Perspectives 11: Mind, Causation and World. Oxford: Blackwell: 305–19.

Weatherson, B. 2003. ‘What Good are Counterexamples?’ Philosophical Studies, 115: 1–31.

Weinberg, J., Nichols, S., and Stich, S. 2003. ‘Normativity and Epistemic Intuitions’. Philosophical Topics, 29: 429–61.

Williamson, T. 2004. ‘Philosophical Intuitions and Scepticism about Judgement’. Dialectica, 58: 109–53.

Wittgenstein, L. 1969. On Certainty. Ed. G. E. M. Anscombe and G. H. von Wright. New York: Harper and Row.

Wright, C. 1985. ‘Facts and Certainty’. Proceedings of the British Academy, 71: 429–72.

Wright, C. 1992. Truth and Objectivity. Cambridge, MA: Harvard University Press.

Wright, C. 2004. ‘Warrant for Nothing (And Foundations for Free?)’. Aristotleian Society Supplement, 78: 167–212.

end p 238
 p p p p p y

14 Truth and Realism: Remarks at St Andrews

Michael Devitt claims that we can be scientific realists without troubling our heads about the nature of truth or of
reference, and without getting into epistemological issues. 1
1
All specific references, unless otherwise noted, are to the essays in this volume.

For we can just take what scientists say about unobservables at face value. Christopher Gauker and Michael Williams
disagree. Gauker argues—convincingly, to my mind—that to figure out whether a working physicist like J. J. Thomson
is a scientific realist it is not enough to know what he says and does in his capacity as a scientist. One also has to quiz
him about his views on the possibility of representing mind-independent objects. This requires explaining to him what
the terms “mind-independent” and “mind-dependent” mean, which is no easy task. We also need to explain that we
are not using “represent” in the familiar pictorial sense, but in a novel sense in which assertions of absence or of
non-existence can map reality just as do those of presence or of existence. We have to wheedle Thomson out of the
routine of his day job, and nudge him over into what Michael Williams calls a specifically epistemological context.
Outside that context, Thomson can get along nicely with Arthur Fine's “the Natural Ontological Attitude.”

Devitt finds that attitude “mysterious,” but I take Fine to be saying simply that one can disbelieve in witches and
believe in electrons without taking sides on any philosophical issue. Scientists, on Fine's account, do not have
ontological commitments, only philosophers do. For “ontological commitment” and “really exists” are part of the same
cluster of concepts that includes “linguistic representation,” “fact of the matter,” “correspondence to reality,” and
“truth-making.” The function of such concepts is to enable philosophers to make invidious distinctions between uses of
language, distinctions that make no difference to practice and in the application of which non-philosophers take little
interest.

Terry Horgan says that “one way to stop worrying about which referential commitments of thought and language are
genuine ontological commitments is to claim that none of them are, and that the idea of a mind-independent,
language-independent world is a metaphysical extravagance.” Horgan takes this claim to be unintelligible, just as
Devitt takes Fine's attitude to be mysterious.

end p 239

But philosophers who have no use for such notions as ontological commitment should not be thought of as
fictionalists. They are not favoring mind-dependence over mind-independence. They regard these paired notions as
serving no purpose save to continue a sterile debate.

Horgan says that his version of realism has the advantage of allowing us to “respect the persistent belief that
numerous ordinary beliefs are literally true.” But the sort of people whom Williams calls “neo-pragmatists” think that
“literally true” is an expression that we would do well to abjure, and that we should also stop using terms such as
“really exists” and “robustly real.” Neo-pragmatists tend to agree with Austin that the contrast between the real and
the non-real is useful only when drawn within some particular context—as when we distinguish between real cream
and non-dairy creamer, or between historical figures and characters in novels, or between real poems and doggerel.
We have disciplines that devote themselves to making these retail distinctions. But we do not have, and do not need,
a discipline that is in the business of distinguishing reality in general from appearance in general. We do not need
what Horgan calls “serious ontological inquiry.”

Fine answers the question “Do you believe in X?”, for such X's as electrons and dinosaurs, by saying “I take the
question of belief to be whether to accept the entities or instead to question the science that backs them up.” On his
view, science says all there is to say about electrons. Philosophy has nothing to add, because no useful purpose is
served by worrying about ontological status. Stated more generally, Fine's point is that if we know how people use
words, and under what circumstances they would be likely to use them differently, then we cannot add to our
knowledge of what they are up to by asking them what they take to be real. Posing that question will not clarify their
previous activities. Rather, it will invite them to join in a new, specifically philosophical, activity.

