Metaethics after Moore
This page intentionally left blank
Metaethics after Moore
edited by
T E R RY H O RG A N
and
MARK TIMMONS
CLARENDON PRESS
●
OXFORD
3
Great Clarendon Street, Oxford ox2 dp
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 trade mark 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 authors 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 the same condition on any acquirer
British Library Cataloguing in Publication Data
Data available
Library of Congress Cataloging in Publication Data
Metaethics after Moore / edited by Terry Horgan and Mark Timmons.
p. cm.
Includes bibliographical references and index.
1. Ethics. 2. Moore, G. E. (George Edward), 1873–1958. Principia ethica.
I. Horgan, Terry, 1948– II. Timmons, Mark, 1951–
BJ37.M47 2006 170⬘.42—dc22 2005023277
Typeset by Newgen Imaging Systems (P) Ltd., Chennai, India
Printed in Great Britain
on acid-free paper by
Biddles Ltd., King’s Lynn, Norfolk
ISBN 0–19–926990–4 978–0–19–926990–7
ISBN 0–19–926991–2 (Pbk.) 978–0–19–926991–4 (Pbk.)
1 3 5 7 9 10 8 6 4 2
PREFACE
Since its publication in 1903, G. E. Moore’s Principia Ethica has continued
to exert a powerful influence on metaethical enquiry. This volume contains
sixteen essays that represent recent work in metaethics after, and in some cases
directly inspired by, the work of Moore. Seven of the essays were originally
presented at the 2002 Spindel Conference commemorating the one hundredth
anniversary of the publication of Principia Ethica and in celebration of a hundred years of metaethics. They are reprinted here (some slightly revised) from
the Southern Journal of Philosophy, 41 (2003). Our introduction situates the
essays in relation to central themes in Moore’s metaethics.
We are grateful to the Southern Journal of Philosophy for permission to reprint
the papers that appeared in the 2003 supplement. We also wish to thank our
editor at Oxford University Press, Peter Momtchiloff, for his guidance and
support in our work on this anthology.
T.H. and M.T.
Tucson, Ariz.
This page intentionally left blank
CONTENTS
List of Contributors
Introduction
Terry Horgan and Mark Timmons
1. How Should Ethics Relate to (the Rest of ) Philosophy?
Moore’s Legacy
Stephen Darwall
ix
1
17
2. What Do Reasons Do?
Jonathan Dancy
39
3. Evaluations of Rationality
Sigrún Svavarsdóttir
61
4. Intrinsic Value and Reasons for Action
Robert Audi
79
5. Personal Good
Connie S. Rosati
107
6. Moore on the Right, the Good, and Uncertainty
Michael Smith
133
7. Scanlon versus Moore on Goodness
Philip Stratton-Lake and Brad Hooker
149
8. Opening Questions, Following Rules
Paul Bloomfield
169
9. Was Moore a Moorean?
Jamie Dreier
191
10. Ethics as Philosophy: A Defense of Ethical Nonnaturalism
Russ Shafer-Landau
209
11. The Legacy of Principia
Judith Jarvis Thomson
233
12. Cognitivist Expressivism
Terry Horgan and Mark Timmons
255
13. Truth and the Expressing in Expressivism
Stephen Barker
299
viii
Contents
14. Normative Properties
Allan Gibbard
319
15. Moral Intuitionism Meets Empirical Psychology
Walter Sinnott-Armstrong
339
16. Ethics Dehumanized
Panayot Butchvarov
367
Index
391
LIST OF CONTRIBUTORS
Robert Audi is Professor of Philosophy and David E. Gallo Chair in Ethics at
the University of Notre Dame. His books include The Good in the Right
(2004), The Architecture of Reason (2001), Religious Commitment and Secular
Reason (2000), and Moral Knowledge and Ethical Character (1997).
Stephen Barker is Lecturer in Philosophy at the University of Nottingham.
In addition to many articles, he is the author of Renewing Meaning: A SpeechAct Theoretic Approach (2004). He is currently completing a book on an expressivist theory of truth.
Paul Bloomfield is Assistant Professor of Philosophy at the University of
Connecticut and in addition to many articles in metaethics he is author of
Moral Reality (2001).
Panayot Butchvarov is the University of Iowa Distinguished Professor of
Philosophy. He is the author of Resemblance and Identity (1966), The Concept
of Knowledge (1970), Being qua Being: A Theory of Identity, Existence and
Predication (1979), Skepticism in Ethics (1989), and Skepticism about the
External World (1998).
Jonathan Dancy is Professor of Philosophy at the University of Reading and
Professor of Philosophy at the University of Texas at Austin. He is the author of
An Introduction to Contemporary Epistemology (1985), Berkeley: an Introduction
(1987), Moral Reasons (1993), Practical Reality (2000), and Ethics without
Principles (2004).
Stephen Darwall is John Dewey Collegiate Professor of Philosophy at the
University of Michigan. His books include The British Moralists and the
Internal ‘Ought’: 1640–1740 (1995), Philosophical Ethics (1998), and Welfare
and Rational Care (2002).
Jamie Dreier is Professor of Philosophy at Brown University. He has published numerous papers on a wide variety of subjects, most recently: ‘Why
Ethical Satisficing Makes Sense and Rational Satisficing Doesn’t’, in Satisficing
and Maximizing: Moral Theorists on Practical Reason (2004), ‘Relativism and
Nihilism’, forthcoming in Oxford Handbook of Ethical Theory, and ‘Pettit on
Preference for Properties and Prospects’, forthcoming in Philosohical Studies.
He is editor of Blackwell’s Contemporary Debates in Moral Theory (2005).
x
List of Contributors
Allan Gibbard is the Richard B. Brandt Distinguished University Professor
of Philosophy at the University of Michigan, and the author of Wise Choices,
Apt Feelings (1990) and Thinking How to Live (2003).
Brad Hooker is Professor of Philosophy at the University of Reading. He is
author of Ideal Code, Real World: A Rule-Consequentialist Theory of Morality
(2002), editor of Rationality, Rules, and Utility (1993), and Truth in Ethics
(1996), and co-editor of Well-being and Morality (2000), Morality, Rules, and
Consequences (2000), and Moral Particularism (2000). He has also published a
large number of articles, mostly in moral philosophy. He is currently working
on a book about fairness and a textbook on moral philosophy and a history of
twentieth-century moral philosophy.
Terry Horgan is Professor of Philosophy at the University of Arizona and
author of many articles in philosophy of mind and metaphysics. He is co-author
(with John Tienson) of Connectionism and the Philosophy of Psychology
(MIT, 1996) and (with Matjaz Potrc) of Austere Realism (forthcoming) and is
completing a book with David Henderson, A Priori Naturalized Epistemology:
At the Interface of Cognitive Science and Coneptual Analysis. He and Mark
Timmons have collaborated on many papers in metaethics and they are currently working on topics in moral phenomenology.
Connie S. Rosati is Associate Professor of Philosophy at the University of
Arizona. She specializes in ethics, philosophy of law, and social and political philosophy. Her ongoing research concerns the nature of personal good and the
nature and normativity of constitutions. Recent publications include ‘Agency
and the Open Question Argument’, Ethics (2003) and ‘Some Puzzles About
the Objectivity of Law’, Law and Philosophy (2004).
Russ Shafer-Landau is Professor of Philosophy at the University of
Wisconsin and author of Moral Realism: A Defence (2003), and Whatever
Happened to Good and Evil? (2003). He is also co-editor of Reason and
Responsibility (12th edition, 2004) and editor of Oxford Studies in Metaethics.
Walter Sinnott-Armstrong is Professor of Philosophy and Hardy Professor
of Legal Studies at Dartmouth College, where he has taught since 1981, after
receiving his BA from Amherst College and his Ph.D. from Yale University. His
newest book, Moral Skepticisms, will be published by Oxford University Press in
2006. He is currently working on empirical moral psychology and brain science.
Michael Smith is Professor of Philosophy at Princeton University and author
of The Moral Problem (1994), Ethics and the A Priori: Selected Essays on Moral
Psychology and Meta-Ethics (2004), and co-author (with Frank Jackson and
Philip Pettit) of Mind, Morality, and Explanation: Selected Collaborations (2004).
List of Contributors
xi
Phillip Stratton-Lake is Professor of Philosophy at the University of Reading
and author of Kant, Duty and Moral Worth (2000). He is editor of On What We Owe
to Each Other: Scanlon’s Contractualism (2004), Ethical Intuitionism: Re-evaluations
(2002), and a new edition of W.D. Ross’s The Right and the Good (2002).
Sigrún Svavarsdóttir is Assistant Professor of Philosophy at the Ohio State
University. Her publications include: ‘Objective Values: Does Metaethics Rest
on a Mistake?’ in Objectivity in Law and Morals, B. Leiter, ed. (2001), ‘Moral
Cognitivism and Motivation’, The Philosophical Review (1999), and ‘How
Do Moral Judgments Motivate?’ in Contemporary Debates in Moral Theory,
J. Dreier, ed. (2006).
Judith Jarvis Thomson is Professor of Philosophy at the Massachusetts
Institute of Technology, specializing in ethics and metaphysics. She is the
author of Acts and Other Events (1977), Rights, Restitution, and Risk (1986),
The Realm of Rights (1990), and Moral Relativism and Moral Objectivity, with
Gilbert Harman (1996). She has also authored many highly influential papers.
Mark Timmons is Professor of Philosophy at the University of Arizona and
author of Morality without Foundations (1999) and Moral Theory: An
Introduction (2002), and editor of Kant’s Metaphysics of Morals: Interpretative
Essays (2002). He and Terry Horgan are currently working on philosophical
issues associated with the phenomenology of moral experience.
This page intentionally left blank
Introduction
Terry Horgan and Mark Timmons
Metaethics, understood as a distinct branch of ethics, is often traced to
G. E. Moore’s 1903 classic Principia Ethica (PE). Whereas normative ethics is
concerned to answer first-order moral questions about what is good and bad,
right and wrong, virtuous and vicious, metaethics is concerned to answer
second-order non-moral questions, including (but not restricted to) questions
about the semantics, metaphysics, and epistemology of moral thought and discourse. Metaethics, then, as a recognized branch of ethics, is part of the philosophical legacy of PE. Moreover Moore’s own combination of metaethical
views has continued to exert a strong influence on metaethical enquiry of the
last hundred plus years, and forms another part of the rich legacy of Principia.
The papers in this volume represent recent work in metaethics that reflects
the rich philosophical heritage of Moore’s PE. They are organized in relation to
central metaethical claims defended by Moore—claims that can be put into
four main groups: the subject matter of ethics, moral semantics, moral metaphysics, and moral epistemology. In what immediately follows we will briefly
summarize the papers, relating them to Moore’s metaethical views.
The subject matter of Ethics
In the first chapter of PE, ‘The Subject-Matter of Ethics’, Moore spends the
first four sections explaining his conception of the field of ethics. In these passages, he refers to an ‘ideal of ethical science’ (56) which he divides into two
main parts. First, there are semantic and related metaphysical questions about
the meanings of moral terms (and the concepts they express) and, second, there
are questions about what sorts of items possess the properties which moral
terms denote. What emerges from Moore’s discussion of the subject matter of
ethics are two theses. First is what we will call the independence thesis, according
to which semantic and related metaphysical questions—questions of
metaethics—can be pursued independently of and are properly prior to
enquiry into substantive matters about the kinds of items that are good or bad,
right or wrong, virtuous or vicious. Second, Moore holds a certain primacy
2
Terry Horgan and Mark Timmons
thesis, according to which the concept of goodness (and badness) is more
fundamental than and can be used to define the concepts of rightness (and
wrongness) and virtue (and vice). Thus, for Moore, the study of ethics,
properly conducted, should begin with an enquiry focused on the concept of
goodness.
The papers by Stephen Darwall, Sigrún Svavarsdóttir, Jonathan Dancy,
Robert Audi, Connie Rosati, and Michael Smith all have to do with one or
another of these two theses. In ‘How Should Ethics Relate to (the Rest of )
Philosophy?’, Darwall challenges both the claims of independence and priority. He argues that although metaethics and normative ethics are properly
focused on different issues, they need to be brought into dynamic relation with
one another in order to produce a systematic and defensible philosophical
ethics. This mutual dependence, claims Darwall, is owing to the fact that issues
of normativity are at the center of the concerns of both metaethics and normative ethics. In making his case, Darwall examines Moore’s doctrine that an
irreducible notion of intrinsic value is fundamental in ethics and argues that
although Moore was correct in thinking that ethical notions are irreducible, he
was incorrect in thinking that this is because they have a notion of intrinsic
value at their core. Rather, according to Darwall, the notion of a normative
reason is ethically fundamental, and a proper philosophical ethics that fully
accommodates the normativity involved in ethical thought and discourse will
require that metaethical issues and normative issues bearing on normativity be
‘pursued interdependently as complementary aspects of a comprehensive
philosophical ethics’. He illustrates this claim by explaining how certain
debates within normative ethics over consequentialism and over virtue depend
upon metaethical issues about the nature of normativity.
Darwall’s paper reflects one important way in which contemporary metaethics
differs in emphasis from Moore’s position. In recent times, philosophers have
come to recognize the importance of evaluations of normative reasons and
rationality, not only in the field of ethics but in relation to the subject matter of
such fields as epistemology, semantics, and philosophy of mind. The contributions of Svavarsdóttir and Dancy reflect this trend. Svavarsdóttir’s ‘Evaluations
of Rationality’ works from the guiding idea that rationality is the excellence of
a rational agent qua rational and goes on to defend a neo-Humean conception
of evaluations of theoretical and practical rationality, according to which such
evaluations make essential reference to an agent’s ends or goals in assessing
the rationality of the agent’s beliefs, actions, and intentions. Evaluations of
theoretical and practical rationality differ according to the types of goals
relative to which we make evaluations of rationality. Svavarsdóttir defends this
view by appealing to intuitions about irrationality with respect to particular
Introduction
3
cases, which she claims are best explained by the neo-Humean—a defense
which is neutral with regard to metaphysical issues about the nature of reasons.
Svavarsdóttir’s defense of her view is admittedly partial because it does not fully
address questions about the justificatory force of rationality evaluations,
leaving as she notes important tasks for the neo-Humean to tackle.
Moore held that considerations of intrinsic value grounded moral reasons to
act. As we noted, according to Darwall, considerations of normativity are
fundamental for both metaethical enquiry and normative ethics. Dancy’s
paper, ‘What Do Reasons Do?’ is focused on the issue of how we are to understand what he calls practical ‘contributory reasons’, particularly as they are
related to oughts. Dancy begins by rehearsing six proposals for understanding
contributory reasons in terms of an ‘overall ought’, and rejects them all.
Dancy’s own proposal is that a ‘reason is something that favours action’, where
favoring is a normative relation in which a reason stands to a particular way
of acting.
Since the contributory cannot be reduced to an overall ought (or any overall
notion, such as goodness), Dancy proposes to go the other way and reduce
overall oughts to the contributory. However, instead of attempting to reduce
overall oughts to favoring reasons (which he doubts can be done), Dancy introduces the notion of a ‘contributory ought’—‘a monadic feature of an action
which is consequent on, or resultant from, some other feature—the “oughtmaking” feature, whatever it is’. How are we to understand how an overall
ought is related to the contributory ought? Here is where Dancy thinks that
appeal to fittingness, a notion employed by the classical intuitionists, offers
promise. In partially defending this claim, he argues that Michael Smith’s
‘Humean realism’ and Allan Gibbard’s expressivism lack the resources needed
for adequately understanding practical reasons and oughts.
The papers by Audi and Rosati concern aspects of Moore’s theory of intrinsic
value. In ‘Intrinsic Value and Reasons for Action’, Audi sketches a theory of
intrinsic value that aims to incorporate certain elements of Moore’s theory, but
which goes beyond it in important ways while also avoiding commitment to
many of Moore’s controversial normative and metaethical views. Moore held
that experiences and non-experiential items such as artworks can be the bearers
of intrinsic value. By contrast, Audi defends experientialism—according to
which the bearers of intrinsic value are concrete experiences—partly by
arguing that it is experiences that seem to have the kind of Aristotelian ‘finality’
and thus ‘choiceworthiness’ that is appropriate for anything’s having intrinsic
value. In order to accommodate the Moorean idea that items such as artworks
are in some sense ‘good in themselves’ (and not merely instrumentally good),
Audi introduces the notion of inherent value—a species of value that is
4
Terry Horgan and Mark Timmons
possessed by something whenever an appropriate experience of it is instrinsically
good. A painting, for example, can be inherently good because an appropriate
aesthetic experience of that object is itself intrinsically good. The concepts of
intrinsic and inherent value, along with a Moorean principle of organic unities
(suitably broadened), provide the basis for a nuanced theory of value whose
merits include the recognition and explanation of a wide range of intuitively
plausible value judgments, as well as contributing to a general theory of practical
reason.
While Audi’s contribution attempts to build on some of Moore’s ideas about
intrinsic value, Rosati in her ‘Personal Good’ challenges one aspect of Moore’s
view. In his critical remarks about egoism as a theory of motivation, Moore
argued that the notion of ‘good for’ that figures in claims about this or that
activity or pursuit being (non-morally, intrinsically) good for an individual is
incoherent.¹ Rosati argues that Moore is mistaken and defends an account of
the good-for relation modeled on the interpersonal relation of successful
loving. Success in an interpersonal loving relationship is characterized by the
fact that such relationships support their participant’s self-esteem, they are
energizing, they provide comfort and feeling of security as well as providing an
important element of a participant’s identity and sense of direction in life, and
such relationships tend to be self-perpetuating. The sort of relation involved in
something’s being good for an individual—part of her personal good—exhibit
these same general sorts of features. According to Rosati, then, the property
good-for is a second-order relational property that is realized in a person’s life
when she stands in the sorts of esteem-enhancing, energizing, and other just
mentioned relations to some pursuit or activity. Rosati defends her view in two
ways. First, she appeals to certain dualities of human nature and experience: we
are partly biological creatures on the one hand who often discover our good, as
when one discovers that she has a natural talent for music and proceeds to
develop her musical talent so that playing music becomes part of her personal
good. But we are also autonomous agents for whom our personal good is often
partly a matter of one’s own making—something invented rather than simply
discovered. In order for playing music to be part of her personal good, the
would-be musician must cultivate her talent and in this way she makes playing
music part of her personal good. Rosati’s account of personal good nicely
accommodates such dualities in that the various relations involved in something’s being good for oneself depend partly on facts about oneself that are
beyond one’s control but partly on what one does. The second way Rosati
defends her view is by responding to certain possible Moorean objections.
¹ See Darwall’s paper, section 3, for an illuminating explanation of Moore’s reason for this view.
Introduction
5
In chapter V of PE, ‘Ethics in Relation to Conduct’, Moore turns from
questions about the definition of ‘intrinsic value’ generally and ‘good’ in particular to questions about right action. He defends two claims in this chapter.
First, he defines right action in terms of intrinsic value: to claim that an action
A (performed by someone S on some occasion O) is right means the same as
claiming that S’s performing A in O resulted in a greater amount of intrinsic
value than would performing any alternative action open to S in O. In short,
for Moore, right actions maximize intrinsic value. Ideally, then, what Moore
calls ‘Practical Ethics’ aims to tell us which actions are right. But, as Moore
explains, in light of severe epistemic limitations on our knowledge about
which, from among alternative actions open to an agent, will maximize intrinsic
value, there is some question about how a morally motivated person should
make decisions on specific occasions. Such epistemic limits impose a ‘humbler’
task on Practical Ethics: one of determining which alternative action likely to
occur to an agent on some occasion is most likely to maximize intrinsic value.