Another way to put this point is to say that, outside of philosophy, the only word-world relations we need are
aboutness and causality. We need to be able to figure out, for example, whether when I speak of freedom fighters I
am talking about the people you call “terrorists.” We also need to know whether our simultaneous exclamations were
both caused by the same snake emerging from the same clump of grass. But for non-philosophical purposes we need
neither “reference” nor “truth-making,” nor “representation” nor “ontological commitment.” These terms come to
seem useful only when we feel the need to talk about language, knowledge, and world in a wholesale rather than a
retail way.

The first prominent figure to feel this need was Parmenides. He rolled the grass, the snakes, the gods, and the stars
into a single well-rounded blob, stood back from it, and proclaimed it unknowable. Plato's admiration for this
imaginative feat led him to coin the term “really real.” Ever since Plato's day, people have been making the sort of
distinctions Paul Horwich satirizes, those between the merely real and the robustly REAL and between queer facts and
non-queer facts. But before Parmenides there was only what Crispin Wright calls “the bland

end p 240

perspective of a variety of assertoric ‘language games’, each governed by its own internal standards of acceptability,
 p p p p p y

each sustaining a metaphysically emasculated version of truth, each unqualified for anything of more interest or
importance.”

In that pre-pre-Socratic golden age, nobody felt metaphysically emasculated, for phallogocentrism had not yet been
invented. There were no metaphysical quietists because there were no activists. Before the expression ontos on
caught on, nobody felt the need for a wholesale contrast between episteme and doxa, as distinct from retail contrasts
between reliable experts and unreliable amateurs. Everybody was a contextualist in Michael Williams’ sense, because
nobody talked about the need to justify “our theory of the world.” Nobody had such a theory, although they did have
theories about, for example, trees, gods, snakes, and stars.

One way to describe Parmenides’ ill-fated flight of fancy is to say that he dreamt up something to which you cannot
apply Davidson's argument that most of our beliefs about an object must be true. Davidson argued that if you do not
have mostly true beliefs about X, then you are not talking about X, and therefore have no beliefs about X. This
argument works for things like numbers, electrons, and poems, but it does not work for “reality” or “the world” or
“things as they really are.”

Parmenidean entities of this sort—the kind that Hilary Putnam calls “superthings”—are distinguished by the fact that
disagreement about them does not take place against a background of massive agreement. You will not be thought to
be talking about trees unless you accept the usual commonplaces about trees, but there are no such commonplaces
about reality. You can go on and on about the nature of reality even if what you say about it has no overlap with what
anybody else says about it. This is why wholesale talk achieves a peculiar pathos of distance—one that cannot be had
at the retail level. This pathos is preserved by those who reply to Davidson: “Your argument works for trees, numbers,
electrons and poems, but since none of these things may REALLY exist—since they might all be mere appearance—it
remains possible that most of our beliefs may yet turn out to be false, just as Parmenides suspected.”

Ascent to the wholesale level is the only way to make intelligible questions like “Are there really numbers?”, “Are
trees really real?”, and “Are electrons as real as tables?” To make them intelligible, however, is also to make them
irresolvable. For it is definatory of the really real that there is no agreed-upon way to tell which statements about it
are true.

To go wholesale is to assume, as Robert Brandom has put it, that “thing” or “real thing” or “self-identical thing” or
“entity” is the name of a natural kind. Brandom has the same kind of doubts about what he calls “pseudo-sortals” as
Williams has about what he calls “a certain hypergenerality” and that Horwich has about what he calls “the scientistic
tendency to over-generalize.” All three of these philosophers share doubts about whether wholesale talk can be made
continuous with retail talk. They dispute the claim that the wholesalers are merely

end p 241

taking the words of the retailers at face value. They think that common sense can be reflective without waxing
ontological.

Conversation about trees, numbers, electrons, and poems is governed by social norms, observance of which marks off
the people who can talk sensibly about those sorts of things from those who cannot. But there are no such norms for
discussion of questions about what really exists. You can say almost anything about the nature of reality—that it is a
blob, that it is atoms and void, that it is the totality of facts rather than of things, that it is organisms and
simples—and get away with it. Ontology is more like a playground than like a discipline.

Issues about realism and anti-realism only arise when the process by which objects come to be talked about is treated
as if it were a species of the same genus as the causal processes by which spatio-temporal objects come into
existence. When subjected to this treatment, notions like “constituted by” or “socially constructed by” are interpreted
as suggestions that there were no trees or electrons before we started talking about them. But if we avoid interpreting
phrases like “constituted by the way we talk” as quasi-causal explanations of where things came from, we shall say
that Sherlock Holmes was brought into existence by his parents, not by Conan Doyle's prose. This is quite compatible
with saying that we should not have been able to talk about Holmes were it not for Conan Doyle. Analogously,
dinosaurs were brought into existence by non-mental and non-linguistic causes, even though we should not be talking
about them had paleontologists not used their imaginations. Redness and blueness were brought into existence by the
Big Bang, but we could not talk about them until early hominids, displaying great imaginative creativity, had dreamed
up words such as “red” and “blue.” Mathematical entities and moral virtues, lacking spatio-temporal locations, were
not brought into existence at all, and certainly not by the use of mathematical or moral language. But imaginative
mathematicians and moralists did enable us to talk about them. Only if we run together questions about how we came
to talk about something with questions about how its causes arose will we want to ask philosophical questions about
what is mind-dependent and what is not.