This is the second of the claims about right action defended in chapter V. Frank
Jackson has argued that there is good reason to reject both of Moore’s claims.²
In their place, Jackson proposes a conception of right action in terms of the
expected intrinsic values of alternative actions, where the relevant expectation
is from the agent’s point of view on the occasion of action. According to
Jackson, this conception of right action is not only immune from various
counterexamples that beset both Moore’s proposal for Practical Ethics as well
his definition of ‘right’, but properly ties the rightness of actions to our critical
practices of holding agents responsible for what they do.
In ‘Moore on the Good, the Right, and Uncertainty’, Michael Smith
proposes a conception of Practical Ethics which, unlike Jackson’s proposal, ties
what epistemically limited agents are to do on some occasion not only to limits
on their non-evaluative information about how much intrinsic value would
result from various actions, but also to epistemic limits on their evaluative
information about what has intrinsic value. This amounts to advising morally
motivated agents that they are to maximize expected intrinsic value-as-theysee-it—advice that recognizes the double epistemic limits humans possess—an
extension of Jackson’s view. However, instead of following Jackson and defining a conception of right action in terms of the doubly constrained notion in
question, Smith argues that we have good reason to accept Moore’s definition
and thus good reason to resist tying our primary notion of right action to the
concept of what we can hold an agent responsible for. But adopting Moore’s
conception of right action might seem to be in tension with his modified
² ‘Decision-Theoretic Consequentialism and the Nearest and Dearest Objection’, Ethics, 101 (1991).
Terry Horgan and Mark Timmons
6
Jackson-style conception of Practical Ethics. After all, if the rightness of an
action depends on the comparative level of intrinsic value it would produce if
performed, then won’t a morally motivated agent be motivated by a desire to do
what Moore’s Practical Ethics recommends, namely, by a desire to do what will
likely maximize intrinsic value? If so, the Moorean element of Smith’s view is in
tension with the Jackson-inspired element. But Smith denies the assumption
about moral motivation featured in this challenge. Rather than thinking of moral
motivation in terms of a desire to maximize intrinsic value, Smith claims that we
should think instead in terms of having intrinsic desires for things one judges to
be intrinsically valuable, such as pleasure, or knowledge, or autonomy, or whatever. Thus, according to Smith, Moore’s conception of right action represents an
appropriate idealization of a plausible account of rational decision making.
Moral semantics
Moore famously began the 100 years of metaethics with his open question
argument—which he thought exposed the fallaciousness of all ‘reductive’
accounts of moral terms and concepts. On the basis of this argument, Moore
concluded that the primary concept of ethics—goodness—is ‘simple and indefinable’. The indefinability thesis, as we call it, is the cornerstone of Moore’s
moral semantics.
Moore’s version of the open question argument works by taking some
purported reductive definition of ‘good’ in terms of some nonnormative term or
phrase (e.g., ‘what we desire to desire’) and posing two questions of the following
kind (where ‘X’ is to be replaced by a term designating some item of evaluation):
X is good, but is it what we desire to desire?
And
X is what we desire to desire, but is X good?
Now if ‘good’ just means ‘that which we desire to desire’, then these questions
ought to strike us as equivalent to:
X is something we desire to desire, but do we desire to desire it?
This latter question is closed in the sense that its answer is trivially affirmative.
But the preceding questions are both open in the sense that they strike us as
non-obvious and open for debate.
For Moore, the pair of questions have an ‘open feel’ to them which he
explained in terms of our grasping of the meanings of the concepts involved.
Introduction
7
Toward the end of the twentieth century, we find that the open question
argument is alive and well. T. M. Scanlon uses a version of this argument in
defense of his buck-passing account of value.³ Scanlon claims that the ‘open
feel’ of the Moorean questions comes from the fact that judgments about
whether something is good express a practical conclusion about the reasons one
has for caring about that thing. To judge that some item of evaluation has such
and so natural properties does not involve judging that the item is good.
Hence, even if the claim that it has natural properties a, b, and c is the ground
for concluding that the item is good, it is a further step to draw the conclusion
that it is good. Hence, the open feel between judging that some item has
natural properties and judging that it is good. Whereas Moore concluded on
the basis of the open question argument that goodness is an unanalyzable, simple, non-natural property, which itself (as distinct from the natural properties
upon which goodness supervenes) provides reasons for action, Scanlon argues
that a better account of the matter is that goodness is a formal, higher-order
property which can be understood as a complex non-natural property: the
property of there being base properties that provide practical reasons. For
Moore, what has reason-giving power is the property of goodness itself—the
reason-giving buck rests with this property. For Scanlon, the reason-giving
buck is passed on to a thing’s good-making properties. Scanlon argues in
various ways for his buck-passing account, partly by explaining its superiority
to the Moorean view and partly by giving two arguments—one appealing to
considerations of parsimony, the other appealing to the plausibility of value
pluralism.
In ‘Scanlon versus Moore on Goodness’, Philip Stratton-Lake and Brad
Hooker offer a partial defense of Scanlon’s buck-passing account of the relation
between base properties, goodness, and practical reasons. Jonathan Dancy and
Roger Crisp have both argued that even if Scanlon’s buck-passing account is
superior to the Moorean account, there are other contending accounts that
Scanlon does not consider. Against Dancy and Crisp, Stratton-Lake and
Hooker argue that these proposed accounts, although genuine alternatives to
the Moorean and buck-passing accounts, are nevertheless deeply problematic
and do nothing to harm the case for Scanlon’s account. Regarding Scanlon’s
two arguments, the authors find that the parsimony argument, once clarified,
does offer some support for the buck-passing view, but they conclude that the
appeal to value pluralism does not aid the defense of this view. They finally
defend Scanlon’s account against an open question worry about the relation
between the fact that something has reason-giving properties and its goodness.
³ What We Owe to Each Other (Cambridge, Mass.: Harvard University Press, 1998).
8
Terry Horgan and Mark Timmons
Paul Bloomfield in ‘Opening Questions, Following Rules’ begins by noting
that the twentieth-century beneficiary of the open question argument has been
(rather ironically) the class of non-realist views, including non-cognitivism and
expressivism. Bloomfield contends that Moore did not properly diagnose the
openness of the relevant questions about goodness; it is not simplicity versus
complexity, and it is not indefinability versus definability. Rather, Bloomfield
contends, it is the normativity involved in moral judgments and concepts that
keeps Moorean questions open and blocks definitions of ‘good’—the same sort
of normativity that keeps questions open in relation to concepts like ‘plus’,
‘mass’, ‘triangle’. According to Bloomfield, then, the issue of normativity in
semantics, epistemology, and ethics is basically the same which he puts as
follows: ‘How can features of the world establish conditions under which it
makes sense for us to think that there are ways we ought to conduct ourselves
(with regard to our actions, our speech, or our beliefs) and other ways which
ought not to be followed?’ A clear implication of Bloomfield’s line of argument
is that those working in metaethics have often labored under the mistaken
assumption that moral terms like ‘good’ are especially problematic.
In addition to the semantic thesis of irreducibility, Moore took for granted
the descriptivist thesis according to which moral judgments of the forms ‘A is
good’ and ‘A is bad’ purport to attribute a property to some item and can thus
be true or false in just the same way in which ordinary non-moral judgments
about the empirical world can be true or false. Since Moore thought that such
judgments sometimes successfully do what they purport to do, he was
committed to certain metaphysical views to which we now turn.
Moral metaphysics
Moore held a version of moral realism—roughly the view that there are moral
properties and moral facts (in which those properties figure) whose existence
and nature are independent of the stances of individuals and groups. But perhaps the most puzzling doctrine in Moore’s metaethics is his view that goodness is a non-natural property. Many have found this hard to accept, partly
because the claim itself is so obscure. In his paper ‘Was Moore a Moorean?’,
Jamie Dreier traces Moore’s attempts, beginning in PE up though his 1922
‘The Conception of Intrinsic Value’,⁴ to characterize the difference between
⁴ This paper appeared originally in chapter VII of Moore’s Philosophical Studies (Kegan Paul, Trench,
Trubner & Co., 1922) and is reprinted in the 1993 revised edition of PE published by Cambridge University
Press.
Introduction
9
natural and non-natural properties, finding the most plausible characterization
in terms of a distinctive kind of non-logical supervenience relation that links
the property of goodness to the natural properties upon which it supervenes.
The problem with the appeal to a kind of non-logical supervenience, according to Dreier, is that it does not really help us understand the idea that goodness
is supposed to be non-natural: the property of being yellow does not logically
follow from a characterization of those properties upon which it supervenes,
but yellow is a paradigm natural property for Moore. Based on certain textual
clues, Dreier proposes that Moore misdescribed the distinction he sought to
capture in his natural/non-natural properties distinction. What Moore was
after, claims Dreier, is more aptly put as a distinction between description and
evaluation—a distinction central to expressivist views. So why wasn’t Moore an
expressivist? Expressivists generally agree with Moore that there is a conceptual
gap between the descriptive and the evaluative. Dreier’s conjecture is that
for the Moorean, this gap is a gap between properties, while for the expressivist
it isn’t.
But despite Moore’s difficulties in understanding this distinction, and
despite the fact that many post-PE moral philosophers rejected Moore’s nonnaturalism, the view is now enjoying a revival.⁵ In ‘Ethics as Philosophy: A
Defense of Ethical Nonnaturalism’, Russ Shafer-Landau provides a partial
defense of non-naturalism. He first provides an epistemological criterion for
understanding the metaphysical thesis of non-naturalism and then proceeds to
mount a defense of the view against two common objections: objections based
on facts about ethical disagreement and on causal criteria for having ontological status. His strategy is to call attention to the close parallel between ethical
enquiry and philosophical enquiry generally and argue that these parallels
provide a basis for rejecting the lines of objection in question and also provide
positive reasons to favor non-naturalism over its metaethical rivals. So first, just
as disagreement in philosophy itself does not undermine (or should not undermine) thinking that there are objective truths about such matters, neither
should disagreement in ethics undermine thinking that there are objective
truths or facts or justified belief in such facts and truths. As for the causal
efficacy criterion of ontological status, Shafer-Landau argues that even if moral
facts do not possess causal efficacy, we need not be skeptics about their ontological status as objectively real. If one insists on the causal efficacy test, then it
looks as if all putative normative facts fail the criterion and are not real. The
implausibility of this implication, then, casts doubt on the causal argument
against moral facts.
⁵ See, for instance, Robert Audi, The Good in the Right (Princeton: Princeton University Press, 2004) and
Russ Shafer-Landau, Moral Realism: A Defense (Oxford: Oxford University Press, 2002).
10
Terry Horgan and Mark Timmons
Arguably, the strongest challenge to any type of moral realism comes from
what is now called expressivism, the heir apparent of non-cognitivism.
Expressivists deny the semantic thesis of descriptivism and propose a different
philosophical picture for understanding and explaining moral thought and
discourse. According to Judith Jarvis Thomson (‘The Legacy of Principia’),
the legacy in question is that the force of the open question argument together
with the rejection of the Moorean idea that there are non-natural properties
motivate two related claims: the no normative truth value thesis according to
which no normative sentences have truth value, and the expressivist thesis that
in uttering or thinking a normative sentence what one does is express a favorable or unfavorable attitude toward the object of evaluation. Thomson
explores two main sources of reason for rejecting the first thesis—appeals to
minimalism about truth and the so-called Frege–Geach problem. She argues
that appeals to minimalism about truth are ultimately circular. However, the
Frege–Geach problem does represent a more serious challenge to those
(particularly expressivists) who embrace the no normative truth value thesis.
According to Thomson, the underlying insight of the Frege–Geach challenge
is the idea that ‘is good’ functions as a ‘logical’ predicate so that sentences
containing this predicate enter into logical relations with other sentences.
But, so the challenge goes, if ‘is good’ is a logical predicate then there is such a
property as goodness and (further) this means that if someone thinks or says a
sentence of the form ‘A is good’, then she has said something that has a truth
value. Hence, by this line of reasoning, the no normative truth value thesis is
false.
Thomson argues that attempts, particularly by expressivists, to rebut this
challenge falter, but rather than embrace the Moorean position (or any
metaethical position that would countenance the property goodness, or rightness), she denies the claim that ‘is good’ is a logical predicate. Rather, according to Thomson, sentences of the form ‘A is good’ are semantically incomplete
and thus ‘is good’ is not (in the requisite sense) a logical predicate. The main
idea is that simply to say of something that it is good without also thinking that
it is good in a certain way is not to attribute any genuine property to the item
in question. Thus, there is no property that people attribute to something
when they use this form of expression and so Moore’s premise—that there is a
property of goodness—is false. However, on Thomson’s view, expressivists who
deny Moore’s premise are mistaken in what someone does in engaging in normative evaluation. Normative claims that predicate goodness or rightness in a
way, as when someone claims that so and so is a good baseball player or that
such and so move in chess was the right move to make, are predicating genuine
properties and properties that are moreover arguably natural. If this is correct,
Introduction
11
then, as Thomson notes, Moore’s open question argument has misled philosophers
to fix upon the pseudo-property of goodness.
Recent developments of the expressivist position are represented in the
papers by Terry Horgan and Mark Timmons, Stephen Barker, and Allan
Gibbard. As already noted, metaethical expressivism is (broadly speaking) the
view that moral judgments do not primarily function to report or describe
moral facts or properties, but instead have an action-oriented expressive function. This primarily negative characterization leaves much open, including
what sort of psychological state a moral judgment expresses, though it is often
taken for granted that such states are not beliefs. And if one embraces what
Horgan and Timmons call the semantic assumption—the idea that beliefs are
necessarily descriptive in that they purport to represent or describe some state
of affairs—then an expressivist must reject the idea that moral judgments are
beliefs. Cognitivism in ethics is the view that moral judgments are beliefs and
so, given the semantic assumption, expressivism is not compatible with cognitivism. Horgan and Timmons challenge the semantic assumption by arguing
that moral judgments share enough of the phenomenological and functional
features that are central to the notion of belief to count as genuine beliefs—a
notion that does not require beliefs to be primarily descriptive. This, they
claim, opens the door to a cognitivist version of expressivism. Horgan and
Timmons sketch a version of cognitivist expressivism, including an account of
logical embedding (meant to deal with the Frege–Geach problem), which they
argue is prima facie more plausible than non-cognitivist and descriptivist
alternatives in metaethics.
In ‘Truth and the Expressing in Expressivism’, Barker proposes a new
framework for metaethical expressivism which involves a combination of several elements. First, Barker claims that evaluative sentences are used to make
genuine assertions, and so there are at least two types of assertion: reportive and
expressive. Second, and following from the first, assertions of both sort are
truth-apt. These claims are embraced also by Horgan and Timmons. But
unlike them, Barker argues that all assertions are representational in that they
purport to represent or describe some state of affairs. So, how do merely
reportive assertions differ from expressive assertions? In response, Barker
proposes what he calls a pragmatic conception of truth according to which
truth-bearers are sentences with representational content that are also used with
an assertoric purpose. The idea is that the essential difference between reportive
and expressive assertions concerns the purposes or intentions for which they
are asserted: ‘in reportive assertions, speakers defend commitments to representational intentions; in expressive assertions speakers defend commitments to
states [cognitive or conative] whose possession they have in fact represented.
12
Terry Horgan and Mark Timmons
In uttering a value sentence, for instance, one is expressing a desire (or related
motivational state) which, according to Barker’s analysis, the speaker is
prepared to defend. Barker explains how his form of expressivism can make
sense of the various objectivist trappings of moral discourse including its truthaptness, logical embedding, and being subject to rational debate.
Moore claimed that what he took to be the fundamental moral concept,
goodness, is a non-natural concept from which, together with his premise that
there is a property goodness, he inferred that this concept signifies a non-natural
property. Gibbard (‘Normative Properties’) distinguishes properties from concepts. The concepts water and H2O are different concepts though they in fact
signify the same property: the property of being water is the same property as
that of being composed of H2O molecules. According to Gibbard, Moore was
correct in noting an important difference between basic moral concepts and
naturalistic concepts of the sort featured in scientific and everyday discourse
about the empirical world. However, it is Gibbard’s view that basic moral concepts in particular and normative concepts in general signify natural properties: some natural property is the property of being good. For Gibbard, the
concept of good is a complex concept involving the concept of ought, and his
main thesis in his paper is what he calls the thesis of natural constitution: some
broadly natural property constitutes being what one ought to do. Gibbard’s main
argument for this claim begins with a traditional non-cognitivist (expressivist)
theme that to understand what the word ‘ought’ means we need to say what it
is to think or claim that someone ought to do something. Gibbard understands
ought-statements in terms of the activity of planning and proposes that we can
best grasp the content or meaning of such statements (both simple and
logically complex) by understanding what it is to disagree in plan. The upshot
of his argument (presented in section 1 of his paper) is that any planner is committed to the thesis of normative constitution. Gibbard’s paper is concerned
with exploring and defending the philosophical assumptions (e.g., about the
nature of properties) presupposed in this argument. His overall metaethical
view represents a blend of non-naturalism about moral concepts with a
naturalist concept of moral properties.
Moral epistemology
In Principia, Moore’s enquiry into the meaning of moral terms was intended to
have a direct bearing on issues in moral epistemology. With respect to substantive
claims involving predication of goodness, Moore held that ‘no relevant evidence whatever can be adduced’ (PE, 1st edition preface, 34), that such claims
Introduction
13
are self-evidently true, and that they can be known on the basis of intuition.
The basic claim of the moral intuitionist is that it is possible for individuals to
be epistemically justified in holding certain moral beliefs independently of
whether they are able to infer those beliefs from other beliefs they hold. This is
typically called ‘non-inferential’ justification. The twentieth century has seen
the fortunes of moral intuitionism wax and wane. In the first half of the
century, prominent moral philosophers such as H. A. Prichard, W. D. Ross,
and A. C. Ewing defended moral intuitionism, but with the emergence, beginning in the 1930s, of non-cognitivist treatments of moral thought and
language, intuitionism fell out of philosophical favor. It is only very recently
that some moral philosophers have been interested in reviving intuitionism in
ethics, and we now find Robert Audi and Russ Shafer-Landau among
intuitionism’s champions.⁶
Notice that moral intuitionism, so defined, is not committed to non-natural
moral properties (as was Moore) or to some version of moral pluralism (as was
Ross). Intuitionism, as here understood, is a purely epistemological position.
Those who reject intuitionism, then, claim that it is not possible for people to
be non-inferentially justified in any of their (non-trivial) moral beliefs. How
might this dispute be resolved?
In ‘Moral Intuitionism Meets Empirical Psychology’, Walter SinnottArmstrong claims that any direct answer to this issue is likely to simply beg the
question on one side or the other, and hence that some indirect strategy is
needed in order to come to grips with the controversy. In particular, SinnottArmstrong claims that recent developments in psychology and brain science
cast considerable doubt on moral intuitionism. In arguing for this claim, he
first develops a set of six principles concerning when non-moral beliefs require
justifying beliefs to back them up. In short, whenever a belief is important, partial, controversial, emotional, subject to illusion, or explicable by dubious
sources, then that belief needs to be backed up by confirming beliefs if the
believer is to be epistemically justified in holding it. By appealing to recent
empirical work, Sinnott-Armstrong argues that moral beliefs of all sorts fall
under one or more of his principles and thus they are in need of support from
other relevant beliefs. If so, then, as he points out, moral intuitionism is incorrect: no moral beliefs enjoy the status of being non-inferentially justified. This
is his strong claim. More cautiously, Sinnott-Armstrong claims that even if
there may be some individuals who, in some contexts, have moral beliefs that
do not require inferential support, still, for educated adults who are well aware
⁶ See also the contributions in Philip Stratton-Lake (ed.), Ethical Intuitionism: Re-evaluations (Oxford:
Oxford University Press, 2002).
14
Terry Horgan and Mark Timmons
of the various possible distorting factors affecting beliefs, no moral beliefs are
non-inferentially justified. Even if moral judgments are not themselves claims
that can be confirmed or disconfirmed entirely by empirical means, including
the methods of science, it does not follow that developments in the sciences,
including biology, psychology, sociology, anthropology, cognitive science, and
brain science, are not relevant to whether a person’s (or group’s) moral beliefs
are epistemically justified. To think they are is typically characterized as
commitment to moral epistemology naturalized.
This last point returns us to what we have called Moore’s independence thesis,
which Darwall found reason to reject. But Panayot Butchvarov (‘Ethics
Dehumanized’) advocates a return to Moorean independence. One dominant
metaethical trend (which we have just seen in Sinnott-Armstrong’s contribution) is moral epistemology naturalized. Another metaethical trend has been
conceptual analysis, often called ‘analytic ethics’, which was preoccupied with
analyzing the meanings of moral terms and the concepts those terms express.
Butchvarov argues that both trends are philosophically misguided. Ethics
naturalized, he claims, is unphilosophical in lacking the kind of supreme
generality and abstractness that is distinctive of philosophical enquiry, taking
human beings to occupy moral center stage, rather than the kind of cosmological ethics we find in Moore, whose views focused on the value of all things in
the universe as a basis for ethical enquiry. Moreover, ethics naturalized lacks
competence in that its scientific pretensions are at odds with how philosophers
go about their business. Analytic ethics, which is explicitly concerned with
armchair, intuitive judgments about meanings, cannot overcome lack of
competence signaled by the philosophical lessons found in Kant, Quine, and
Wittgenstein about conceptual analysis. In light of these failures, Butchvarov
advocates returning to the cosmological orientation of Moore’s ethics which,
he thinks, can be properly understood as avoiding the traditional metaethical
debate between realism and anti-realism, as well as avoiding the battery of
objections to the effect that Moore’s ethics is not relevant to action. Such a
return to a Moorean view of ethics would represent a version of ‘ethics
dehumanized’: cosmological in its focus and thus properly philosophical.
Conclusion: expanding metaethics
If we stand back far enough from the metaethical fray of the past one hundred
or so years to see if we can view general trends or developments with this field,
we notice some basic contrasts between metaethics as practiced throughout
much of the twentieth century and metaethics now. In at least three ways
Introduction
15
metaethics has expanded: in its methodology, in the extent of its recognized
philosophical import, and in the currently available metaethical options. First,
whereas metaethical enquiry went through a phase of rather narrowly focused
a priori linguistic or conceptual analysis, such enquiry today, while still very
much concerned with how to understand the meanings of moral concepts, has
expanded methodologically. Gone apparently is the search for analyses of
moral terms and concepts in terms of necessary and sufficient conditions.
Second, as many of the papers in this collection make clear, the concerns of
metaethical enquiry really extend to all areas of philosophy in that issues concerning normativity are now being raised in philosophy of mind, philosophy of
language, and epistemology. In his paper for this volume, Allan Gibbard suggests that metaethics is properly understood as a branch of what we might call
meta-normativity, thus bringing into close connection the concerns of
metaethics and other areas of philosophy. Finally, the options in metaethics
have expanded at least in the sense that there is a growing ‘centrist’ trend among
defenders of opposing metaethical views to find as much common ground as
possible with their traditional adversaries. This is particularly evident in the
papers by Horgan and Timmons, Barker, and Gibbard.⁷
Metaethics, appropriately expanded, continues to flourish.
⁷ For a recent attempt to wed expressivist themes to moral realism, see David Copp, ‘RealistExpressivism: A Neglected Option for Moral Realism’, Social Philosophy & Policy, 18 (2001).
This page intentionally left blank
1
How Should Ethics Relate to (the Rest of )
Philosophy? Moore’s Legacy
Stephen Darwall
1. Moore and metaethics
From our perspective a century later, Principia Ethica seems a revolutionary
work. It didn’t seem that way at the time, however.¹ Moore’s claims about the
irreducibility of good struck his contemporaries as familiar. And Principia’s
reviewers thought Moore’s objections to naturalists like Mill and Spencer were
‘the standard criticisms.’² In these respects and several others, Moore was, as
Thomas Hurka has brilliantly shown, thoroughly within a tradition of moral
philosophy that ran ‘roughly from the first edition of Sidgwick’s Methods of
Ethics in 1874 to [W. D.] Ross’s Foundations of Ethics in 1939.’³
Moore acknowledged that he wasn’t original in insisting on an irreducible
core of all ethical concepts. He didn’t appreciate the roots of this thought in
eighteenth-century intuitionists like Clarke, Balguy, and Price, not to mention
sentimentalists like Hutcheson and Hume, but Moore did see himself as
¹ G. E. Moore, Principia Ethica, revised edition with the preface to the (projected) second edition and
other papers, edited with an introduction by Thomas Baldwin (Cambridge: Cambridge University Press,
1993), initially published in 1903. References will be placed in the text and include section number along
with page numbers to this edition, thus: (§ 14, 69). For a very insightful account of Moore’s ideas in their
context, to which I am much indebted, see Thomas Hurka, ‘Moore in the Middle,’ Ethics, 113 (2003):
599–628.
² Norman Wilde, Review of Principia Ethica, Journal of Philosophy, Psychology and Scientific Method,
2 (1905): 581–3, at 582. Bernard Bosanquet also termed Moore’s criticisms of Mill ‘not quite original’
(Bernard Bosanquet, Critical Notice of Principia Ethica, Mind, 13 (1904): 254–61, at 261). And
J. S. Mackenzie remarked that Moore’s critical observations of Spencer, Mill, and Green ‘have already
been brought out by other critics’ ( J. S. Mackenzie, Review of Principia Ethica, International Journal of
Ethics, 14 (1904): 377–82, at 378). I am indebted to Hurka’s article for these references.
³ Hurka, ‘Moore in the Middle.’ Hurka describes this tradition further as follows: ‘Moore’s principal
predecessors in this sequence, alongside Sidgwick, were [Hastings] Rashdall, who began publishing on
ethics in 1885; Franz Brentano, whose Origin of Our Knowledge of Right and Wrong appeared in 1889; and
J. M. E. McTaggart, who discussed ethics in his 1901 Studies in Hegelian Cosmology. His successors included
H. A. Prichard, C. D. Broad, W. D. Ross, A. C. Ewing, and later members of the Brentano school such as
Alexius Meinong and Nicolai Hartmann.’
18
Stephen Darwall
following Sidgwick. Sidgwick was, Moore said, the ‘only . . . ethical writer’ who
had clearly seen the irreducibility of ethics’ defining notion.⁴
Nevertheless, ethical philosophy of the last century looked considerably less
to Sidgwick than to Moore.⁵ This is partly due, no doubt, to Principia’s radical
self-presentation—the sense it gives of wiping the slate clean, exposing the
fallacies of all prior ethical thought. But the main reason, I think, is Principia’s
exemplification of the emerging Russell/Moore program of philosophical
analysis, which would prove so influential in twentieth-century philosophy.
This brought ethical philosophy into dynamic relation with more general
philosophical trends in metaphysics and the philosophies of language and
mind in ways that would dramatically affect how ethics was conceived and
practiced as a subject. Among other things, it gave Principia a readership that
extended far beyond consumers of systematic ethical thought, including many
who were relatively unfamiliar with Principia’s place in the tradition Hurka has
described.
Principia Ethica might fairly be called the first work in analytical ethical
philosophy. As we’ve noted, other writers had considered and discussed the
nature and content of central ethical ideas and terms. Perhaps most prominently for Moore’s contemporaries, Sidgwick had claimed that ‘ordinary moral
or prudential judgments . . . cannot legitimately be interpreted as judgments
respecting the present or future existence of . . . any facts of the sensible world.’
Sidgwick believed this was because ‘the fundamental notion represented by the
word “ought” . . . which such judgments contain expressly or by implication,’ is
‘essentially different from all notions representing facts of physical or psychical
⁴ Samuel Clarke, A Demonstration of the Being and Attributes of God (London: J. Knapton, 1705) and
A Discourse Concerning the Unalterable Obligations of Natural Religion (London: J. Knapton, 1706). Both
can be found in The Works of Samuel Clarke, 4 vols. (London: J. & P. Knapton, 1738), facsimile edition
(New York: Garland, 1978); John Balguy, The Foundations of Moral Goodness (London: John Pemberton,
1728), facsimile edition (New York: Garland, 1976); Richard Price, A Review of the Principal Questions in
Morals, originally published in 1758, ed. D. D. Raphael (Oxford: Clarendon Press, 1974). For a discussion
of the relevance of earlier British intuitionists to Moore, see A. N. Prior, Logic and the Basis of Ethics (Oxford:
Clarendon Press, 1949).
I mention Hutcheson and Hume because Hutcheson argued that the concept of moral goodness cannot
be reduced to natural goodness (and so requires a special sense) and Hume held that moral judgments do not
concern any matter of fact and that an ‘ought’ cannot follow from any ‘is’. Francis Hutcheson, An Inquiry
into the Original of our Ideas of Virtue (London, 1725), relevant passages in D. D. Raphael (ed.), The British
Moralists: 1650–1800, 2 vols. (Oxford: Clarendon Press, 1969) and L. A. Selby-Bigge, British Moralists, 2
vols. (Indianapolis: Bobbs-Merrill, 1964), an electronic version is available through InteLex Past Masters
(http://library.nlx.com); David Hume, A Treatise of Human Nature, ed. with analytical index by
L. A. Selby-Bigge, 2nd edn. with text revised and variant readings by P. H. Nidditch (Oxford: Clarendon
Press, 1978).
⁵ The Philosopher’s Index, which has catalogued philosophical articles since 1940, has 120 entries in
which both ‘Sidgwick’ and ‘ethics’ appear, and well over 350 with ‘Moore’ and ‘ethics’ (subtracting those concerned with other Moores, like A. W. Moore, Asher Moore, etc.).
How Should Ethics Relate to Philosophy?
19
experience’ (a claim to which we shall return in due course).⁶ Clear as he was
about this, however, Sidgwick did not give analysis of ethical terms and ideas
the methodological emphasis that Moore would.
The methodological priority of analysis is front and center in Principia.
Moore begins the Preface by citing the failure to appreciate its importance as
the major obstacle to progress in prior ethical thought.
[I]n Ethics, as in all other philosophical studies, the difficulties and disagreements, of
which its history is full, are mainly due to a very simple cause: namely to the attempt to
answer questions, without first discovering precisely what question it is that you desire
to answer.(P, 33)
Moore took his own advice, devoting much of the first half of Principia to the
analysis of ethical concepts and to illustrating how failures of analysis, specifically, what he called the ‘naturalistic fallacy,’ had confounded clear ethical
thinking.
As the passage just quoted makes clear, Moore held the analysis of ethical
terms and questions to be but one instance of a general philosophical program.
The emergence of metaethics (also ‘analytic’ or ‘critical’ ethics) in the twentieth
century was thus part of a general ‘analytic,’ later ‘linguistic,’ turn in anglophone philosophy that was itself due partly to Moore.⁷
Metaethics’ development as a potentially freestanding area of philosophical
enquiry can be traced to Moore, therefore, in a way that it cannot be even to
such analytically minded predecessors as Sidgwick. After Moore, it became
possible to pursue metaethical questions completely independently of issues of
normative ethics and possible also to specialize in this area without any interest
in normative issues whatsoever—or, at least, any philosophical interest.
Charles Stevenson, for example, introduced Ethics and Language in 1944 by
saying that he was concerned ‘with a narrowly specialized part’ of ethics, whose
purpose is ‘to send others to their tasks with clearer heads.’⁸ At one point the
‘linguistic turn’ became so dominant that a philosopher could expect to raise
no eyebrows by introducing a book titled Modern Moral Philosophy as follows:
‘[A] moral philosopher . . . thinks and speaks about the ways in which moral
⁶ Henry Sidgwick, The Methods of Ethics, 7th edn. (London: Macmillan, 1967), 25. All of Sidgwick’s
chapter on ‘Ethical Judgments’ is worth noting in this connection.
⁷ Russell and Moore’s analytic turn was in part a critical response to Idealism. On this, see Peter Hylton,
Russell, Idealism, and the Emergence of Analytical Philosophy (Oxford: Clarendon Press, 1990). Particularly
relevant here is Moore’s ‘The Refutation of Idealism,’ Mind, 12 (1903): 433–53, which was published the
same year as Principia.
⁸ Charles Stevenson, Ethics and Language (New Haven: Yale University Press, 1965), 1 (emphasis added).
This was originally published in 1944. It is worth noting that Stevenson was initially attracted to philosophy
by Moore and Wittgenstein when he was studying English literature at Cambridge in the early 1930s.
20
Stephen Darwall
terms, like “right” and “good” are used by moralists when they are delivering
their moral judgments.’⁹ This not only proclaimed the legitimacy of metaethics
as an independent philosophical area, it read normative ethical reflection out
of the philosophical canon altogether.
That metaethical analysis should dominate ethical philosophy was hardly
Moore’s intention. It is consistent with Principia’s methodological principles,
indeed, that metaethics’ philosophical significance is entirely instrumental, a
clarificatory preliminary to answering the normative ethical questions that give
ethical philosophy its real task. Principia begins by distinguishing ‘two kinds of
question, which moral philosophers have always professed to answer’: first,
‘What kind of things ought to exist for their own sakes [alternatively: are good
in themselves or “intrinsically valuable”]?’ and second, ‘What kind of actions
ought we to perform?’ (P, 33–4). Moore argues that, contrary to what many
philosophers have implicitly believed, questions of the first kind can be analyzed no further, that the concept of intrinsic good is simple and unanalyzable.
And he argues that questions of the second kind are analyzable, specifically, in
terms of questions of the first kind and empirical causal questions: ‘To assert
that a certain line of conduct is, at a given time, absolutely right or obligatory
is obviously to assert that more good or less evil will exist in the world, if it be
adopted than if anything else be done instead’ (§ 17, 77). From this, Moore
infers that in order to answer questions of the first, intrinsic value, kind, ‘no
relevant evidence whatever can be adduced’ (P, 34). All we can do is to make sure
that we have the question of value clearly before our minds and not some other
naturalistic or metaphysical issue with which philosophers have mistakenly
confused it. But since questions of right conduct are analyzable as a complex of an
empirical causal question together with a question of intrinsic value, what a person ought to do does admit of evidence, namely, ‘causal truths’ regarding actions
in the agent’s power and ‘ethical truths of our first or self-evident class’ (P, 34).
This sets up Principia’s strategy after chapter I. In chapters II and III, Moore
discusses earlier philosophers’ misbegotten attempts to support normative
intrinsic value claims with empirical naturalistic evidence, which he criticizes
for committing the ‘naturalistic fallacy.’ Here Moore’s diagnosis is that naturalistically minded philosophers like Mill and Spencer have simply failed to get
the ethical question of intrinsic value they purport to be addressing clearly in
view and confused it with naturalistic issues that are both distinct from it and
incapable of shedding any light to answer it. In chapter IV, Moore pursues a
similar strategy with regard to metaphysical ethicists like Kant and Green who,
in his view, had been guilty of essentially the same error, this time confusing the
⁹ W. D. Hudson, Modern Moral Philosophy (Garden City, NY: Doubleday, 1970), 1.
How Should Ethics Relate to Philosophy?
21
ethical question of intrinsic value with a distinct metaphysical issue, for example,
one concerning a transcendental will. In chapter V, ‘Ethics in Relation to
Conduct,’ Moore reaffirms his (agent-neutral) consequentialist analysis of
right or ought to do, arguing that philosophers who have advanced
fundamentally agent-relative normative principles of conduct, whether egoists
or deontologists, have been guilty of conceptual confusion or incoherence.¹⁰
He then draws out what he takes to be the practical consequences of an analytical consequentialism, arguing that, as a practical matter, individuals are
nonetheless always best advised to follow commonsense moral ‘rules which are
both generally useful and generally practised’ (§ 99, 213). Finally, in chapter
VI, Moore sketches his normative theory of intrinsic value, according to which
‘by far the most valuable things, which we know or can imagine, are certain
states of consciousness, which may be roughly described as the pleasures of
human intercourse and the enjoyment of beautiful objects’ (§ 113, 237).
This left Moore with a remarkable normative position. On the one hand, he
championed the unparalleled intrinsic value of friendship and aesthetic appreciation together with an analytical consequentialism of the right according to
which ‘it is only for the sake of these things—in order that as much of them as
possible may at some time exist—that any one can be justified in performing
any public or private duty’ (§ 113, 238, emphasis added). These were the
ethical ideas that thrilled Bloomsbury. On the other, Moore held that, practically speaking, everyone should always conform to commonsense morality.
(This part, Bloomsbury quietly ignored.¹¹)
It is worth noting that Moore’s premier intrinsic values were organic wholes.
Moore calls them ‘states of consciousness’—the ‘pleasures of human intercourse’
and ‘enjoyment of beautiful objects’—but it is important to him (and to his critique of idealist accounts of the mind) that consciousness involves a relation
between something mental and some object that is (in these cases, anyway)
outside the mind. For Moore, the relevant conscious states are pleasures, but
they are pleasures taken in the existence of something really existing outside the
pleasurable experience itself. In order for the pleasure of human intercourse to
exist, there must really be human interaction that the participants actually
enjoy. Similarly, in order for someone to enjoy a beautiful object, it must be the
case that there really exists such an object, that it is really beautiful, and that
pleasurable experience comes from a rapport with that object and its beauty.
Moore did not deny that either relatum by itself had some intrinsic value, but
he held that whatever value unappreciated beauty or the pleasurable regard of
¹⁰ We will consider these claims of Moore’s in more detail below.
¹¹ Or so said Keynes. See John Maynard Keynes, ‘My Early Beliefs,’ in The Collected Writings of John
Maynard Keynes, vol. x (Cambridge: Macmillan, 1972), 435.
22
Stephen Darwall
‘fool’s beauty’ might have is ‘so small as to be negligible’ in comparison to the
complex whole (§ 113, 237). Moore’s premier intrinsic values thus exhibit his
‘principle of organic unities’: ‘The value of a whole must not be assumed to be the
same as the sum of the values of its parts’ (§ 18, 79).
Hence Moore’s normative theory of intrinsic value was organicist and
pluralist. He held that there was more than one intrinsically valuable kind, and
he thought that these included organic wholes, whose value substantially
exceeded that of the parts of which they were composed. In value theory,
Moore’s took his major disagreement to be with monistic theories such as
hedonism that held that the good was simple. Moore believed that value or
good, as he called it, was a simple property, but he believed that the good, that
is, the things that are good or have intrinsic value, is complex in the ways just
indicated. One form of the naturalistic fallacy consisted in confusingly taking
the simplicity of good for simplicity of the good. Some might mistakenly infer
that the good is simple, say, that it is identical with pleasure, because they
correctly saw that pleasure is simple (§ 12, 64) and mistakenly identified good
with pleasure (perhaps dimly perceiving that good is simple, but failing to see
that it is a distinct simple from pleasure).
In Principia, therefore, the main rationale for metaethical analysis is to prepare the way for accepting Moore’s normative ethics by clearing away the main
source of support for its competitors. Non-consequentialist theories of right
(more precisely, non-agent-neutral consequentialist theories), such as deontology and egoism, are to be rejected for their failure to appreciate the conceptual
tie rightness of action must have to intrinsic value. And monistic views of
intrinsic value can be seen to lack support if they are advanced on the basis of
mistaking some property that might characterize the good with good itself.