If we abandon Parmenides’ unfortunate attempt to distinguish, in a wholesale way, between real things and apparent
things, we shall say that although there are plenty of fictional characters, mythical creatures, fake Rolexes, and
obsolete theoretical posits, there are no merely apparent objects. Myth vs. history is a useful retail contrast, as is the
contrast between possible geometrical figures and impossible ones. But all objects—from rocks to round squares—are
accessed in the same way, by learning how to talk about them.

Parmenides and Plato went wholesale because they were obsessed by the idea of non-discursive access to truth.
Empiricists think that such access is provided in sensory perceptions. Ernie Sosa, like Plato, thinks that it is provided
 p p p p p y

in contemplation of simple mathematical propositions. Only such direct access, such thinkers believe, could provide
knowledge that no further discourse could place

end p 242

in doubt. The idea that such certainty might be possible led Plato to contrast the indefeasible reality of purportedly
directly accessible objects with the dubious reality of objects that are accessible only discursively. But the notion of
“access” is itself a conflation of causal impact with the acquisition of information. The latter is always indirect, because
it can occur only when there has been a lot of stage-setting in the language.

I shall return to this Wittgensteinian-Sellarsian thesis at the end of my comments, but now I want to turn from the
topic of ontology to that of semantic theory. If one regards ontology as hopeless and notions such as “linguistic
representation” as useless, one will be dubious about Michael Dummett's suggestion that old arguments about realism
vs. idealism can be spruced up by taking the linguistic turn and formulating alternative theories of meaning.

The idea of a semantic theory for a language started out as the thought that both language and non-language can be
cut up into bits which can then be paired off. This is sometimes possible in areas of inquiry in which ostention is
possible, but it has never been clear how to do it elsewhere. In this exigency, philosophers of language have invented
the notion of “being made true by”—thought of as a relation between a bit of non-language and a bit of language that
resembles ostention in respect of non-discursivity. But nobody has ever explained what kind of non-causal relation
“being made true by the world” is supposed to denote. It resembles “word-world fit,” or “word-world correspondence,”
in being an unanalyzable and inexplicable primitive—of the sort that Davidson argued added nothing to the simple
notion of “true.” Philosophers who construct theories of meaning feel free to postulate novel entities to serve as terms
for these mysterious relations. But, as with ontology, it is not clear what the rules of the game are supposed to be.

It seems to me that Dummett's attempt to contrast his own anti-realism with Davidson's putative realism was
unfortunate, and led nowhere. Davidson himself had no use for the notion of corresponding to reality, nor for that of
beliefs being “made true by” non-beliefs. His commitment to holism led him to drop the idea that we were ever going
to do any useful pairing of bits of non-language with bits of language. He tried to get people to stop calling him a
“realist,” though with little success. He wound up saying that there was no interesting difference between knowing a
language and knowing one's way around the world generally.

If one carries through on that line of thought, one will no longer say, as Dummett once did, that philosophy of
language is first philosophy. That claim seemed plausible only when it seemed philosophically profitable to do what
Horgan calls “paraphrasing claims containing referential commitments that seem . . . ontologically dubious into claims
whose ontological commitments seem less dubious.” But most philosophers now agree with Horgan that it is not, if
only because there are no rules for accuracy of paraphrase. Nobody now tries to use a semantic theory for purposes of
philosophical therapy (of the sort that Russell administered to

end p 243

Meinong, and Carnap to Heidegger). So it is not clear what we would do with a semantic theory if we had one—what
relevance it would have to any other area of philosophy.

The idea that both non-language and language can be split up into bite-size chunks is essential to keeping a
distinction between semantics and pragmatics in place, and thus to the hope of constructing a theory of meaning for a
language. Radical holists like Brandom give up on this distinction. For Brandom, the content of the assertion made by
the user of a sentence is different every time that sentence is used. As Michael Williams has remarked, that amounts
to giving up the attempt to divide the use of an expression into an essential core, called the “meaning” of the
sentence, and the accidental features that lie outside this core. This is equivalent to abandoning the attempt to find
what Horgan calls “identity-preserving differences” that will ensure, for example, that “the word ‘corporation’ has the
same meaning when employed under different semantic standards in different contexts.”