It is thus ironic that Principia’s main effect was to help create an environment
in which metaethics could be pursued independently of normative ethics as a
freestanding philosophical area, one which would even for a time at mid-century
claim exclusive legitimacy as genuine ethical philosophy. At the same time,
however, although in Principia Moore pursued metaethics in service of
normative ethics, he did not think that normative ethics had anything positive
to learn from metaethics. More precisely, he didn’t think that its purely ethical
part—normative theory of the good—could receive positive support from
metaethical theory. He did believe that metaethics bears positively on normative theory of the right, since he thought that analysis reveals that the concept
of right is a complex composed of empirical causal concepts and the concept of
intrinsic value. This enables us to identify the empirical causal issues and issues
of intrinsic value we are to look to as evidence of what a person should do and
to rule out as incoherent agent-relative normative theories of right like egoism
How Should Ethics Relate to Philosophy?
23
and deontology. But when it comes to theory of the good, the irreducible
core of normative ethical theory according to Moore, metaethics can play no
more than a destructive role, showing only how attempts to ground intrinsic
value claims in putative analyses, or indeed, to give any justificatory reasons for
intrinsic value claims at all, must always come to grief. Metaethical reflection
can never count in favor of any (wholly) ethical claim. It can only properly
count against attempts to count it in favor.
Now this picture is actually quite congenial to the philosophical environment
that emerged after the heyday of analytic metaethics, during the great resurgence and expansion of normative ethical theory in the 1970s.¹² With, on the
one hand, the linguistic turn and philosophical analysis on the wane owing to
Quinean objections to the analytic/synthetic distinction, and, on the other, an
impressive example of systematic normative theory in Rawls’s Theory of Justice,
an atmosphere emerged in which normative theory could be pursued without
apology and independently of metaethical reflection. Rawls proclaimed the
‘Independence of Moral Theory,’ arguing that normative theory might
proceed entirely on the basis of considered ethical judgments or intuitions and
without concern about more general philosophical foundations, or connections to other areas of philosophy.¹³ In retrospect, this attitude was not really so
different from Moore’s. It extended Moore’s view that metaethics has no
positive relevance to value theory or to normative ethical theory in general.
Normative theories of the good, right, virtue, and so on could be constructed
on a base of considered judgments without much of a glance at other, nonevaluative philosophical areas.
In another way also, the indifference of some normative theorists to
metaethics in the 1970s was substantially similar to the ‘that’s not my department’ view analytic metaethicists had earlier sometimes taken toward normative ethics, but from the other direction. Both believed that their part of ethical
philosophy, normative ethics or metaethics, respectively, could be pursued successfully without any attention to the other.
¹² For a short historical overview, see Stephen Darwall, Allan Gibbard, and Peter Railton, ‘Toward Fin de
Siecle Ethics: Some Trends,’ Philosophical Review, 101 [1992]: 115–89. Reprinted in S. Darwall, A. Gibbard,
and P. Railton (eds.), Moral Discourse and Practice: Some Philosophical Approaches (New York: Oxford
University Press, 1997).
¹³ John Rawls, A Theory of Justice (Cambridge, Mass.: Harvard University Press, 1971); ‘Independence of
Moral Theory,’ Proceedings and Addresses of the American Philosophical Association, 48 (1975): 5–22,
reprinted in John Rawls, Collected Papers, ed. Samuel Freeman (Cambridge, Mass.: Harvard University
Press, 2001), 286–302. Rawls’s method in Theory of Justice, however, might better be viewed as a ‘wide’
reflective equilibrium that takes in, among other things, the sorts of broadly metaethical considerations
Rawls presents in his ‘Kantian Interpretation.’ On the distinction between narrow and wide reflective equilibrium, see Norman Daniels, ‘Wide Reflective Equilibrium and Theory Acceptance in Ethics,’ Journal of
Philosophy, 76 (1979): 256–82, also in Norman Daniels (ed.), Reading Rawls (Stanford, Calif.: Stanford
University Press, 1989), 253–82.
24
Stephen Darwall
2. Why should normative ethics be part of philosophy?
Why then should normative ethics and metaethics be in the ‘same department’?
Or to put it another way, what reason, other than tradition, is there for normative ethical theory to be conducted in a philosophy department? Metaethics, the
philosophy of ethics, might seem to have roughly the same relation to
normative ethics that the philosophy of physics does to physics. And although
there was no way to sharply demarcate physics from metaphysics before the rise
of the experimental method, no one these days would argue for a merger
between the departments of physics and philosophy. So why should things be
any different with ethics?¹⁴
I believe that the case of ethics is different and that the practice of ethical
theory should reflect this fact. Unlike the natural and social sciences, ethics has
not spun off from philosophy, nor should it. The various sciences became
autonomous disciplines when the experimental method gained wide acceptance as the only reliable way of confirming theory about their respective subject
matters.¹⁵ Disagreements in the philosophy of science or in metaphysics or
epistemology leave this consensus largely unaffected. Whether theories are
interpreted realistically, as the best explanation of observed experimental
results, or instrumentally, as the best device for predicting them, is mostly
irrelevant to scientific practice, which proceeds largely independently of these
philosophical disputes.
Nothing like this consensus has emerged in the case of ethics, nor is it likely
to. Unlike the sciences, ethics’ subject matter is itself the focus of a lively philosophical debate to which we can envision no end. Cartesian skepticism
notwithstanding, there is sufficient consensus about the phenomena the
sciences are to explain that they can proceed without attention to fundamental
metaphysics and epistemology. In my view, however, this is not the case with
ethics. As I shall attempt to illustrate in the final section, there is no consensus
regarding the normativity of (various) ethical propositions and judgments (nor
even about what normativity or normative judgment itself is). These are matters of metaethics—of the ‘metaphysics of morals,’ ethical epistemology, and
the philosophy of language and mind as they relate to ethical talk and ethical
states of mind. And once they are brought in, other philosophical issues and
areas come in their wake—for example, in the philosophy of action, concerning
¹⁴ In the next several paragraphs I draw on my ‘Why Ethics is Part of Philosophy: A Plea for Philosophical
Ethics,’ in Klaus Brinkmann (ed.), The Proceedings of the World Congress of Philosophy (Bowling Green, Oh.:
Philosophy Documentation Center, 1999), 19–28.
¹⁵ These are matters of degree, of course. Scientific questions are not entirely independent of issues in the
philosophy of science. My claim is that theoretical physics is independent of the philosophy of physics to a
larger degree than ethical theory is independent of metaethics.
How Should Ethics Relate to Philosophy?
25
the nature and freedom of the will and action. When these broader philosophical
issues are in the background, I shall claim, it is impossible to completely seal
them off from intelligent normative discussion and debate. My conclusion will
be that, although metaethics and normative ethics focus on different issues,
systematic ethical philosophy thrives when these areas are brought into
dynamic relation and pursued in an integrated way we might call ‘philosophical
ethics’—framing normative ideals we can accept in light of both the best normative reasons as we see them and an adequate philosophical understanding of
their subjects and of the possibilities for knowledge or justified acceptance in
this area.¹⁶
It is only in this century that we have come to distinguish sharply between
metaethics and normative ethics. That, again, is Moore’s legacy. Earlier
systematic ethical thinkers pursued issues of both kinds, attempting to integrate
them into a coherent overall view. The past hundred years has seen a sharpening of issues and insights in both areas but also a degree of specialization
and compartmentalization that has sometimes closed off opportunities for
mutually beneficial interaction and the development of more comprehensive
philosophical ethical outlooks.
We are, of course, right to distinguish metaethics from normative ethics. It
is possible to combine any given normative view with a variety of metaethical
positions. Among consequentialists, we have non-cognitivist prescriptivists
(Hare), theological voluntarists (Berkeley), naturalists (Mill), and intuitionists
(Sidgwick and Moore). Nonetheless, there do seem to be affinities between
metaethical and roughly corresponding normative ethical theories. Metaethical
naturalists have almost always been consequentialists, for example, although,
as we’ve noted, a consequentialist need not be a naturalist. Deontologists, on
the other hand, require a metaethical account of moral obligation that can
explain how the right can diverge from the beneficial, and the wrong from the
harmful. Some deontologists have been theological voluntarists. Some have
been rational intuitionists, holding, like Clarke, Price, Prichard, or Ross, that
we can perceive the truth of fundamental deontological moral principles
immediately by reflection. And some have followed Kant in maintaining that
deontological normative principles are grounded in the structure of deliberation of a free rational agent. These associations seem far from accidental.
As I mentioned, a major reason that normative ethical theories can never
completely dispense with metaethics is that the normativity of their subject
matters is often in dispute, along with the character of their respective forms of
¹⁶ For a more extended development of this idea, see my Philosophical Ethics (Boulder, Colo.: Westview
Press, 1998).
26
Stephen Darwall
normative judgment. In the next section, I will discuss Moore’s views in this
light and argue that, while Moore is right that ethical concepts all have an
irreducible core, he is wrong to identify it with intrinsic value. Here, I shall
urge, we really would do better to look to Sidgwick rather than Moore.
The central ethical notion is that of normativity or normative reason, which we
require to understand both the notion of good and that of right action. Moore
held intrinsic value to be a normative notion, but I believe that he mislocated
its normativity. Moreover, for Moore, intrinsic value is the only genuine
normative notion. Although he was a normative pluralist in the theory of
value, he was a metaethical monist. In this, I shall argue, Moore was dramatically mistaken. In principle, there are as many normative notions as items
(action and attitudes) that can be normatively regulated—the desirable, the
estimable, the dignified (that worthy of respect), and so on. Most regrettable
here was Moore’s failure to understand reasons for action and ought-to-do’s,
which he reduced to the ought-to-be’s he identified with intrinsic value. This,
it seems to me, is a metaethical failure whose consequences have been at least as
serious as any of the ‘naturalistic fallacy.’ In the final section, I will move from
Moore to the contemporary scene in order to illustrate how debates in a variety
of areas of normative ethics cannot be settled independently of metaethical
disputes about the normativity of their respective subjects.
3. Moore and normativity
By ‘good’, Moore says he means the ‘unique object—the unique property’ we
have before our minds when we say that something has ‘intrinsic value,’ ‘intrinsic worth,’ or ‘that a thing ought to exist’ (§ 13, 68). For Moore, therefore,
something’s being intrinsically good is identical to its being such that, by virtue
of its intrinsic nature, it ought to exist: ‘[W]hen we assert that a thing is good,
what we mean is that its existence or reality is good’ (§ 70, 171).¹⁷ It follows
that what most fundamentally possesses intrinsic value for Moore is a state of
affairs. Or, to be more precise, the normative proposition entailed by a thing’s
having intrinsic value is that the state of its existing ought to be.¹⁸
Now there appear to be ways of valuing something intrinsically whose
content does not reduce to, nor even arguably entail, that that thing ought to
¹⁷ This is reinforced by Moore’s famous ‘isolation test.’ A’s being better than B amounts to its being the
case that, as Moore put it in Ethics, ‘it would be better that A exist quite alone than that B exist quite alone’
(G. E. Moore, Ethics (New York: Oxford University, 1965), 39).
¹⁸ In this and following paragraphs, I draw on my ‘Moore, Normativity, and Intrinsic Value,’ Ethics 113
(2003): 468–89.
How Should Ethics Relate to Philosophy?
27
exist. Kantian respect for a person as an end in herself seems a form of intrinsic
valuation, but it cannot be reduced to the proposition that the person ought to
exist, and it may not even entail it. It may not be surprising that respect is hard
to fit into Moore’s consequentialist framework but so also is the kind of intrinsic
valuation involved in benevolent concern for a person (or other creature) for
her own sake. In caring for someone we properly want her to exist if, but only
if, existing is good for her.¹⁹ The difficulties of accommodating these within
Moore’s scheme are some indication already of idiosyncrasies in his theory.
Moore says that what is intrinsically valuable is what ‘ought to exist for its
own sake.’ But how are we to understand this? Oughts gain their sense from
norms; only what can be regulated by norms can be subject to normative judgment.²⁰ True, we can say of some event that it ought to happen and simply
mean that its happening follows from the laws of nature and initial conditions
(as we believe them to be), as in, ‘the car ought to start.’²¹ But there is nothing
normative in such a statement.
Now this doesn’t mean that the only oughts are ought-to-do’s. Normative
guidance need not be voluntary, and there is much that we judge normatively
and regulate by norms other than action, for example, reasoning, beliefs,
choices, emotions, responses, feelings, intentions, and attitudes.²² But it does
mean that there cannot be a brute ought-to-be that is genuinely normative
(unlike the ought in the ‘car ought to start now’). The state of something’s
existence can be the object of various attitudes, and we can sometimes say that
a state ought to be, thereby expressing a normative judgment about that
attitude. Most obviously, one can desire that something exist. So we might say
that that state ought to be, meaning that its obtaining ought to be desired, that
it is desirable. Or we might mean that that state is worth caring about. Or that
it is worthy of concern from a particular perspective, say, the moral point of
view. In the final analysis, we must understand ought-to-be’s as elliptical and
underspecified, requiring completion by reference to something that can be
normatively regulated: some attitude or agent-state.
¹⁹ This is an important theme of Elizabeth Anderson’s in Value in Ethics and Economics (Cambridge,
Mass.: Harvard University Press, 1993), 26–30. I defend an account of welfare inspired by this insight in
Welfare and Rational Care (Princeton: Princeton University Press, 2002).
²⁰ A fortiori, only what exists can be subject to normative judgment. This is, I think, what lies behind
Prichard’s somewhat curious remark that ‘only something which is can be something which ought, or ought
not, to exist. To say, e.g., that a feeling of generosity which I am not having “ought” to exist is to say nothing,
just because ex hypothesi there is nothing here for “being something which ought to exist” to be attributed to’
(H. A. Prichard, ‘Moral Obligation,’ in Moral Obligation and Duty and Interest (London: Oxford University
Press, 1968), 163).
²¹ On this point, see Roger Wertheimer, The Significance of Sense: Meaning, Modality, and Morality
(Ithaca, NY: Cornell University Press, 1972).
²² I am indebted here to Allan Gibbard, Wise Choices, Apt Feelings (Cambridge, Mass.: Harvard
University Press, 1990).
28
Stephen Darwall
If intrinsic value is to be a normative notion, then, it will have to be
interpreted in terms of a normative connection to some valuing attitude or
agent-state. The distinction between intrinsic and extrinsic value, between
values that depend only on the intrinsic properties of what has it, and those that
depend also on extrinsic properties, must consequently be drawn within such
norms. When we say that something is intrinsically valuable solely because of,
or in virtue of, its intrinsic nature, we should understand the relevant ‘because’
and ‘in virtue of ’ normatively, that is, as asserting that its intrinsic nature
provides reasons for so valuing it.²³
There are good reasons for Moore to hold that intrinsic value is an explicitly
normative notion, additional even to the obvious one that it is hard to see how
it could be intrinsically relevant to ethics otherwise. The most important one
was identified by Frankena, namely, that it is hard to see what else underlies the
open question and other arguments Moore employs to argue that intrinsic
value cannot be reduced to naturalistic or metaphysical notions. Frankena
writes: ‘[T]o my mind, what makes ethical judgments seem irreducible to
natural or metaphysical judgments is their apparently normative character.’²⁴
But if this is right, and there cannot be a brute ought, then, as Frankena also
pointed out, intrinsic value must be, not a simple, but a complex notion that
embeds the idea of normativity or ought within it. Roughly, to be intrinsically
valuable must be to be something that ought to be valued (that there is reason
to value) just in light of its intrinsic features.²⁵ And it will follow further that
Sidgwick, not Moore, was right about the irreducible core of all ethical notions.
The central idea contained in all ethical notions is not intrinsic value but
normativity: ‘the fundamental notion represented by the word “ought”.’²⁶
So although Moore was right that ethical notions are irreducible to non-ethical
ones, he was wrong in thinking that this is because they all have the idea of
intrinsic value as their irreducible core. If the concept of intrinsic value is to be
normative, as it must be to make the open question argument (and the naturalistic fallacy) plausible, then it must include the idea of normativity or ought.
²³ We should say, as W. D. Falk puts it, that it is something that would be so valued on a proper review of
its intrinsic properties (W. D. Falk, ‘Fact, Value, and Nonnatural Predication,’ in Ought, Reasons, and
Morality (Ithaca, NY: Cornell University Press, 1986) ). See also Wlodek Rabinowicz and Toni RønnowRasmussen, ‘The Strike of the Demon: On Fitting Pro-attitudes and Value,’ Ethics, 114 (2004): 391–423.
²⁴ William Frankena, ‘Obligation and Value in the Ethics of G. E. Moore,’ in Paul A. Schilpp (ed.), The
Philosophy of G. E. Moore (LaSalle, Ill.: Open Court, 1942), 102.
²⁵ Cf. Scanlon’s ‘buck-passing’ account of value, according to which ‘to call something valuable is to say
that it has other properties that provide reasons for behaving in certain ways with regard to it.’ That the relevant reasons must be for action, as opposed also to attitudes of other kinds, seems too narrow, but otherwise
the ideas are quite similar (T. M. Scanlon, What We Owe to Each Other (Cambridge, Mass.: Harvard
University Press, 1998), 96).
²⁶ Methods of Ethics, 25.
How Should Ethics Relate to Philosophy?
29
Moore was also wrong in thinking that the concept of right conduct, or what
one ought to do, can be reduced to that of intrinsic value, or what ought to be.
Moore believed, again, that to assert that a given action is right, something one
ought to do, ‘is obviously to assert that more good or less evil will exist in the
world, if it be adopted than if anything else be done instead’ (§ 17, 77). But if
this were true, then deontological claims to the contrary, for example, that one
should not torture even if it would bring about greater good, would not be
false, so much as incoherent. And, indeed, this is what Moore claimed, arguing
as follows.²⁷
1. To say that an action at a time is an agent’s ‘absolute duty’ is to say ‘the
performance of that action is unique in respect of value.’
2. A dutiful action cannot possibly be unique in respect of value in the sense
of being the only valuable thing (since ‘every such action would [then] be
the sole good thing, which is a manifest contradiction’).
3. A dutiful action cannot possibly be unique in respect of value in the sense
‘that it has more intrinsic value than anything else in the world’ (‘since
every act of duty would then be the best thing in the world, which is also
a contradiction’).
4. Therefore, the only sense in which the performance of an action can possibly be unique in respect of value is ‘that the whole world will be better,
if it be performed, than if any possible alternative were taken’(§ 89, 197).
From 1 and 4 together, it follows that an action’s being an agent’s absolute
duty is equivalent to its being the case that the world would be better were the
action performed than if the agent were to do anything else she could. Any nonconsequentialist duty or ought-to-do claim to the contrary is self-contradictory.
Moore’s argument assumes that action’s essence is instrumental, that a doing
is simply the intentional bringing about of a state. Premise 1 is uncontroversial
(at least in one direction) if we understand it as saying that if an agent has an
absolute duty to do something then that means the action would be the best
thing for her to do, that it is the best act of those available. In 2 through 4, however, most obviously, in 4, Moore slides to a different kind of evaluation—
namely, his broader category of intrinsic value. In this sense, an act’s having
intrinsic value consists in its being true that the state of its being performed
ought to be. Now 4 allows for the possibility that an act can have intrinsic value
in this sense. But even if an act has intrinsic value as an existent, Moore’s argument clearly assumes that its value as an action is purely instrumental. Its value
²⁷ In Principia Ethica. Moore changed his view in Ethics to hold that consequentialism is a synthetic a priori
truth rather than an analytic one (G. E. Moore, Ethics (New York: Oxford University Press, 1966), 89).
30
Stephen Darwall
as an action, the sense in which one’s absolute duty is the best available act, is its
instrumental value in bringing about intrinsically valuable states, including,
perhaps, the state of the act itself being performed.