Brandom is an example of a philosopher of language who has no interest in isolating anything that could be called
“semantic structure.” In this respect, he is following the lead of Davidson's “A nice derangement of epitaphs.” By the
time Davidson wrote that article, he had ceased to believe in the possibility of explaining interpretive know-how by
reference to an understanding of underlying structures. 2
2 Donald Davidson, ‘A Nice Derangement of Epitaphs,’ in LePore, Ernest (ed.), 1986, Truth and Interpretation: Perspectives on the Philosophy of Donald

Davidson, Oxford: Basil Blackwell.

He had come to agree with the point that Horgan makes against Fodor: that learnability does not require
compositionality. Davidson had become content to say that a Tarskian truth-theory for a language would capture what
the successful interpreter knew, while cheerfully admitting that nobody would ever formulate such a theory, much
less be able to use it to resolve philosophical controversies.

If we want a linguisticized reformulation of the old debate between realism and idealism, we would do better to see
that debate as reincarnated as a dispute between John McDowell and Brandom than as a supposed disagreement
between Davidson and Dummett. The principal issue between them concerns McDowell's “realist” claim that we are
answerable to the world. I interpret Brandom as saying that if it does not talk, we are not answerable to it—that we
need answer only to other language-users. That, on his quasi-idealist view, is what was true in Hegel's claim that
Spirit is never presented with anything other than itself (‘Ein durchaus Anderes ist fuer den Geist gar nicht
 p p p p p y

vorhanden’).

In a recent paper, McDowell seems to be saying that it is impossible for Spirit to be this self-contained, because
inference is as inconceivable without representation as is representation without inference. 3
3
McDowell's paper was presented at the Truth and Realism Conference at the University of St Andrews and is forthcoming in the Library of Living

Philosophers volume on Michael Dummett.

The core of his argument is that

end p 244

Brandom, even more egregiously than Dummett, fails to realize that “representationalist conceptions of meaning”
need not be burdened by the reductionist claim that they can explain meaning “in terms supposedly intelligible
independently of the concepts we need for describing the rationally structured context in which we place linguistic
performances when we understand them.” But suppose we disburden them, and eschew any claims about what is
intelligible independently of what. I still cannot see what representationalist conceptions are supposed to be good for,
at least after we have given up attempts to go wholesale by doing ontology.

I am dubious about McDowell's attempt to milk the notion of representation out of Aristotle's off-hand little truism.
Suppose we grant to McDowell that “there is no reason to suppose that something recognizable as an idea of
expressing the thought that things are thus and so could be intelligible in advance of the idea that if things are indeed
thus and so, such a performance constitutes speaking the truth.” I do not see how to get from this to the further claim
his criticism of Brandom would seem to presuppose—namely, that I couldn't think of myself as expressing thoughts if I
weren't thinking of myself as representing something, and as somehow answerable to it.

I suspect that in the golden age before Parmenides and Plato, people managed this trick nicely. That was because they
did not know either that there was such a thing as “the world” or that there was something called “mind” or
“language” or “the subject” or “human cognitive faculties” to set over against the world. A fortiori, they did not think
of the one as being represented by, or representing, the other.

I am loath to agree with McDowell's final conclusion that “competitions between representational and broadly
inferential conceptions of meaning are misconceived from the start.” That seems to me to miss the revolutionary
potential of Brandom's project. Brandomian inferentialism gives us a way of seeing the notion of “object” as emerging
from the use of singular terms, and of seeing that use as emerging from the need to make de re ascriptions. Brandom
has given us an alternative to the traditional picture of singular terms as names of non-discursively accessible objects.

McDowell's notion of a conceptually structured faculty of receptivity still retains enough of this picture to leave room
for bad ontological questions about what objects are out there, striving to make an impression on our wax tablets.
Brandom leaves no room for them. The competition between those two ways of conceiving of singular terms is
important because if we do things Brandom's way we shall treat a decision to cease speaking of certain things
—phlogiston or Zeus or racial purity—not as a decision about what an entity called “the world” is really like but as a
decision about how to achieve local, concrete, retail aims.

Brandom's repudiation of the idea that “entity” or “real thing” names a natural kind links up nicely with Williams'
repudiation of the idea that there are natural kinds called “justification” or “knowledge” or “inquiry” or “our theory

end p 245

of the world”—kinds whose relation to the natural kind called “real thing” are to be investigated by ontology and
epistemology.