Moore’s reduction of right to good, of ought-to-do’s to ought-to-be’s, thus
depends upon a purely instrumental view of action. What action is for is the
production of intrinsically valuable states. The concept of a reason for acting,
therefore, reduces to an empirical part, which is concerned with the consequences of acts within the agent’s power, and an ethical part, which is
concerned with the intrinsic value of those consequences. On this picture, the
ethical aspect of the question of what to do is entirely agent-neutral. It is the
question of what states the world should contain, viewed as from nowhere.
Even if the commonsense description of an act or an agent’s reason is agent-relative,
the question of whether the state of the act’s being performed, or being
performed for that reason, ought to exist is not itself agent-relative, but agentneutral. If we ask, for example, whether someone’s caring for her children, or
benefiting a child because it is hers, ought to exist for its own sake, the grounds
for answering that question will not themselves be agent-relative. If it is a good
thing that Jesse helps his children then it will likewise be a good thing that
Mervis helps hers.²⁸
Moore to the contrary notwithstanding, however, ‘What to do?’ is in its
nature an agent-relative question. It is the question of what the agent should do.
Moore doesn’t deny this in a sense. It is just that he holds that its agent-relativity
is restricted to an instrumental, causal question concerning the consequences of
all actions in the agent’s power. These are not, however, normative issues.
According to Moore, the only ethical issues that agents ever face are agent-neutral:
what states should be?
Moore’s analytic consequentialism can be shown to be mistaken by a version
of the very same open question argument he himself uses to illustrate the naturalistic fallacy. When a deontological theorist of the right asserts that one ought
not to torture another person even if that would have better consequences, she
is not saying, incoherently, that, in the situations where torturing would
produce better consequences, not torturing would actually have better consequences. Neither is she just saying that the state of someone’s torturing has
intrinsic disvalue. She is saying that even in the case where one could prevent,
say, a morally equivalent torturing, it would be wrong, one ought not, to
torture another oneself. But if this is a coherent claim, as it surely seems to be,
then analytic consequentialism cannot possibly be right.
²⁸ This is not to deny that some evaluations of what should be are evaluator-relative or even that the
agent’s own evaluations of what should be are distinctively relevant to what he should do.
How Should Ethics Relate to Philosophy?
31
Of course, consequentialism might still be correct as a substantive normative
theory of the right. However, I believe that it will appear so only so long as we
are in the grip of the same instrumental picture of action that Moore plainly
assumed. So long as we view the question of what to do entirely from the
perspective of how it would be best for the world to be, consequentialism of
some sort will seem inevitable. But agents are not simply producers of world
states. Many of the most important practical questions concern how to relate
to the world from within it, most especially how to relate to other agents. That
this is a central feature of the ethical context has been stressed by deontological writers, like W. D. Ross, and it is an important part of the Kantian and
contractualist pictures. Thus Ross’s list of prima facie duties includes those we
have by virtue of various special relationships we stand in to others (parent,
trustee, etc.) and special relations arising from past actions (promise, contract,
restitution, gratitude), as well as general relations of reciprocity, beneficence,
and non-maleficence we can stand in to any person (or, in the latter two
cases, to any sentient being).²⁹ Kant and contractualists, on the other hand,
see standards of moral right and wrong as principles that define expectations
moral persons rightly have of each other as mutually respecting free and equal
agents.
The affinity between Moore’s consequentialism, his instrumental conception of action, and his reduction of the right to the good is a deep feature of his
philosophical ethics. Ultimately, Moore fails to appreciate the idea that really
drives his arguments for the irreducibility of the ethical, including the open
question argument—namely, the concept of normativity or ought. For any
property that is characterized non-normatively, we can ask whether (or coherently deny that) things with that property are good because we tacitly assume
that good is an explicitly normative property. And Moore fails to appreciate the
implications of normativity for the instrumentalist picture of action that
underlies his consequentialism. The concept of ought and explicitly normative
notions that contain it are ideas we have use for because we are self-reflective
beings who are capable of normative guidance. Only a being who can be aware
of her own attitudes and take a critical perspective towards them can grasp the
distinction between what she actually values and what she ought to value—
what is valuable. Relatedly, as Butler pointed out, only a being who can distinguish between the strength of a motivational state and its authority—between
how much she actually desires something and how much she should desire it—
can be an agent.³⁰ Our finding the open question argument persuasive is thus
²⁹ W. D. Ross, The Right and the Good (Oxford: Clarendon Press, 1963).
³⁰ Bishop Butler, Sermons, ed. Stephen Darwall (Indianapolis: Hackett Publishers, 1983), II. 14, p. 39.
32
Stephen Darwall
an expression of our own freedom.³¹ As free agents, we cannot possibly accept
the Moorean definition of action as simply for bringing about valuable states of
the world. That would involve a kind of ‘bad faith,’ as though we weren’t free
even to consider whether there are agent-relative reasons for action, ought-todo’s (for example, those having to do with our relations to other persons) that
cannot be given a Moorean reduction.³²
Ultimately, then, Moore failed to grasp that ought-to-do’s can be independent of ought-to-be’s—that reasons for acting can be independent from reasons
we have to desire or value states of the world—because he failed to appreciate
the idea of normativity and its full implications for ethics. Intrinsic value
cannot be the fundamental ethical notion from which all others derive. What
makes it an ethical notion in the first place is its containing the idea of normativity or ought. But in this it is no more or less fundamental than the idea of
right, what we ought to do; the estimable, what we ought to esteem; the
desirable, what we ought to desire, and so on. Working out the relations between
different normative notions is a complicated matter.³³ As I will suggest
presently, some of this can be done at the metaethical level in a way that can fit
with various normative claims. But it is also partly, no doubt, a substantive
normative matter.
4. Philosophical ethics: why normative ethics
should be part of philosophy
The foregoing discussion shows that the relation of metaethics to normative
ethics is more complex than Moore thought. There are normative disputes
about the right, for example, that cannot be settled without attention to
metaethical issues about its normativity that Moore thought he had decisively
settled. Indeed, much of the debate between consequentialists and their critics
over the past two decades has had an important metaethical element, with
consequentialists arguing that agent-relative restrictions can be given no philosophically satisfying rationale, and their opponents trying to do precisely that,
³¹ That the appeal of the open question argument is deeply related to freedom is an important theme of
Connie Rosati’s ‘Naturalism, Normativity, and the Open Question Argument,’ Nous, 29 (1995): 46–70. See
also, Stephen Darwall, ‘Internalism and Agency,’ Philosophical Perspectives, 6 (1992): 155–74, and
‘Autonomist Internalism and the Justification of Morality,’ Nous, 24 (1990): 257–68.
³² In ‘Moore, Normativity, and Intrinsic Value,’ I defend the claim that we are only capable of a conception of ourselves as free agents by virtue of standing in second-personal relations to other free agents.
³³ See John Skorupski, Ethical Explorations (Oxford: Oxford University Press, 1999), 38–46, on ‘bridge
principles.’
How Should Ethics Relate to Philosophy?
33
for example, by grounding them in some form of contractualism.³⁴ As I have
said, I think this is a very good thing. Both metaethics and normative ethics
thrive when they are pursued interdependently, as complementary aspects
of a comprehensive philosophical ethics. In this final section, I will try to
illustrate some ways in which current matters of lively debate in normative
ethics depend upon metaethical issues regarding the normativity of their
subject matter. In these areas as in many others, I believe, solid progress can be
made only when normative ethics and metaethics are kept in close contact with
one another.
For example, critics often point out that act consequentialism is self-defeating
in various ways. The overall consequences of publicizing the act-consequentialist
theory of right, or even of each person’s privately using act consequentialism as
a deliberative standard, would be worse than either publicizing some other
non-consequentialist doctrine or having individuals deliberate according to a
non-consequentialist standard of choice, respectively. Many consequentialists
accept that this is probably so, and some, indeed, follow Moore in holding that
the standards for private or public choice should therefore be commonsense
deontological rules (such as Ross’s principles of prima facie duty). At the same
time, however, many consequentialists argue that this is in no way an argument
against their theory. Act consequentialism may be a correct theory of right even
if it would not be practical in consequentialist terms to say so too loudly,
whether to oneself or to others.
Now surely whether being self-defeating is an objection to act consequentialism depends significantly on metaethical issues concerning the nature and
concept of morality and of moral right and wrong, as I will illustrate presently.
In a somewhat different vein, critics sometimes argue that consequentialism is
too ‘demanding,’ since it holds that agents act wrongly when they fail to do all
they can to respond to unmet needs even at levels of sacrifice substantially
beyond what commonsense morality would require. This may seem to be simply a disagreement in considered judgments, but here again it is difficult to
contain the debate within normative ethical boundaries. Some consequentialists who hold that any non-maximizing act is wrong, nonetheless also say that
it would often be mistaken to blame agents for such actions, and not just on
strategic grounds but because such wrongdoing may not be blameworthy even
when the agent lacks any standard excuse. Or again, they may hold that,
although such conduct is wrong, and so something agents have the best moral
³⁴ See, for example, Samuel Scheffler, The Rejection of Consequentialism (Oxford: Clarendon Press,
1982); T. M. Scanlon, ‘Contractualism and Utilitarianism,’ in A. Sen and B. Williams, Utilitarianism and
Beyond (Cambridge: Cambridge University Press, 1982), 103–28, and What We Owe to Each Other.
34
Stephen Darwall
reason not to do, that doesn’t mean that agents shouldn’t so act, all things
considered, in light of all the reasons that bear on the case.
These consequentialist judgments have metaethical implications. If someone
dubs an action ‘morally wrong,’ but not something for which a person would
be blameworthy lacking adequate excuse, she severs a connection between
wrong and blameworthiness that others may think is part of the very concept
of moral wrong. Thus Mill famously wrote that ‘we do not call anything wrong
unless we mean to imply that a person ought to be punished in some way or
other for doing it; if not by law, by the opinion of his fellow creatures; if not by
opinion, by the reproaches of his own conscience.’³⁵ If act consequentialists
and their critics use ‘wrong’ in these different ways, they may simply be talking
past each other. But how should we conceive of moral wrong? My own view is
that Mill is right and that the usage he points to is part of a deep connection
between morality and mutual accountability. What is wrong is not simply conduct that we have strong reasons of a certain kind for avoiding but conduct that
we are responsible or accountable (to the moral community) for avoiding.
More recently, writers like Gibbard and Skorupski have echoed Mill’s thought,
arguing that the concept of moral wrong must be understood in terms of such
distinctive moral sentiments as guilt and blame.³⁶ In each case, the idea is that
judging an action wrong is an aspect of holding the agent accountable. It is seeing the agent as rightly subject, lacking adequate excuse, to something, perhaps
a sanction, perhaps some form of a self- or other-directed, second-personal
‘reactive attitude,’ like blame or guilt, that might help constitute holding him
morally responsible.³⁷
This metaethical issue has normative implications. If wrong is tied to
accountability, then some position other than act consequentialism would
seem the most sensible position for a consequentialist to take. Or again, the
metaethical issue of morality/reasons internalism may be partly at issue in what
³⁵ John Stuart Mill, Utilitarianism, ed. George Sher (Indianapolis: Hackett Publishing Co., Inc., 1979),
ch. 5, pp. 47, 48.
³⁶ Gibbard, Wise Choices, Apt Feelings, 42, and John Skorupski, Ethical Explorations, e.g. 142. This is also
an important strain in Bernard Williams’s thought. See his Ethics and the Limits of Philosophy (Cambridge,
Mass.: Harvard University Press, 1985), and ‘Internal Reasons and the Obscurity of Blame,’ in Making Sense
of Humanity (Cambridge: Cambridge University Press, 1995). See also Russ Shafer-Landau, Moral Realism:
A Defense (Oxford: Oxford University Press, 2003).
³⁷ The term comes from P. F. Strawson, ‘Freedom and Resentment,’ in Studies in the Philosophy of Thought
and Action (Oxford: Oxford University Press, 1968), 71–96. More precisely, Strawson calls these ‘participant
reactive attitudes’ (‘natural human reactions to the good or ill will or indifference of others towards us’) (80).
These attitudes are ‘second-personal’ in the sense both that they are felt in response to something secondpersonal (an attitude directed towards one) and that their natural expression is also directed toward their
object. For example, both anger and fear have intentional objects, but anger is directed towards its object in
a way that fear is not.
How Should Ethics Relate to Philosophy?
35
seem simply to be normative debates about consequentialism and ‘demandingness.’ Act consequentialists are frequently externalists on this issue, while
their critics frequently take it to be a presupposition of moral debate that morality
purports to be action- or feeling-guiding. Here again, it is impossible to isolate
the normative moral disagreement and bracket philosophical reflection on the
nature and source of morality’s authority. My own view is that standards of
moral right and wrong do purport to be authoritatively action-guiding, and
that one way to see this is to reflect on the connection just mentioned between
morality and accountability. Blame commits one to thinking there was a good
reason for the person not to have done what she did. It would simply be
incoherent to judge someone blameworthy while acknowledging there really
was no reason whatsoever for her not to have acted as she did.³⁸ It seems
incoherent, indeed, to blame while allowing that the wrong action, although
recommended against by some reasons, was nonetheless the sensible thing to
do, all things considered (supremacy). Part of what one does in blaming is to
say that the person shouldn’t have done what she did period. And accepting
blame involves an acknowledgment of this proposition also. To feel guilt is, in
part, to feel that one shouldn’t have done what one did. If, however, supremely
normative purport is internal to the very idea of moral standards of right and
wrong, then this will have implications for considered moral conviction. It will
mean that our judgments of right and wrong will be staked on their having
authority, such that, if we come to believe they lack it in some case or area, we
will have to withdraw them.³⁹
To take another example in a very different area, consider the issue of how
virtue ethics should be positioned within moral or normative ethical theory.
That would seem analogously to depend on virtue’s normativity. Sometimes
virtue is understood as a distinctively moral goodness of the person or of
character, as, for example, in Francis Hutcheson’s virtue theory.⁴⁰ If we
understand it in this way, then virtue’s normativity will be related to morality’s,
but different from that of moral obligation. Hutcheson thought that
morality, fundamentally moral goodness, was normative for distinctive
feelings of moral approbation and condemnation. More usually, however,
³⁸ John Skorupski makes a similar point in Ethical Explorations, 42–3, as does Bernard Williams in
‘Internal Reasons and the Obscurity of Blame,’ in Making Sense of Humanity, 40–4. See also Russ ShaferLandau, Moral Realism: A Defense.
³⁹ This thought underlies some of Samuel Scheffler’s arguments in Human Morality (New York: Oxford
University Press, 1992).
⁴⁰ Francis Hutcheson, An Inquiry into the Original of our Ideas of Beauty and Virtue (London, 1725). For
selections, see Moral Philosophy from Montaigne to Kant, ed. J. B. Schneewind, 2 vols. (Cambridge:
Cambridge University Press, 1990); and British Moralists: 1650–1800, ed. D. D. Raphael, 2 vols.
(Indianapolis: Hackett Publishers, Inc., 1991).
36
Stephen Darwall
virtue theories are put forward either as including good-making features of
persons that extend beyond or intersect the moral or as an alternative to
morality.⁴¹ How are we to understand the normativity of this sort of goodness?
It is impossible to know how virtue theory relates to other normative theories
without taking a position, if only implicitly, on this metaethical issue.
Obviously, the virtues concern how we ought to be, in some sense. But in
what sense? The virtues of a car are what make it a good thing of its kind, that
is, relative to the standard purposes and expectations we have for cars, either in
general or of a certain kind. Should we think about ethical virtues in the same
way? Can we, without thinking human life serves a larger purpose? We might
make use of broader functionalist notions, but even if we can, it is not obvious
what normativity this would have for us as agents, since, in Moorean fashion,
we could always sensibly ask why we should uphold the standards for a good
thing of that kind, and it is not obvious what could provide a satisfactory
answer.
Alternatively, the virtues are sometimes thought of as features of the person
or character that are necessary for well-being, qualities without which, in
Philippa Foot’s words, a person does not ‘get on well.’⁴² Virtue’s normativity
here apparently derives from that of a person’s good or well-being. But exactly
what normative force welfare has for a person is itself controversial.⁴³ Or
perhaps we should understand the virtues in relation to a kind of value that
differs from goodness of a kind, from well-being, and even from moral goodness narrowly conceived, as more general worthiness of admiration or esteem.
In the final analysis, what normative judgments of virtue we are prepared to
accept on reflection cannot be separated from these metaethical issues of
virtue’s normativity. So here again, it seems impossible to seal off the normative
theory of virtue from philosophical reflection about the very phenomena that
such a theory would hope to explain. Ultimately, a satisfactory theory of the
virtues must be placed within a philosophical ethics that integrates these
normative and metaethical issues.
The development of metaethics as a freestanding area was Moore’s legacy,
even if it wasn’t his intent. For the past hundred years, metaethics and normative ethics have frequently had an agonistic, sometimes even an antagonistic,
⁴¹ See, for example, Michael Slote, From Morality to Virtue (New York: Oxford University Press, 1992),
and Morals from Motives (New York: Oxford University Press, 2001); and Rosalind Hursthouse, On Virtue
Ethics (Oxford: Oxford University Press, 1999).
⁴² Philippa Foot, ‘Virtues and Vices,’ in Virtues and Vices (Berkeley and Los Angeles: University of
California Press, 1978), 2.
⁴³ In Welfare and Rational Care, I argue against the view that it entails agent-relative reasons for acting and
for the claim that it is normative for care—that what is for someone’s good is what one should want for that
person for her sake, that is, insofar as one cares for her.
How Should Ethics Relate to Philosophy?
37
relationship. We have seen periods in which one was in the ascendancy and the
other in eclipse, only to be followed by a reversal of roles. Perhaps this century
can see a period in which we do not lose the substantial benefits that a specialist’s sophistication in these areas can bring but also gain more of the interaction
that a genuinely philosophical ethics requires.
This page intentionally left blank
2
What Do Reasons Do?*
Jonathan Dancy
1. Reason and reasons
When I talk in this paper about reasons, I will be focusing primarily on
practical reasons, reasons for action; and I will talk mainly about what I call
contributory reasons. A contributory reason for action is a feature whose
presence makes something of a case for acting but in such a way that the overall case for doing that action can be improved or strengthened by the addition
of a second reason of the same sort. Also, a contributory reason on one side
is not necessarily destroyed by the presence of a reason on the other side.
This does happen sometimes, I agree, but it is far from the standard case.
Contributory reasons are officially reasons capable of doing what they do either
alone or in combination with others. But they can combine in peculiar and
irregular ways, as we will see. There is no guarantee that the case for doing an
action, already made to some extent by the presence of one reason, will be
improved by adding a second reason to it. Reasons are like rats, at least to
the extent that two rats that are supposedly on the same side may in fact turn
and fight among themselves, so that the addition of the second reason makes
things worse rather than better. Remember the old joke about a New York
restaurant: there are two things wrong with this restaurant—the food is terrible
and the portions are too small.¹
Much of our talk of reasons is about contributory reasons in this sense,
reasons on one side or on the other, reasons that stack up with others to make a
better or worse case for an action. But as well as talking about reasons in this
way, we also speak of what there is overall reason to do. There is nothing wrong
with that, of course, but it should not delude us into thinking that there are
such things as overall reasons in addition to the contributory ones. To talk of
what there is overall reason to do (and note that ‘reason’ in this phrase is not a
count noun) is to talk about where the contributory reasons come down—on
* The material in this paper has appeared, in substantially the same form, as parts of chs. 2, 3, and 4 in
Dancy 2004b.
¹ Thanks to Jerry Dworkin for this and other jokes.