The view that I am suggesting is as quietistic and metaphysically emasculated as Horwich's. The only significant
difference between his position and mine, perhaps, is that he puts most of the blame for the persistence of ontology
on followers of Democritus—philosophers who take physical facts to have a desirable feature called “ontological
weight” that other facts lack. But as I see it, the obsession with facts about material corpuscles (the one shared by
Thomas Hobbes and Frank Jackson) is just one example of the bad effects of going wholesale. Plato's conviction that
facts about spatio-temporal relations are insufficiently robust, and his consequent obsession with abstract objects, is
an equally good example. The blame rests with Parmenides for having seduced both Democritus and Plato into
thinking that it is important to discover which things are really real.

I have one other disagreement with Horwich. I would not say, as he does, that the tradition that Parmenides
inaugurated has landed us with incoherent concepts such as “weird fact,” “queer fact,” “robust fact,” and the like. Any
expression, even those used by philosophers, has a coherent use if you give it one. Ability to continue a conversation,
and to tell a plausible story about how that conversation has gone so far, is quite enough to show coherence.

But you can be coherent without being useful. The scholastic philosophy of the fifteenth century was a coherent
conversation, yet many of us are glad that it eventually withered away. I should like to think that by 2100
philosophers will no longer be interested in the issues that were discussed in 1900 under the rubric “realism vs.
idealism” and in 2004 under the rubric “realism vs. anti-realism.” In the course of the twenty-first century, interest in
the debate between McDowell and Brandom, a debate I am treating as continuous with the one between Russell and
Bradley, may wither away.

As I see it, the shift in terminology from “idealism” to “anti-realism” as the term that contrasts with “realism”
 p p p p p y

signalizes the one great accomplishment of twentieth-century philosophy. In the course of that century, philosophers
broke the link between the idealists' repudiation of the correspondence theory of truth and their claim that reality was
spiritual in nature. The struggle between holism and atomism thus got detached from the struggle between the gods
and the giants—between philosophers who agree with Plato and Berkeley that the really real is more like a mind than
like a lump of dirt and those who agree with Democritus, Hobbes, and Jackson that dirt enjoys ontological priority. In
1900 there were still many people who were on Plato's side in this debate, but in 2000 none were left.

This desirable change was the result of assaults on the Platonic notion of direct, non-discursive access to the real that
were launched by Wittgenstein, Ryle, Sellars, Putnam, Davidson, Wittgenstein, Heidegger, Gadamer, Derrida, and
many other twentieth-century philosophers. These philosophers did their best to get rid of notions like “intuition”—the
one that Ernie Sosa would like to rehabilitate.

end p 246

Reiterating Peirce's criticisms of this notion, they rejected the possibility of what Sosa describes as “conscious states
with probative force” and of propositions such that any rational mind considering them would be “directly attracted” by
them. Unlike Sosa, they are content to think that to know that something is square is to know that the word “square”
applies to it.

On my reading of the history of analytic philosophy, what Gustav Bergmann called “the linguistic turn” cleared the
way for what Sellars called “psychological nominalism”—the doctrine that all awareness that is not merely
discriminative behavioral response is a linguistic affair. This in turn made it plausible to explicate our use of “true” in
terms of “norm-governed assertion,” in the manner of Davidson and Brandom, as opposed to treating “truth” as the
name of a property that explains the attraction that certain propositions exert on rational minds, in the manner of
Sosa.

The upshot of the linguistic turn was thus to get rid of the idea that the mind was some sort of immaterial substance,
and thus of any motive for the various naturalizing projects on which many contemporary philosophers are still
engaged. The claim made by the idealists of 1900 that reality is more like thoughts than like dirt is no longer taken
seriously. The leading Hegelians of our day—Pippin, Pinkard, and Brandom—differ from Royce, Joachim, and Blanshard
in taking Berkelian immaterialism to be entirely irrelevant to Hegel's project.

It was unfortunate that the demise of metaphysical idealism encouraged a revival of metaphysical materialism, in the
form of claims that the supervenience of the behavior of everything else on the behavior of elementary particles
shows that the latter are somehow ontologically or semantically privileged. This revival has resulted in the
divinization of physical facts that Horwich deplores, as well as to philosophers' insistence, mocked by Fine, that
natural science is somehow special.

The linguistic turn, when combined with a continuing attempt to go wholesale, produced the substitution of “language”
for “mind” as the name of a directly knowable superthing standing over against an indirectly knowable superthing
called “reality.” Davidson's claim that there is no such thing as a language amounts to the suggestion that we should
retreat to the retail level, where there are no superthings. Retreat to that level is, I suspect, the only way to bring
about the disappearance of the various “isms” whose prevalence Horwich deplores.

end p 247

end p 248

Index

abduction 107–10 , 132–3

aboutness 80

acceptance 26–7

affirmability 6 , 140–1

alethic realism 2

alethic relativism 2

Alston, W. P. 236

analetheism 67–8

anti-realism 79 , 86–7 , 119 , 193–5

idealism and 246

instrumentalist 86 n. 14

and realism 179–81

reductionist 86 n. 14

approximate truth 106 n. 12

 p p p p p y

Aristotle 2

assertibility, idealized 56–9

assertibility-conditional semantics 180 , 181

atomic facts 70–1

attractions 210 , 224

intuitions and 228–9

attributor contextualism 59 n.

austere indirect-correspondence (AIC) realism 138–43 , 145 , 155 , 160

austere metaphysical realism 5–6 , 137

austere ontology 138 , 139–43

common sense and 145–55

Austin, John 94 n. 30 , 240

basic facts 196

Beall, JC 61–73

beliefs 26–7 , 192–3 , 215 , 219–20 , 224

intuition and 208–9 , 210 , 211 , 212 , 213

benign logical incoherence 153–5 , 163 , 168–73

Bergmann, Gustav 247

Berkeley, George 79 n.

Blackburn, Simon 201–2

bloodless quietism 202 , 203

Boghossian, Paul 13–37 , 191

Boyd, Richard 108 , 109 , 115

Brandom, Robert 83 , 89 , 241–2 , 244 , 245

broad inferentialism 83

causation 229

charity, principle of 18

classical pragmatism 2

common sense :
and austere ontology 145–55

and metaphysically lightweight posits 150–1

and non-arbitrariness of composition principle 151–3

common-sense realism 4 , 77–8 , 92

competence 231
and intuition 224–5

competence-based performance errors 147–50

scorekeeping 148–9 , 167

scorekeeping confusions 149–50

completion relativism 28–9

conceptual reductionism 200 n. 22

conceptual relativism 4

conceptual truth 231

constructive empiricism 26–7 , 86 n. 14 , 191–3

constructivism 101–2

contextual semantics 6 , 137–43 , 145–55

 p p p p p y

and austere ontology 138 , 139–43 , 145–155

and common sense 145–55

variation in 143–5

contextualism :
attributor 59 n.

relativism and 14–16

David, Marian 203–7

Davidson, Donald 241 , 243 , 244 , 247

deflationary facts 194 , 195 n. 16

deflationary realism 87–8 , 93 , 95

deflationary truth 2 , 87–8

deflationism 2 , 82–3 , 128–30 , 142 , 143

Descartes, René 80–1 , 92 , 214 , 222

on intuition 210–11

Devitt, Michael 77–8 , 100–21 , 239

and causal theory of reference 130–2

realism (now) 126–7

dialetheism 66–7

direct-correspondence (DC) semantic standards 147 , 149 , 154

disagreement, faultless 36 , 37 , 38

discourse :
paraphrasing strategies 139–40

semantic pretense theories 140

disputes of inclination 38–40 , 42–6 , 47–8 , 62–3 , 64 , 65–8

diversity, moral 23

Dummett, Michael 179 , 181 , 243

end p 249

Einstein, Albert 13 , 16

eliminativism 32

emotivism 189

empirical equivalence 110–17 , 129

empirical reductionism 200 n. 22

empiricism 242
arguments against 88–93

constructive 26–7 , 86 n. 14 , 191–3

entitlement 236–7

entity realism 102–3 , 126

epistemic justification 218–20 , 222

epistemology of realism 132–5

error theories 191–2 , 193

statements and 25–6 , 28

evil demon hypothesis 112

excluded middle, law of 43 , 51 , 194

expressivism 194 , 201–2

expressivist realism 39
 p p p p p y

fact realism 102–4 , 126 , 130

facts 203–7
atomic 70–1

basic 196

deflationary 194 , 195 n. 16

negative/positive 70–3

real 195 , 196

robust 194 , 195 n. 16

weird 190–1 , 194 , 199

factual relativism 18–21

fallibility 222–3

faultless disagreement 36 , 37 , 38

faultlessness 36 , 37 , 72
analetheism and 67

and dialetheism 66

in disputes of inclination 38 , 39 , 40 , 41 , 42 , 47–8 , 59

intuitionism and 48–9

ordinary view and 63 , 64

Feyerabend, Paul 101

fictionalism 190–2 , 193

fictionalist moral relativism 24–8 , 30–1

Field, Hartry 191

Fine, Arthur 96 , 202 n. 25 , 239 , 240

natural ontological attitude (NOA) 93 , 95 , 104 n. 8

on realism 107

Fine, kit 36
on reality 195–8

Fodor, Jerry 157 , 192–3

foundationalism 88 , 91–2

Frege, Gottlob 179

Galileo Galilei 13 , 14

Gauker, Christopher 125–35 , 239

Gibbard, Allan 201–2

global irrealism 142

global relativism 2 , 139

Goldbach's conjecture 43–4

Graff, Delia 172

Grice, Paul 183–4

Hacking, Ian 102 n. 3 , 110 n. 21

Harman, Gilbert 15 , 23 , 28
on truth conditions 16–18 , 22

Hegel, G. W. F. 244

Heyting, Arend 180

Horgan, Terry 137–60 , 239 , 240 , 243

Horwich, Paul 188–202 , 241–2 , 246

 p p p p p y

on deflationism 82–3

humanity, principle of 18

Hume, David 229

ideal warranted affirmability 140–1

idealism 4 , 246

idealized assertibility, relativism and 56–9

imperatival moral relativism 31–5

inclination, disputes of :
contradiction in 36 , 38 , 40 , 42 , 46 , 48 , 62 , 67 , 69 , 72