40
Jonathan Dancy
this side or on that. We can say we have more reason to do this than to do
that, but most reason to do some third thing. These verdicts² do not themselves
specify further reasons (of an overall sort), on pain of changing the very situation on which they pass verdict. So there are no overall reasons. All reasons,
then, are contributory, unless we can find a further sort of reason that is neither
contributory nor overall.
2. The role that reasons play
How are we to understand the role played by a contributory reason? This
question is not often asked, and when it has been asked the answers given
have tended to be a bit incautious. Those who have addressed the issue have
mostly tried to explain the role of the contributory in terms of what happens
at the overall level. Jean Hampton, for instance, knows perfectly well that
there are contributory reasons. She mentions Ross’s notion of the prima
facie, though only with the now standard gripes about Rossian terminology
(1998: 51 n.). But when she talks about reasons, she constantly uses terms
that are, one would have thought, quite inappropriate for the merely
contributory. She says that reasons are ‘directive’ (51, 85), have ‘obligatory
force’ (99), have ‘compelling rightness’ (93, 99), are ‘prescriptive’ (87), are or
express commands (88), concern the ‘ought to be’ (92), ‘feel like orders’
(106), and have a ‘compelling quality’ (91–2). All these remarks seem to be
more appropriate at the overall level. An overall ‘ought’, we might say, is a
directive; in a perfectly clear sense (at least for present purposes), it tells us
what to do—what we should do. It expresses a command, maybe, it has a
compelling quality, it has ‘obligatory force’ in some sense or other, it feels like
an order and directs us to do this rather than that. The question then is
whether contributory reasons do the same as the overall does. And the answer
is surely that they don’t. If an overall ‘ought’ commands, it cannot be that
contributory reasons command as well; there is no such thing as a pro tanto³
command. We have reasons on both sides of the question, often enough. Is it
that each reason on either side commands and then that somehow the sum
² I get this term from Philip Stratton-Lake (2000: 14), who in turn owes it to Philippa Foot. StrattonLake also accepts Foot’s contrast between the verdictive and the evidential. But I prefer the term
‘contributory’ to their ‘evidential’, since I do not identify reasons for doing an action with evidence that the
action ought to be done, on the grounds that something can be evidence without being a reason.
³ ‘Pro tanto’ means ‘as far as that goes’. It has become common to prefer this term to Ross’s term ‘prima
facie’, because the latter looks as if it means ‘at first glance’—which, in Ross’s usage, it most definitely does
not. In this paper I try to avoid both of them in favor of ‘contributory reason’. But all these terms are trying
to capture the same phenomenon.
What Do Reasons Do?
41
total of them commands as well? I don’t think this is a coherent scenario;
there are too many commands floating about.
Hampton may have made a mistake in supposing that one can simply
capture what reasons do by talking in terms that really apply only at the overall
level. But this does not show that with a bit more subtlety we could not do
better, while still appealing to some relation to an overall ‘ought’ to explain the
role of the contributory. In what follows I consider a long list of suggestions of
this sort and reject them all. But they are rejected not because there is reason in
advance to say that no attempt of this sort can succeed. Each will be rejected for
specific failings rather than for being of that general style.
The first and most common suggestion about the relation between
contributory reasons and oughts is that to be a contributory reason is to be a
consideration that would decide the issue (i.e. ground an overall ‘ought’) if it
were the only relevant consideration. My own favorite version of this idea is in
Ross, who wrote:
I suggest ‘prima facie’ duty or ‘conditional duty’ as a brief way of referring to the
characteristic (quite distinct from that of being a duty proper) which an act has, in
virtue of being of a certain kind (e.g. the keeping of a promise), of being an act which
would be a duty proper if it were not at the same time of another kind which is morally
significant. (1930: 19–20)
We can think of this as a functional definition or characterization of the
role played by a reason: the characterization runs by appeal to something that
such a thing would do in a certain circumstance. And this is supposed not
just to get it right about which things are reasons and which are not, but to
capture what is going on when a consideration is not alone, or defeated, but
still making a contribution of the style that we are trying to understand. Now
there can be no objection to functional definitions of this sort, I think, but
there can be objections to instances—and there are to this one. I have four
objections, in what I think of as increasing strength. The first is that the
supposed definition makes essential appeal to the very concept it is trying to
explicate. It does this because of the presence of the word ‘significant’ at the
end, for this sort of significance is exactly what we are trying to understand,
and the account appeals to that concept in a way that certainly looks viciously
circular. It would be just the same if Ross had spoken of what is morally
‘relevant’; what we are trying to understand is what it is to be ‘relevant’ to
how to act, in the sort of way that a contributory reason is. So an answer that
makes essential use of any such notion does not seem able to advance us
very far. My understanding of functional explanations, however, is not good
enough to tell me whether this objection is really important. So I pass to the
42
Jonathan Dancy
second. This is that it is true of every feature whatever, whether relevant or
not in the actual case, that it would decide the issue if it were the only relevant
one. If so, the definition does not help us to distinguish those things that are
reasons from those that are not. It has the consequence that everything is a
reason for everything. The third problem is that the definition is trying to
characterize something that a feature can do in concert with others by appeal to
something that can only be done in isolation, and this is a peculiar procedure.
I think of it as no better than trying to characterize the contribution made by a
football player to his side’s victory by talking only about how things would have
been had he been the only player on the field. It is as if one tried to give an
account of what a conversation is which ran entirely in terms of how things
would have been if there had only been one participant. Finally, and perhaps
most forcibly, the definition assumes that each relevant feature could be the
only relevant feature. But this assumption seems not to be true. There can be,
and are, some reasons that are only reasons if there is another reason present as
well. The best example I know of this, which is contrived but still effective,
I owe to Michael Ridge. Suppose that I promise to do something only if there
is some other reason to do it. If there is no other reason to do it, my having
promised gives me no reason either. The promise only comes into play as a
reason if there is a second reason present. And this means that the definition we
are considering fails to capture reasons of this sort. But such reasons are as
much reasons as any other. So the definition is defective.
It is important to be clear what is going on here. There is something undeniable in the offing, namely that if there is a reason to , and no reason for doing
anything else, and no reason not to , one ought to . At least, I think this is
undeniable, though that does not mean that no philosopher has denied it.⁴ But
this undeniable truth should be distinguished from the falsehood that I
exposed in the previous paragraph, namely that to be a reason to is to be
such that if there is no reason not to , one ought to . Perhaps the problem
really lies with the intermediate claim that if a feature is a reason to , then if
there were no other reason in the case, one ought to . What is wrong with
this claim is that there are some reasons of which the antecedent and the
consequent of this conditional explication cannot both be true at once. The
example of the promise to act only if there is some other reason is exactly of
this sort. This being so, the intermediate claim cannot be held to be a full
characterization of what it is to be a reason.
⁴ John Broome denies it implicitly, I think, because he thinks that there can be some oughts that are not
grounded in reasons. If so, there might be a reason around, and that the only reason, without that reason
grounding an ought, because there is an opposing ought of the special Broomean sort already in play. See
Broome 2004.
What Do Reasons Do?
43
I now leave that attempt to capture the role of a reason, which we could call
the ‘isolation approach’, for another equally down to earth attempt. Perhaps a
contributory reason is a consideration in whose absence the relevant action
would be less obligatory, or even not obligatory at all. This version requires us to
make sense of the notion that obligatoriness can be a matter of degree. But I let
that pass, for reasons that will emerge later. The problem, as I see it, with this
definition is that sometimes a consideration is a reason, even though in its
absence we would have more reason, not less. To see what I mean by this,
consider a case in which I am thinking of doing something for a friend. My
action, were I to do it, would be good, and partly good because it is an expression of our friendship. But now, if I were to be doing the action and not doing it
for a friend, I would (let us suppose) be doing it for a stranger. And in such a case
the action might be even better. Now one does not have to accept this actual
example in order to accept the point it is trying to make. The point is that the
sort of support we are trying to capture is not easily capturable in conditional
terms. However we are to understand the matter, our friendship seems to be a
reason to do the action even though if we were not friends I would have even
more reason to do it. The contribution of a reason seems to be non-comparative,
in this sense. (This is all very contentious, but there is more on the point below.)
A further and less contentious objection to this definition will emerge later,
but I pass on to the idea that a contributory reason is a consideration on which
we ought to act if it is stronger than other reasons. Here I could content myself
with saying that the appeal to the notion of strength that is made here is
viciously circular (and I think the charge of vicious circularity is this time more
obviously effective than it was in the case of the isolation approach). But I have
another suggestion to make at this point. This is that as well as the sort of
peremptory reasons I have so far been discussing, which certainly do stand in
some close relation to oughts (even if we are finding it hard to characterize
them in terms of that relation), there are reasons of another style, which I call
enticing reasons, and these do not stand in the same relation to oughts at
all. Enticing reasons are to do with what would be fun, amusing, attractive,
exciting, pleasant, and so on. They can be stronger and weaker, and they are
often strong enough for action. But (as I understand the matter) they never
take us to an ought; it is not true of an enticing reason that if one has one of
them and no reason of any other sort, one ought to do what the reason entices
one to do. One can do that; but one has the right not to. With peremptory
reasons we could not say any such thing.⁵
⁵ One might allow that there is a weak sense of ‘ought’ in which one ought to choose the most enjoyable
way of spending the afternoon, where no considerations other than pleasure are at issue. But in this weak
44
Jonathan Dancy
Of course if there were enticing reasons, no attempt to define the general
notion of a reason in terms of some relation to oughts could hope to succeed.
We would have to restrict ourselves to the claim that peremptory reasons can
be so understood. And this itself would not be insignificant. So it is worth
continuing the chase, leaving the possibility of enticing reasons to one side.
My next suggestion is that a contributory reason is a consideration that
ought to motivate one. The difficulty that I see with this is that it is hard to
believe that the structure of motivation should match the profile of the reasons.
It is not as if for every reason one ought to have a little bit of motivation
(whatever that might mean), nor is it the case, so far as I can see, that one’s
overall motivation to do the action ought somehow to match the extent to
which the reasons for doing it are stronger than the reasons against, or the
reasons for doing something else. Note that the ‘ought’ in the present suggestion, as in all the others, is an overall ought. (At the moment we are supposing
that all oughts are overall oughts.) It is not that there is some reason to be
motivated by each reason for doing the action, but rather that all things considered one ought so to be motivated. And this seems to me to be false.
A somewhat similar idea is that a reason is a consideration that ought to
affect how one deliberates. Again, the ought here needs to be an overall ought,
for otherwise we are merely using the notion of a contributory reason to
explain itself. And there would be a further defect in the present case, which
is that we would be attempting to understand a reason for action by identifying it with a reason for deliberating in one way rather than another, and this
seems to subvert the focus of the reason. The reason, we might say, is trying
to get us to act in the way it wants; it would not be satisfied if we told it
that we had fully recognized it by deliberating in the way it told us to.
(Apologies for the anthropomorphism here.) But this complaint holds good
even if we understand the ought involved as an overall ought; the focus of
the reasons still seems to be subverted. What is more, there may be reasons
(even overall reason) to deliberate in a certain way that are not matched by
reasons for action. For instance, suppose that there is a feature that is very
commonly a reason but is not a reason in the present case. We should, perhaps,
bear the presence of this feature in mind when deliberating, but only so as
to determine that it is not in fact here a reason. So being a reason for action is
not the same as being a consideration that one ought to bear in mind when
deliberating.
sense of ‘ought’, it might be both that one ought overall to choose this and that one is permitted not to
choose it. And this is hardly a sense of ‘ought’ at all. I discuss enticing reasons in much greater detail in
Dancy 2004a.
What Do Reasons Do?
45
Finally, consider a remark by Jean Hampton: ‘we speak of reasons using
the language of obligation, as in “you ought not to drink that stuff because it
is petrol” ’ (1998: 80). The best way to take what Hampton is saying here,
I think, if it is to be different from anything we have seen before and rejected,
is that to be a reason is to be the sort of thing that leads to an ought. Now this is
plausible, but there is a subtle problem with it. There is a presumption here that
reasons are essentially ‘ought-makers’. But it seems to me that we are in fact
dealing with two normative relations rather than one. The first is the relation
between reasons and ought-judgments; we specify the reasons and pass to the
judgment that we ought to act. The second is a relation between reasons and
action which is not necessarily mediated by any ought at all; it is the one that
is in play when we engage in the sort of practical reasoning whose ‘conclusion’
is an action. Crucially, the relation between reason and ought-judgment is
different from the relation between reason and action. And it is really the latter
that we are after when we try to understand the notion of a reason for action—
a practical reason. Ought-judgments are judgments, and reasoning to those
judgments, however practical its purpose, is theoretical rather than practical,
since it is reasoning to something that is accepted as true. It is compatible with
this to suppose that people often reason to action by passing from reasons to
ought-judgments, and only then to action. The point is only that the relation
between the reasons and the action is different from the relation between the
reasons and the ought-judgment. And we should also avoid supposing that the
relation between the reasons and the action is the same as the relation between
the ought-judgment and the action. For, as we said earlier on, an overall ought
is not a reason. This, I think, is evidence (should any be needed) that the
attempt generally to capture what a reason does in terms of some relation to
an ought is looking in the wrong place. The relation that lies between reason
and action is not going to be captured in ways that discuss only the different
relation that lies between reason and ought-judgment.
3. Reasons as favorers
I have been arguing that there is no known way of understanding what a
contributory reason does by appeal to some relation in which the contributory
stands to the overall ought. Every suggestion we have come up with has fallen
to one objection or another. I conclude that the notion of a reason is incapable
of explication in this sort of way. Is there then nothing we can say to explicate
the role played by reasons? The only thing left to say, I think, is that a reason is
a consideration that favors action. Favoring here is a normative relation in
Jonathan Dancy
46
which a reason stands to acting in a certain way. (Reasons for belief stand in
the same relation to believing something.) But when we have been told this we
do not seem to be very much further forward. Myself, I think that the focus
on a normative relation is a positive step forward. But nonetheless we have
ended up saying not a lot. One thing we have said, of course, is that the notion
of a reason cannot be understood in ‘other’ terms. Here I see myself as saying
something rather like what Moore said about value. Moore thought that every
attempt to characterize value, the good, in other terms fell foul of the naturalistic fallacy. That has not been my argument; far from it. In fact I had no such
general argument. I tried to deal with particular suggestions one by one rather
than all together. But the end result is similar.⁶
If one cannot explicate a philosophically significant concept, there may
however be other ways of giving people a sense that the concept is itself in good
order and that they have a reasonably clear grasp on it. One way of doing this
is to work through a range of examples, showing how the concept applies to
them and showing that there is a graspable distinction between cases in which
it applies and cases in which it does not apply. In this way one can hope to
train people up in the use of a concept on which (we are supposing) they
already have an implicit grasp. My experience is that, in the case of favoring,
not too many examples are needed. Consider the following piece of practical
‘reasoning’:
(1)
(2)
(3)
(4)
(5)
I promised to do it.
My promise was not given under duress.
I am able to do it.
There is no greater reason not to do it.
So I do it.
Note here that there is a similar train of thought that ends not with (5) as
presented, but with:
(5*) So I ought to do it.
Think of (5) and (5*) as ‘conclusions’ of the reasoning presented in (1)–(4).
This is a bit odd, because I intend (5) to be (or represent) an action undertaken
in the light of the reasoning that leads to it, and it is hard to think of an action
as the conclusion of anything (except a series of other actions, I suppose). If (5)
is an action done in the light of (1)–(4), somewhere in (1)–(4) we should be
able to find something that favors that action. And indeed there is; the only
question will be how many favorers there are there. For ease of reference, I will
⁶ Thanks to Walter Sinnott-Armstrong here.
What Do Reasons Do?
47
call (1)–(4) the ‘premises’ of the reasoning, for the while at least. Later I will
suggest that they are not premises at all, whether the conclusion is an action, as
in (5), or an ought, as in (5*).
Premise (1) presents a clear favorer. That I promised to do it is (in this
context at least) a reason in favor of doing it. I am not going to argue for this; it
is an assumption of the example. What I am going to argue for is that none of
the other ‘premises’ is a favorer. They play other roles; they are relevant, but not
in the favoring way. Take (2). I want to say that the fact that my promise was
not given under duress is not a second reason for doing the action, to be set
alongside the first one. What is true here is that if my promise had been given
under duress, I would have had no reason to keep it. What this means is that
in the absence of (2), (1) would not have favored the action. In this sense, the
presence of (2) enables (1) to favor (5). In my preferred terminology, (1) is a
favorer, and (2) is an enabling condition or enabler.
Various people have suggested to me a rather different picture: that the
favorer here is not that I promised, but rather that I freely promised—a
combination of (1) and (2). I have no objection to this idea in principle; there
will be many cases where a favorer is a complex of which no part is itself a
favorer. The suggestion requires, then, that promising does not of itself favor
acting, since we must be dealing now with a complex favorer which does not
contain simpler favorers. (If promising itself favors, it is hard to think that
freely promising also favors.) I would only object to all of this if it were
premised on the idea that one can always put a favorer and an enabler together
to make a more complex favorer. This does seem to me to be false; I say a
little more about it later. The reason why I don’t need to insist that (1) is an
independent favorer is that the official purpose of my example at the moment
is only to train people up in distinguishing favoring from other forms of
relevance. Those who want to argue that the real favorer is really (1) ⫹ (2) are
thereby showing that they don’t need a lot of training.
Not but what it may be that their grasp on the notion of favoring could be
improved. For I don’t myself accept the idea that the real favorer is that I freely
promised rather than that I promised, so that it is a mistake to think that
promises only give a reason if they are freely made. I still hold that what favors
my doing the action is that I engaged myself to do so. But it is hard to know
how to tell whether this is correct or not. One clue is that those who recognize
that their promise was deceitfully extracted from them often feel some
compunction in not doing what they promised, even though they themselves
recognize that in such circumstances their promise does not play its normal
reason-giving role. I think their attitude would be different if what plays the
reason-giving role were not that one promised but that one ‘freely’ promised
48
Jonathan Dancy
(where to be free a promise must not be extracted by deceit). For on that
hypothesis there would be no sign of a favorer in the case at all. Another
example: that someone is asking you the time is a reason to tell them, a reason
that would not exist if their purpose were to distract you so that their accomplice can steal your bag. I would not be very tempted to say that the reason is
really that they are asking you the time for a genuine rather than a surreptitious
purpose.
Now consider (3). That I am able to do an action is not often a reason in
favor of doing it. There are unusual occasions, such as when I have been
paralyzed for a while and the mere ability to flex my arm is a reason for me to
do so (for a while). But this is not one of those, I am supposing. What then is
the importance of (3)? Suppose we agree that (in some suitable sense) ‘ought’
implies ‘can’. This tells us that (5*) cannot be true unless (3) is true. But we
might also think that ‘has a reason’ implies ‘can’—that one cannot have a
reason to do an action that one is (in the relevant sense) incapable of doing. I
have, perhaps, a reason to run as fast as I can but no reason to run faster than
that. If so, then in the absence of (3), again (1) would give me no reason to
act. What (3) does is to enable (1) to favor (5). It is another enabler, perhaps
not quite of the same type as (2).
Are we sure that (3) is not a favorer? I think there is a conclusive reason for
saying that it is not, which is that if it were, all favored actions would share one
and the same favorer: that they are actions of which the agent is capable. But it
is not the case that if an action is favored, we already know at least one of the
reasons for doing it, namely that the agent is capable of doing it. The agent’s
capacity to act must therefore be playing some other role, and I suggest that the
role that it is playing is that of a general enabler (unlike the role of (2), which
we could think of as that of a specific enabler, one specific to the particular case,
or at least to a limited class of cases).