faultlessness in 38 , 39 , 40 , 41 , 42 , 47–8 , 59

intuitionism and 42–52

ordinary view of 38–42 , 46 , 47 , 48 , 62–3

and simple deduction 40–2 , 45–6 , 48 , 59 , 63 , 64

sustainability in 38 , 39 , 41 , 42 , 46–8 , 59

incoherence, benign logical 168–73

indirect-correspondence (IC) semantic standards 6 , 145 , 146 , 154–5

inferences 90–1 , 218

inferentialism 83–4 , 92 , 94 , 95–6 , 245

instrumentalism 104 n. 10

instrumentalist anti-realism 86 n. 14

intentionality 80 , 99

internalism 88

introspection 213–14 , 215 , 221

intuitionism 180–1
bivalence and 43–4 , 45 , 48

and disputes of inclination 42–52

faultlessness and 48–9

Goldbach's conjecture 43–4

quandaries and 44–5

intuitions 208–18 , 223–4

attraction and 210 , 228–9

and beliefs 208–9 , 210 , 211 , 212 , 213

characteristics of 232–4

competence and 224–5

definition of 227–31

as epistemic hinges 237

and fallibility 222–3

introspection and 214

knowledge and 212

minimalist conception of 232–4 , 235

end p 250

paradoxes and 209 , 220–1

philosophical 234–5

propositional 212–13

semantic 234
 p p p p p y

source-opacity and 232 , 233

theoretical relevancy 235

trustworthiness of 228 , 232 , 233 , 235–7

truths and 209 , 210

irrealism, global 142

justification 22 n. 4 , 218–20 , 221 , 222

Kekulé, Friedrich August 133–4

Kitcher, Philip 105

causal theory of reference 96–7

on deflationary realism 93 , 95

natural epistemological attitude (NEA) 93–4 , 95 , 97–8

on naturalism 85 n. 12

on real realism 78 , 86–7

knowledge 80 , 84–5 , 88–92

basic 88–9 , 90

intuition and 212

observational 88–9 , 90 , 91 , 92

Kölbel, Max 55 n. 12

Kuhn, Thomas 101

Kukla, Andre 111 , 112

on empirical equivalence 114 , 115 n. 34 , 116 n. 36

language, philosophy of 39 , 179–80

Laudan, Larry :
on approximate truth 106 n. 12

on empirical equivalence 115 n. 34 , 116 n. 36

pessimistic meta-induction 107

realism, attack on 105 , 107

on underdetermination argument 113–14

Leplin, Jarrett 103 n. 5 , 107

on empirical equivalence 115 n. 34 , 116 n. 36

on underdetermination argument 113–14

Lewis, David 148 , 157

Lewis, Peter 107 n. 15

linguistic turn 247

linguistics 179–80

local relativism 2–3

logic of paradox (LP) 66 , 67

logical incoherence, benign 153–5 , 163 , 168–73

Lynch, Michael P. 227–37

McAllister, J. W. 105

McDowell, John 195 , 244–5

MacFarlane, John 36 , 55 n. 12

Mackie, John 191

McMullin, Ernan 105 , 110 n. 21

meaning :
 p p p p p y

inferentialism about 83–4

non-factualism about 194

use theory of 198–9

memory 222–3

meta-induction, pessimistic 107 , 117–20

methodological naturalism 99

methodological skepticism 81

methodology 174 , 177–87

mind, theories of 85

minimalism 142 , 143

moral codes 22–4 , 26–7 , 29 , 31–5

moral diversity 23

moral nihilism 33

moral relativism 16 , 19 , 21–2 , 29–30

fictionalist 24–8 , 30–1

imperatival 33–5

Mueller-Lyer illusion 147–8 , 209 , 213

Musgrave, Alan 110 n. 21

natural epistemological attitude (NEA) 93–4 , 95 , 97–8

natural ontological attitude (NOA) 93–8 , 104 n. 8

natural science 177–8

naturalism 4–5 , 109 n.