Now consider (4). The presence of (4) does not enable (1) to favor (5). (1)
would have favored (5), we may suppose, even if something else had more
strongly favored not doing (5); that one promised can be some reason to act
even if there is greater reason not to. What (4) does enable is the move from (1)
to (5). In the absence of (4), that move should not be made. This is a different
sort of enabling from what is done by either (2) or (3).
Are we sure that (4) is not a favorer? Again, there is a conclusive reason for
saying that it is not. Judgments like (4) are verdictive; to assert (4) is to pass
judgment on the balance of the reasons present in the case. If (4) was itself a
further reason over and above those on which it passes judgment, we would
be forced to reconsider the balance of reasons once we had asserted (4), in a
way that would continue ad infinitum. Which is ridiculous. So (4) is not itself
What Do Reasons Do?
49
a favorer. We can add to this the sort of point I made about (3): that if (4) was
a favorer, all actions that we have overall reason to do would have at least one
common favorer, which I take to be intrinsically implausible. In saying this
I don’t mean to be appealing to any first-order theory, idiosyncratic or otherwise.
What is implausible is that the structure of reasons alone should guarantee that
all right actions share a common favorer, which it would if (4) were taken to
be favoring rather then enabling.
So what we have in the present example is one favorer and three enablers,
none of which is doing exactly what any of the others is doing. For a simpler
example of a similar sort, consider Plato’s suggestion in his Crito that agreements should be respected so long as they are just. The ordinary reason for
respecting one’s agreements is, I suppose, that one agreed to them. Plato’s idea
seems to be that this would be no reason if the agreement itself were unjust. But
that the agreement is not unjust does not seem to be another reason for
respecting it; that something is not unjust is not ordinarily a reason for doing
it, and it does not seem to be one in this case either.
In addition to favoring and enabling as two different sorts of things that
relevant considerations can do, there is a third role that a relevant consideration can play. To see what this might be, consider a rather different
example:
1. She is in trouble and needs help.
2. I am the only other person around.
3. So I help her.
That she is in trouble and needs help is a consideration that favors my helping
her. That I am the only other person around does not seem to be another
reason, on top of the first one. It is not as if, even if she were not in trouble, that
I am the only other person around would still favor my helping her. The
reverse, however, is true; even if there were others around, I would still have a
reason to help her, a reason given by the trouble she is in. But my being the
only other person around does make a rational difference, all the same. I suggest that what it does is to intensify the reason given me by her need for help.
Instead of two reasons, what we have here is one reason and an intensifier.
Presumably there is the opposite of an intensifier, an attenuator. An example
might be this:
1. She is in trouble and needs help.
2. It is all her own fault, and she got in this situation through trying to spite
someone else.
3. But still I help her.
Jonathan Dancy
50
One might think that there is reason to help her even though it is all her own
fault but less reason than there would otherwise be.
So far, then, we have identified three sorts of role that a relevant consideration can play: a relevant consideration can be a favorer/disfavorer, it can be an
enabler/disabler for another favorer/disfavorer, and it can intensify/attentuate
the favoring/disfavoring done by something else. All such considerations are
relevant, but there is more than one way of being relevant.
These distinctions are important. Here I mention just two ways in which
they matter. Consider again the third attempt to characterize reasons in terms
of oughts. This was that a reason for action is a consideration in whose absence
the relevant action would be less obligatory, or even not obligatory at all. We
can now see a much stronger argument against this suggestion. This is that
enablers are not reasons, but they are considerations in whose absence the
relevant action would not be rendered even partly obligatory. Further, intensifiers are not reasons, but they are considerations in whose absence the action
would be less obligatory than otherwise. And here we have an instance of the
second, and more general, way in which our distinctions matter. They show
that in the attempt to understand what reasons do, subjunctive conditionals
are blunt tools. Every time we see an attempted explication that centers on a
subjunctive conditional, we should be wary.
4. Some strategic considerations
It is worth confessing that the picture I am offering deprives us of two considerable advantages. If we had been able to understand contributory reasons in
terms of some relation to the overall ought, we would have been handed on a
plate answers to two important questions:
1. What makes the notion of a reason normative?
2. What makes the notion of a reason deontic?
We take it that the notion of an ought is both normative and deontic. It is
deontic by definition, more or less. The normative realm is divided into two
halves: the evaluative and the deontic. The former is concerned with what is
good and bad, the latter with what is right or wrong, with what one ought
to do, with duty and obligation and so on. The idea is that one can sense a sort
of family resemblance between the different deontic concepts, even if one
cannot do much to explicate that resemblance. Now which side of the divide
would the notion of a reason fall on? Traditionally, it is shown to be both
deontic and normative because of its definability in terms of an overall ought.
What Do Reasons Do?
51
However, we have now decided that we cannot define a reason in that way, and
I think we have also seen the possibility of non-deontic reasons. For enticing
reasons, if indeed there are any such things, do not seem to be concerned with
what we ought to do; they are more concerned with what it would be pleasant
to do, without any suggestion that somehow one ought to take the most pleasant course (at least, not in any sense that is incompatible with one’s being
permitted not to). If so, these reasons don’t take us to oughts, and it looks as
if we are going to have to think of them as lying on the evaluative side of
the evaluative/deontic distinction. They are more to do with what is best than
with what one ought to do. This leaves us trying to say in what sense they
are normative. But that issue also now arises for the non-enticing reasons, the
peremptory ones, once we have accepted that they cannot be shown to be
normative by simple appeal to the sort of definition we have rejected. If we
want still to think of the notion of a reason as normative (or even as the basic
normative notion, as I will be urging later), we will have to think of it not as
extrinsically normative (i.e. as normative in virtue of some relation it bears to
an intrinsically normative ought) but as intrinsically normative. We will, that
is, have to appeal directly to the notion of favoring and maintain that it is itself
normative, normative in its own nature. Perhaps this can be done, but there is
a struggle ahead. We may find ourselves saddled with two styles of favoring: the
more deontic style that is appropriate for the peremptory and the more
evaluative style that is suited to enticers.
5. Two basic notions?
We have decided that we can neither understand the contributory in terms of
some one relation to the overall nor do without the contributory altogether.
This leaves us with a further strategic choice. One option is to take it that there
are two basic notions, that of a contributory reason and that of an overall
ought, neither of these being definable in terms of the other. Our other main
option is to reverse our original direction and to try to define the overall
in terms of some relation to the contributory. Of course to do this we will have
to show how to define all overall notions in contributory terms, not just that
of the overall ought; we will have to pull the same trick for the notions of
goodness and rightness, which may not be so easy. Nonetheless the general
program of reducing overall concepts to contributory ones is extremely attractive, for reasons that I now turn to investigate.
The matter can be brought out by considering the difficulties that Ross
gets into when he tries to relate what he calls prima facie duty to what he calls
52
Jonathan Dancy
duty proper. First, he says that ‘the phrase “prima facie duty” must be apologized for, since it suggests that what we are speaking of is a certain kind of duty,
whereas it is in fact not a duty, but something related in a special way to duty.
Strictly speaking, we want not a phrase in which duty is qualified by an adjective, but a separate noun’ (1930: 20). Second, he says that prima facie duty ‘may
be called a parti-resultant attribute, i.e. one which belongs to an act in virtue of
some one component in its nature. Being one’s duty is a toti-resultant attribute,
one which belongs to an act in virtue of its whole nature and of nothing less
than this’ (1930: 28). Each of these claims causes problems.
The first one is awkward because by denying that prima facie duty is duty in
any sense, Ross makes it impossible for himself to make sense of the idea that
duty proper is in some sense the product of various normative pushes or forces
coming up from below—from the various features that stand as reasons. Ross
effectively maintains that there is nothing like duty at the contributory level,
and so leaves nothing there that we can think of as contributing to duty. But
this is to sever what he has to say about duty from its proper ground. To put the
matter another way: we have to find some way of understanding right-making
features so that the ‘right’ in ‘right-making’ is the rightness of duty proper, and
Ross has prevented himself from doing that.
The second one is awkward because it reinforces the divide between duty
proper and the things that contribute to it. If the ground for duty proper is
really every feature of the relevant action, and the ground for prima facie duty
is some particular feature or other, we have deprived the relevant right-making
features (which occur at the contributory level) of the ability to be what make
the action right. It is impossible to conceive how they can be what make the
action right because we have been explicitly told that they play no special role
in the construction of duty proper—no role, that is, that is not played by every
feature of the action whatever.
It may be that there is some other resolution of these difficulties, but by far
the easiest is to work with a contributory ought, which is a matter of degree. I
think of this contributory ought as a monadic feature of an action which is
consequent on, or resultant from, some other feature—the ‘ought-making’
feature, whatever it is. So oughts of this sort are not relations. But for them to
be present there must be a certain relation between the ought-making feature
and the action. One might suppose that this relation must be the favoring
relation. Things are a little delicate here. I am still inclined to say that we are
dealing with more than one normative relation. The monadic ought is reached
by detaching from the ought-making relation. The train of thought here is
‘Feature F ought-makes action A; Feature F is in place; so one ought (so far as
that goes) to do action A’. This expresses the idea that the relevant relation is
What Do Reasons Do?
53
not favoring, but ought-making (or right-making, if you like—so long as
‘right’ does not mean only morally right). So we can stick to our intuition that
favoring is a relation between a feature and an action, and ought-making is a
making-the-case relation that holds between a feature and the (contributory)
rightness/oughtness of an action.
6. Fitting and most fitting
Suppose, then, that we pursue this strategy. We are going to understand the
overall in terms of the contributory. For this, we need a sound conception of
what happens at the contributory level and a way of understanding the overall
as some function of that. I want to suggest briefly that classical intuitionism is
well placed to offer what is needed here, and I will end by trying to show that
some other more recent positions are in much worse shape. The intuitionists
tended to understand reasons and oughts in terms of what is fitting. The action
which overall one ought to do, which one has most reason to do, is that action
which best fits (or is most fitting to) the situation in which one finds oneself.
An action that one has at least some reason to do is an action that fits the
situation in some respects but not necessarily in others. It is this notion of being
fitting in a certain respect, of partial fittingness, that is the crucial one here. If
this notion makes as much sense as it seems to, intuitionism has the right
strategic shape to capture the notion of a reason and to explain the overall
ought in its terms. Can others claim as much?
7. Humean realism
Michael Smith (1994: 151 ff.) understands rightness as the property an action
has of being such that one’s most rational self would advise one to do it. Now
if we converted this directly to yield an account of reasons, we would get
something like this: for there to be a reason for me to is for my most rational
self to advise me to do it, so far as that goes. But this is clearly wrong, for reasons
that should now be familiar. The notion of advice is an overall notion. There is
no such thing as partial or contributory advice. One cannot say ‘To some
extent (or in some respect) I advise you to do this, but overall I advise you not
to do it.’ So if Smith were restricted to the terms in which his official account
is presented, he would be stuck at the outset. But there is a way of understanding advice in terms of desirability; it is not as if the actual act of giving and
receiving advice is especially important here. We could try to re-express Smith’s
54
Jonathan Dancy
position thus: for an action to have the property of rightness is for it to be
such that one’s most rational self would judge it desirable that one do it. And
now things are looking more promising, for there does seem to be some
prospect of finding a suitable notion of what is desirable in a certain respect,
but not overall.
If there is such a notion, we can use it in the following analysis of a reason: for
there to be a reason for me to is for my most rational self to find it desirable in
a certain respect that I . So far, then, things are going well for Smith. But what
exactly is this notion of desiring or finding desirable in a respect? I know two
ways of understanding this notion, and I do not think that either will work. The
first is to construe desiring in a respect in terms of valuing that respect. The idea
here is that I value honesty, integrity, kindness, and thoughtfulness. The action
I propose to do will be dishonest and, to that extent, lacking in integrity, but it
will be kind and thoughtful. I have reasons to do it and reasons not to do it. My
most rational self, that is, desires honesty and integrity and kindness and
thoughtfulness, and those independent desires for features are construed as
partial desires for any action in which the features desired are present.
This conversion of the valuing of features one by one into a partial valuing—
and hence desiring—of complexes in which they occur will not work, because
of the organicity (or holism) of value. It is generally agreed that the value of a
complex whole is not to be identified with the sum of the independent values
of its parts. If that is so, we cannot move automatically from an independent
evaluation of the parts to any judgment of how valuable/desirable they are
in the whole. But our first way of making sense, on Smith’s behalf, of a notion
of desiring in one respect and not in another involved exactly that move. It
involved the identification of desiring a part wholly and desiring a whole partly,
and this is flatly at odds with the organicity of value (once we understand value
as desirability).
There is however another way of working. To desire a whole partly might be
to desire the whole more with the part than without it. There is a subjunctive
conditional operating here: I desire O partially for being F iff I desire it as it is
more than I would if it were not F. This subjunctive conditional should remind
us of one of the failed definitions of what a contributory reason does. And I
suggest that the problems we had with that definition are repeated here. The
crucial point is that the following things are compatible:
1. Part of the value of the whole derives from the presence of feature F.
2. If it were not F, the whole would be even more valuable than it currently is.
The way to show these two compatible is to find a case where, if the whole had
not been F, it would have been G—and being G is even more valuable than
What Do Reasons Do?
55
being F. In such a case, we can say that the whole is the better for the presence
of F, but we had better not mean by this that it would be worse without F. Now
let us see if such a case can be found. The simplest case of that sort is one where
being not-F is even more valuable than being F. This sort of win-win situation
does not seem unimaginable to me, and I have already suggested some
examples. I might praise you for doing something nice for a friend and praise
you even more fervently were you to do it for someone who is not a friend. But
perhaps this suggestion appears fanciful. The first approach I mooted involves
the thought that if the whole had not been F, it would (as things stand) have
been G. Let us suppose that had Mary been at the party, her presence would
have been one of the respects in which the party was enjoyable. But as things
stood, her absence enabled a more intimate gathering that was even better than
the party would have been had she been there. Now is this a relevant example?
If it is, it is easily repeatable and will defeat any attempt to understand partial
desirability in terms of an overall ranking of alternatives.
There seems to be only one effective objection to such examples, which is
that they involve appeal to the wrong subjunctive conditional. Instead of
appealing to how things would have been if they had not been F, it may be said,
we should start from how they are, with feature F present, and create a contrast
with that very same situation minus F. So in the case of the party with Mary, we
should extract Mary from that party and then ask about the value of the
resulting object, the actual party minus Mary. The greater value of the other
party, the one we actually had, is irrelevant.
I do not think that this maneuver can possibly be got to work across the
board; and it has to work across the board if it is to be the basis of a sound
general understanding of ‘desirability in a certain respect’. The idea had been
that we are to rank alternatives, and alternatives have to be ways in which things
might have gone. Now there are certainly two ways things might have gone, to
pursue the party and Mary a little further. There was the larger party with
Mary, and the smaller, more intimate one without her. What is not a way things
might have gone is the ebullient party without Mary. Look at another example.
Suppose that my wife and I hire a car and drive across the country to the West
Coast, spend a day there, and then drive back. (I wrote this on the East Coast
of New Zealand.) Now I want to assess the value of having a car to make the trip
in. Do I compare the value of the trip with a car and without a car, leaving it
somehow unspecified how we are to get from East to West Coast and back
again? No, I compare doing it by car with doing it by bus, let us say—or by
bicycle, perhaps. Do I compare doing it by car with the weighted average of
doing it in the various alternative ways, or with doing it in no particular way,
but in some way or other (shades of Berkeley on abstraction here)? No, again.
Jonathan Dancy
56
There really only seems to be one coherent comparison. And the same applies
to Mary and the party; we need whole alternatives to compare for value, and
the Mary party without Mary is no such thing. Or rather, if it is anything at all,
it is the intimate party, not a less well-attended ebullient one.
Perhaps we could cope with this by supposing that context, rather than
logic, should determine the appropriate subjunctive conditional. This is of
course a wise sort of thing to say in general; the problem is how to apply it to
the Mary case so as to get the right answer. The right answer is that Mary would
have contributed to the value of the ebullient party but that ‘it’ would have
been even better without her. So what might the ‘it’ be here? To fill out the
context a little: the three of us who made up the intimate party would, with
Mary, have had an ebullient rather than an intimate evening. We want to
find something in between the incomplete ebullient party—that party but
Maryless, as it were—which seems to be incapable of existing, and the intimate
party, which could exist but gives us the wrong answer (that Mary did not
contribute to the ebullient party). And I don’t think that there is anything
intermediate to be found here. So the tactic of allowing the subjunctive
conditional to vary according to context does not seem to be even applicable to
the present case.
All this seems to me to show that as things stand Smith is devoid of any
suitable notion of ‘desirability in a respect’. And if this is true, it is at best moot
whether his general approach is capable of capturing the notion of a contributory reason. If it is not, the probable diagnosis is that it was constructed
with only the overall in mind, in a way that turns out not to be as flexible as it
needs to be if it is to cover all the normative ground.
8. Expressivism
For an expressivist attempt to capture what it is to be a reason or, perhaps better,
what it is to take something to be a reason, we need to look at Allan Gibbard’s
Wise Choices, Apt Feelings (1990: 160–6). To see the problem from Gibbard’s
point of view, we need to ask what we might expect an expressivist to say about
the notion of a reason. An initial suggestion might go something like this:
To say that R is a reason for S to in circumstances C is to express acceptance
of a system of norms that directs S to in C.
The trouble with an account of this sort is that it does not address the notion
of a contributory reason. We want to know what it is to say that R is some
reason for S to in C, and what we are told is surely located more at the notion
of overall reason. After all, I might say that R is a reason for S to in C while
What Do Reasons Do?
57
actually adhering to a system of norms which forbids S to in C. What is
required, then, is to find some other term than ‘directs’. What then does
Gibbard in fact offer? He writes: ‘To say that R is some reason for S to in C ’ is
to express acceptance of a system of norms that direct us to award some weight
to R in deciding whether to in C ’ (1990: 163). Note that our initial suggestion supposed that the norms I sign up to in saying that there is a reason for S
to in C are norms that require (or recommend) that S should be doing the
-ing in C (or the deciding whether to in C ). On Gibbard’s account, by
contrast, they require me to be doing something, not S. But that is by the by.
The real weakness is that the account makes explicit use of some appropriate
conception of weight—of normative weight, as one might put it—and that
was pretty well exactly the thing we were trying to understand. Further, we
should note the way in which the account is entirely in terms of ‘awarding
weight to R in deciding whether to in C ’. But to award weight to R in one’s
decision is not the same as taking R to favor -ing in C. I might think, for
instance, that I should always consider the question whether R if I am deciding
whether to in C, but suppose nonetheless that on many occasions the
question whether R or not will be in fact irrelevant to how I ought to act. This
would happen if R was very commonly relevant, so that it is a consideration one
ought always to bear in mind, even though sometimes its relevance is defused
by other considerations. If, by contrast, awarding weight to R is another way of
saying ‘take R to count in favor or against’ we are left grasping for an expressivist
account of these notions, which were surely the ones the analysis was supposed
to be grappling with in the first place.
We might therefore try the following slightly different version:
To say that R is a reason for S to in C is to express acceptance of a system
of norms that would lead one to approve of S ’s -ing, at least to some
extent, in respect of R.⁷
The problem now is that, though one thinks there is indeed some reason for S
to in C, it might well be that one also thinks that there is conclusive or overall
reason for S not to in C at all. The difficulty is to make sense of the idea that
one might approve of S ’s -ing to some extent while being entirely against it.
But isn’t there such a phenomenon as approving of an action in one respect
and not in another? Let us say that the action is that of my son’s buying a car
⁷ Note that this suggestion could be turned into a non-Gibbardian one by dropping the explicit reference
to the system of norms. This would give us:
To say that R is a reason for S to in C is to express approval of S ’s -ing, at least to some extent.