methodological 99

scientific 85–6 , 98 , 99

necessity 178 , 210–11 , 216 , 229–30

negative facts 70–3

neo-Cartesianism 5 , 81 , 84–6 , 96–7

neo-pragmatism 5 , 81–4 , 96 , 99 , 240

deflationism about truth 82–3

epistemological orientation 81–2

inferentialism about meaning 83–4

nihilism, moral 33

non-arbitrariness of composition, principle of 151–3

non-reductive realism 204 , 205

Nye, Mary Jo 133–4

observables, skepticism about 100–1 , 108

observational knowledge 88–9 , 90 , 91 , 92

ontology, austere 138 , 139–43 , 145–55

paraconsistency 65–8

paradoxes :
intuitions and 209 , 220–1

logic of paradox 66 , 67

sorites 154 , 168–73

Parmenides 240 , 241 , 242 , 246

perception 208 , 213 , 214 , 219 , 236

 p p p p p y

performance errors 147–50 , 167

permissibility 221–2

Perrin, Jean 134–5

pessimistic induction argument 107 , 117–20

end p 251

philosophical intuitions 234–5

Plato 240 , 242–3 , 246

Poincaré, Henri 134

positive facts 70–3

positivistic instrumentalism 104 n. 10

Potr , Matja 137–60

practical justification 221

pragmatism, classical 2

predictions 216–18

Presocratics 177–8

presupposition, rule of accomodation for 157

propositional intuitions 212–13

propositions :
moral codes as 22–4 , 26–7 , 29

truth of 54 , 55 , 68 , 188–9 , 215

Psillos, Stathis 108 , 111 n. 25

Pust, Joel 236 n. 13

Putnam, Hilary 100 , 140–1 , 241

quantum theory 101 n.

quasi-particularism 139

quasi-particularist semantic normativity 155–60

queerness 152 , 201

quietism 202 , 203

Quine, W. V. O. 82 , 83 , 231

Raffman, Diana 172

rampant realism 38–9 , 41 , 48 , 49–51

analetheism and 68

and facts 70–1 , 72

real facts 195 , 196

real realism 78–9 , 85

argument to, from NOA 93–8

Kitcher on 78 , 86–7

realism 4 , 77 , 86–8 , 107 , 126–7 , 194–8

AIC 138–43 , 145 , 155 , 160

alethic 2

and anti-realism 179–81

attacks on 105 , 107 , 108

austere metaphysical 5–6 , 137

common-sense 4 , 77–8 , 92

deflationary 87–8 , 93 , 95
 p p p p p y

disjunctive theory of 127–8

entity 102–3 , 126

epistemology of 132–5

expressivist 39

fact 102–4 , 126 , 130

rampant 38–9 , 41 , 48 , 49–51 , 68 , 70–1 , 72

real 78–9 , 85 , 86–7 , 93–8

reductive/non-reductive 204 , 205

response-dependent 39

scientific 4 , 5 , 77–8 , 92 , 100–21 , 125–32

standard-relativized 39–40

and truth 78

realism (now) 126–7

reality 195–200

reductionism 200 , 203 , 204 , 207

reductionist anti-realism 86 n. 14

reductive realism 2–4 , 205

reference, theories of 103–4 , 118–19

causal 95 , 96–7 , 130–2 , 194

scientific realism and 130–2

relative correspondence truth 69–73

ordinary view 72–3

polarity view 70–1

relatively positive/negative states 71–2

relativism 42 , 189–92 , 193

alethic 2

completion 28–9

conceptual 4

contextualism and 14–16

factual 18–21

fictionalist moral 24–8 , 30–1

global 2 , 139

and idealized assertibility 56–9

imperatival moral 31–5

about justification 22 n. 4

local 2–3

moral 16 , 19 , 21–2 , 24–8 , 29–30 , 30–1 , 33–5

about propositional truth 68

relational meanings 14–16

relational truth conditions 16–18

replacement 21 , 32

taste-function 61–2 , 68–9

true 52–5 , 68–9

relativistic truth 55 n. 12

relativity, and truth conditions 53–4

 p p p p p y

reliability inference 90

replacement relativism 21 , 32

response-dependent realism 39

revisionism 201

Rey, George 108 n. 19

Richard, Mark 162–74

robust facts 194 , 195 n. 16

Rorty, Richard 96 , 195 , 239–47

Rosen, Gideon 191

same-level inferences 90–1

Schiffer, Stephen 192–3

scientific naturalism 85–6 , 98 , 99

scientific realism 4 , 5 , 77–8 , 92 , 100–21

arguments against 110–20

arguments for 105–10

basic abductive argument for 109–10

and causal theory of reference 130–2

definition of 100–5

pessimistic meta-induction against 117–20

and semantics 125–30

success argument for 105–9

success of methodology argument for 109

end p 252