This version is more amenable to the general thrust of expressivism; it is something that Simon Blackburn,
for instance, might espouse. But the criticism offered in the text applies to this simpler suggestion as well
(or badly) as it does to the Gibbardian one in the text.
58
Jonathan Dancy
(a 1971 Citroën DS). The elegance of the car is some reason to buy it, I think,
and its unreliability is some reason not to buy it; the unreliability wins, and
there is overall reason not to buy it. One way to understand what is going on
here is to say that I approve overall of the elegance (my approval of the elegance
is unmixed, as it were) and disapprove overall of the unreliability. These will
be two overall approvings, and we are going to try to construct a notion of
contributory or partial approving of buying the car out of them. This should
remind us of the previous discussion of Smith. The difficulty for Smith in this
connection arose because of the organicity of value, and I suggest that the same
problem arises now. Wholesale approvals of parts considered apart cannot be
converted directly into focused approval of those same parts when they are
parts of wholes, or of wholes in virtue of the presence of those parts, because a
feature need not have the same value in every whole in which it appears.⁸
One can say ‘I partly like him and partly dislike him; I like his wit and charm
but I dislike his manipulativeness.’ This is overall liking and overall disliking,
directed at a part. Whether overall one likes him or not will be determined by
the balance between these partial likings (likings of parts) in an organic way.
What the expressivist needs here is a notion of overall liking whose object is
the whole but which is somehow also focused on the part. I like him for his
wit and charm; I dislike him for his manipulations. But so far we have not
seen how to do this.
The second approach we might try on behalf of the expressivist tries to
piggy-back on an expressivist conception of an overall ‘ought’-judgment. We
are, for present purposes, allowing that there is nothing wrong with the
standard expressivist approach on this topic:
To say that S ought overall to in C is to express acceptance of a system of
norms that would require S to in C.
So we might try something like the following, for the contributory:
To say that R is some reason for S to in C is to say that, if it had not been
the case that R, S ’s -ing in C would have been less right than it is.
But this can hardly succeed, for reasons most of which we have seen too often
before. First, our expressivist has not yet established a good sense for the notion
of more and less right. We just have the notion of requiring that appears in the
account of overall oughts, and we have not been shown any expressivist way
of taking requiring to come in degrees of stringency. (Note that there is a
difference between degrees of acceptance of a norm and degrees of stringency
⁸ This is a tendentious version of organicity. I defend it in Dancy 2003.
What Do Reasons Do?
59
of the norm accepted.) Second, the subjunctive conditional we are now considering is anyway incorrect as an analysis of the contributory. For this point,
we can again appeal to the discussion of Smith; this sort of appeal to subjunctive conditionals does not work, because it can be the case both that this feature
counts in favor of the action and that the action would have been better (more
right) without it.⁹
Some will feel that this is all mere skirmishing. Suppose, then, that we try to
capitalize on the approach to pro tanto or ceteris paribus preference considered
earlier. What we are looking for, again, is a notion of approval of an act in virtue
of the presence of some feature F, to take this as a model for taking the presence
of F as a reason for doing the action. We could try to do this in terms of approving of the act with that feature more than one approves of ‘the same’ act without
it. But we have already seen all sorts of difficulties with this approach. And what
if the presence of a certain reason against doing the action was required for the
presence or force of other stronger reasons in favor of it? For a formal example
of this, what if I promise that even if there is some significant reason against
doing this, I will still do it?
My conclusion is that the expressivists have failed to offer any effective
account of a contributory reason and that they do not seem to have the
resources to do better.
References
Broome, J. (2004). ‘Reasons.’ In Wallace et al. (2004).
Dancy, J. (2003). ‘Are There Organic Unities?’ Ethics, 113: 629–50.
—— (2004a). ‘Enticing Reasons.’ In Wallace et al. (2004).
—— (2004b). Ethics without Principles. Oxford: Clarendon Press.
Gibbard, A. (1990). Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard University
Press.
Hampton, J. E. (1998). The Authority of Reason. Cambridge: Cambridge University Press.
Ross, W. D. (1930). The Right and the Good. Oxford: Clarendon Press.
Smith, M. (1994). The Moral Problem. Oxford: Blackwell.
Stratton-Lake, P. J. (2000). Kant, Duty and Moral Worth. London: Routledge.
Wallace, R. J., Pettit, P., Scheffler, S., and Smith, M., eds. (2004). Reason and
Value: Themes from the Moral Philosophy of Joseph Raz. Oxford: Clarendon Press.
⁹ Further, this analysis says of a reason something that is just as much true of an enabling condition that
is not a reason.
This page intentionally left blank
3
Evaluations of Rationality
Sigrún Svavarsdóttir
I
One hundred years ago, G. E. Moore declared, ‘how “good” is to be defined, is
the most fundamental question in all Ethics’¹ and, then, proceeded to argue
that ‘good’ is indefinable. The rest is history: the history of twentieth-century
metaethics. That history is often recounted as a series of responses to Moore’s
open question argument for the indefinability conclusion and the further
conclusion that goodness is a non-natural property accessible through rational
intuition. New developments in the philosophy of language, metaphysics,
epistemology, and the philosophy of mind have opened up new ways of diagnosing the fault in Moore’s argument and new ways of understanding the
semantics, metaphysics, and epistemology of ethical discourse. I will not
recount the details of that history here.
Philosophical reflections on the nature and grounds of ethical judgment did
not start with Moore. However, he set them on a new footing by turning attention to the analysis of ethical concepts. This marked the ‘analytic turn’ within
moral philosophy and the advent of metaethics as a distinct philosophical
enterprise. Nevertheless, the history of twentieth-century metaethics cannot
be fully told as a series of responses to Moore. That leaves out an important part
of the ongoing discussion about the nature and grounds of moral judgment: a
This is a revision of ‘Evaluations of Rationality’ as printed in the Southern Journal of Philosophy, Spindel
Supplement, 41 (2003): 121–36. Aside from a new introduction, the changes are very minor. The original
version was written while I was a Fellow in residence at the National Humanities Center. I gratefully
acknowledge the Center’s support. Lectures based on this article were given at the Spindel Conference 2002,
at a conference titled ‘Mikjálsmessa’ held at the University of Iceland in March 2003, and at North Carolina
State University, March 2003. A version of the article was also presented to the Triangle Ethics Group, North
Carolina. I thank everyone who participated in the discussion at these four forums. Special thanks are due to
Russ Shafer-Landau for fair and challenging comments presented at the Spindel Conference and published
in the conference proceedings. Finally, I would like to thank Justin D’Arms for comments on the paper and
Jenefer Robinson for a helpful discussion.
¹ G. E. Moore, Principia Ethica (Buffalo, NY: Prometheus Books, 1988), 5.
62
Sigrún Svavarsdóttir
part that pertains to the nature of reasons for action, practical rationality, and
the relation between morality and rationality. Many have thought that we can
only understand the normativity of moral judgments, if we attend to such
matters. This paper is a small contribution to this part of metaethics. I will not
discuss the thorny notion of a reason for action nor the relation between morality and rationality. Instead, I focus attention on the evaluation of agents as
more or less rational. How is it best to construe this type of evaluation?
I will not come to a final conclusion about this matter, although I offer a first
approximation of a theory of rationality. This is a neo-Humean theory, since it
relativizes evaluations of rationality to the ends or attitudes of the agent evaluated. Neo-Humeans about rationality often do little to defend their view. They
take themselves to be offering a minimalist theory—minimalist in the sense
that whatever they have to say about the nature of rationality has to be included
in their opponents’ theory of rationality. The neo-Humeans, then, challenge
the anti-Humeans to defend the part of their theory that goes beyond its neoHumean component, which is considered a common ground and, hence,
beyond controversy. This will not be my strategy. Below, I motivate the idea
that all evaluations of rationality are relativized to the agent’s ends or attitudes
as neo-Humeans maintain. Indeed, I argue that this is true for evaluations of
both theoretical and practical rationality. This distinction should, I suggest, be
drawn in terms of the types of goals that need to be taken into account in
evaluations of rationality. I end the paper by discussing what shape a defense of
my and other neo-Humean accounts of rationality has to take. In this context,
I argue that the sort of relativity proposed here has no obvious metaphysical
implications and should not be confused with the type of mind-dependence
that has figured into metaphysical discussions of objectivity. I also broach
issues concerning the connection between rationality and the pursuit of (or
respect for) values as well as issues concerning the justificatory or critical force
of verdicts of rationality.
II
It seems uncontroversial that only RATIONAL as opposed to arational beings are
subject to evaluation in terms of rationality or irrationality. (In order to avoid
confusion, I will use small capitals when using ‘rational’ to contrast with ‘arational’.) Now, a natural thought is that rationality is the excellence of a being qua
RATIONAL being. This will be my guiding idea. Notice that this approach makes
rationality out to be a virtue. Of course, no one might instantiate this virtue
perfectly. People are rational to a greater or lesser extent. At some point they are
Evaluations of Rationality
63
not merely lacking in rationality but are downright irrational. Moreover, there
is both the issue of how rational an individual is overall and also the issue of the
extent to which he displays rationality in having a specific belief or intention.²
It is the latter type of evaluation that mostly interests me here.
Perhaps we can make some headway in thinking about rationality as a form
of evaluation by getting clear on what it is to be a rational being. Minimally,
it consists in having well-developed cognitive capacities—capacities for mental
representation and reasoning. I will not attempt here to specify fully what these
capacities must be like in order for their possessors to count as rational
beings. Suffice it to say that the representational capacity must go beyond a
mere disposition to register features of the environment in the way that a
shadow of a tree may indicate the time of day. The representations must be able
to feed into thought processes that amount to sound reasoning—sequences of
mental states whose representational contents stand in relations such as implies, is
consistent with, explains, is a generalization of, is an instance of, etc. There must also
be room for failure: the representation may be inaccurate and the reasoning faulty.
These normative properties of representations and of reasoning invite the
idea that the excellence of a being qua RATIONAL being consists in representing
and reasoning well or correctly. But this does not give us a plausible conception
of rationality, not even theoretical rationality. It is possible to represent something incorrectly without being guilty of irrationality. This might happen when
the subject is so unfortunately situated in the world that he receives a misleading input into his cognitive system and is not in a position to discover the error
however well he uses his cognitive capacities. A representation may also be poor
though correct. This does not detract from the rationality of the cognizer if it is
due to the poverty of the input available or the limits of his cognitive powers.
Indeed, it is rational to leave out details, when they are irrelevant or distracting.
Perhaps less obviously, it is possible to be rational even if one’s reasoning does
not quite measure up to the standards of inductive and deductive reasoning. As
Gilbert Harman has noted, if it takes a genius to recognize that a triad of propositions imply a fourth proposition, the average Joe is not irrational in failing to
draw the inference.³
² We also evaluate actions, intentions, beliefs, and possibly some other intentional states as rational or
irrational. These evaluations, I submit, are best understood as derivative evaluations. A rational action or attitude is an action or an attitude that an agent would undertake or have if he were rational.
³ Gilbert Harman, Reasoning, Meaning and Mind (Oxford: Oxford University Press, 1999), 18. It seems
odd, however, to say that he is rational in failing to draw the inference. The verdict that someone is not irrational in doing something is not the same as the verdict that he is rational in doing it. The former verdict
absolves the person from a certain type of criticism rather than commends him. The latter verdict commends
him. The difference between these verdicts is illuminated by how the overall assessment of the agent’s rationality is affected by his failing to draw the inference in question. Failing to draw the inference does not detract
from the agent’s overall rationality, although it does not bolster his rationality either.
64
Sigrún Svavarsdóttir
These observations suggest that making an error and being irrational are not
quite the same thing, even in the theoretical arena. Rationality is less about
being free from error than about not making mistakes that could be avoided by
exercising one’s cognitive capacities well; thus, evaluations of rationality must
be sensitive to the limits of the cognitive powers of the subject evaluated, as well
as to any other feature of the subject, or his circumstances, that may affect his
ability to exercise these cognitive powers to the fullest. (From now on, I will refer
to these as his epistemic limits.) This leads to the suggestion that rationality—
the excellence of a being qua RATIONAL being—consists in using one’s cognitive
capacities as well as possible given one’s epistemic limitations.⁴
III
Something still seems amiss. It is irrational to clutter one’s mind and waste one’s
time by teasing out all the trivial and uninteresting implications of one’s beliefs
or to weed out any inconsistency between them.⁵ It seems also irrational to go
off endlessly on tangents and explore at length insignificant side issues rather
than continue a main line of enquiry, no matter how exquisitely one handles
the tidbits. In these cases, one need not be making avoidable errors in representation or reasoning. One might, indeed, be using one’s cognitive capacities
superbly, at least in the sense that the reasoning is sound and the representation
is both subtle and correct. So why do I fall short qua RATIONAL being if I spend
my days teasing out trivial implications of my beliefs? A first stab at an answer
focuses on my limitations: I have a finite mind and a finite amount of time.
Spending them in this way diminishes my opportunity to grapple with matters
more important than trivial implications of my beliefs. This raises the question
of whether the excellence of a being qua RATIONAL being is in part constituted
by not letting an enquiry into less important matters detract from one’s enquiry
into more important ones. If so, we could even save the suggestion that
rationality—the excellence of a being qua RATIONAL being—consists in using
one’s cognitive capacities as well as possible given one’s epistemic limitations.
We would only have to add that it is possible to use one’s cognitive capacities
poorly, not only by falling into avoidable error, but also by using them on one
subject matter at the cost of another, more important, subject matter.
I do not dispute that some subject matters are more important than others
and that we may be using our cognitive capacities poorly if we concentrate on
the less important at the expense of the more important subject matter. But it
⁴ Of course, the notion of an agent’s epistemic limits needs to be clarified. I leave that for the future. It
also needs to be rendered more precisely how the agent’s epistemic limits affect evaluations of his rationality.
⁵ This observation is borrowed from Gilbert Harman. See his Reasoning, Meaning and Mind, 18–19.
Evaluations of Rationality
65
does not seem as though this in itself amounts to irrationality. It seems more
important to find the cure for cancer than design household equipment.
Moreover, it seems a shame if a genius decides to use his formidable mental
powers on designing household equipment rather than on finding the cure for
cancer. Perhaps, it is also a moral failing. Certainly, it is not the best use to
which the genius can put his cognitive capacities. But this hardly suffices for
making the genius irrational. Imagine encountering the genius in his lab
excitedly explaining to you the newest intricate design of the perfect bottle
opener. You think of what he could have accomplished as a medical researcher.
Alas, he was never able to develop interest in anything but the complex design
of silly household equipment. Do you accuse him of irrationality for designing
bottle openers rather than developing new drugs? That does not seem like the
appropriate kind of criticism in this case.
Now imagine that the genius starts obsessively to work out all the implications
of his beliefs. He gets to the point of tediously listing all the disjunctions that
follow from his belief that snow is white, an endless task indeed. There are halffinished designs waiting on his drawing table, finished designs that need to be
forwarded to the production department, unpaid bills, missed appointments.
He neglects his friends, loses sleep, and forgets to eat. Calling him irrational is
at best an understatement. This is a descent into madness. But consider a
slightly milder case of this disorder: the genius has a tendency to break off his
work on his intricate designs by starting to work out the implications of his
various beliefs, even if they do not bear on what he is doing and even if they are
of no interest to him beyond being implications of what he already believes.
These obsessive reflections seriously hold up his projects from time to time,
although they never take over completely. Alternatively, imagine that the
genius has a tendency to go off on tangents. Frequently, he explores thoroughly
possible addenda to the device he is designing, though it should be clear to him
from the start that they have no good use or do not fit with the overall design
of the object. These become little side projects that are abandoned as fruitless after
holding up the main project for hours on end. These are, I submit, clear cases of
irrationality. Being a genius, the designer’s reasoning and representations are
beyond reproach even when he goes off on tangents or obsessively works out all the
implications of his beliefs. So, what is his shortcoming qua RATIONAL being?
IV
The key to understanding this form of irrationality is to appreciate that it is a
shortcoming of a being that uses its cognitive capacities purposefully, that is,
of an agent. Imagine a machine that at least closely simulates our powers of
66
Sigrún Svavarsdóttir
representation and reasoning. Like us, it can fall into error, and we may even
imagine that it can in some cases detect and correct its mistakes much like we
do. Say that this machine replaces fieldworkers who gather and analyze data
about various features of the physical environment in remote regions of the
world. Now, compare this machine to a fieldworker whom the machine has
replaced. The representation and reasoning they use may be very similar. We
may even assume that the machine and the fieldworker can be misled in a
similar way by an input from the environment, but they both have the ability
to catch the mistake and correct it. When the machine fails to correct the
mistake, it is malfunctioning, whereas the fieldworker is being sloppy, dense, or
irrational. Also, imagine that some input triggers the machine to spend an
inordinate amount of time recording data of little interest given the main line
of research. It goes off on a tangent so to speak. If a similar thing happens to the
fieldworker, we deem her irrational. But such an evaluation does not apply to
the machine. It is neither rational, when it functions well, nor irrational, when
it functions badly.
The fieldworker’s activity of representation and reasoning is embedded in a
web of propositional attitudes. She takes various cognitive and conative
attitudes towards the representations that she computes. She believes some of
them, doubts others, presupposes some for the sake of argument, entertains
some as explanations or perhaps as mere theoretical possibilities. She would
like to know whether or why some of them are true. In the absence of such
propositional attitudes that motivate the next step in the research, we do
not have an agent trying to figure out what is the case or how to understand
something.⁶ I do not know how to render explicit the ascription-conditions for
propositional attitudes, but clearly the machine does not meet them. The
machine is merely a sophisticated tool used in research. If the machine continues to record and analyze data long after it has been abandoned as a research
tool, its operations are not involved in any research. There is no agent trying to
figure out one thing or another, even if the machine keeps cranking out
representations of its environment.
When engaging in an enquiry, however theoretical it may be, we are exercising our agency. We are not simply crunching out representations or drawing
inferences willy-nilly but, rather, engaging in purposeful activity. We are
attempting to figure out one thing or another: whether it is the case or how to
understand it. Assessments of our cognitive goals aside, there is room for two
types of shortcomings in the use of our cognitive capacities for the purposes of
⁶ This is not to deny that many of the psychological processes needed to sustain an enquiry are subconscious and even sub-personal. Some of them might not even involve full-fledged intentional states.
Evaluations of Rationality
67
enquiry: first, mistakes in representation or reasoning and, second, use of our
cognitive capacities that does not suit our cognitive purposes. When the genius
interrupts his line of enquiry by teasing out the trivial implications of his beliefs
or by going off on tangents, he is falling short in the second way. He is using his
cognitive capacities in ways that seriously interfere with the pursuit of his
cognitive goal. Such a shortcoming need not be a case of irrationality. Just as
mistakes in representation and reasoning are not failures of rationality unless
they are avoidable given the epistemic position of the agent, this second type of
shortcoming is a failure of rationality only if it should be clear to the agent that
his cognitive activity is not suitable to his cognitive goals. Since the genius is in
a position to appreciate the futility of his little side projects, as well as how badly
they sit with his main line of enquiry, he is guilty of irrationality, even if his representation and reasoning are flawless. His irrationality does not have anything
to do with the importance of his cognitive goals. Rather, his shortcoming qua
RATIONAL being pertains to his ineptitude in using his cognitive powers in ways
that are suitable to his cognitive goals. Thus, evaluations of rationality are sensitive
not only to the agent’s epistemic position but also to his cognitive goals.
V
It may seem as though we have passed from a discussion of theoretical
rationality to a discussion of practical rationality. It is common to draw the distinction between theoretical and practical rationality along the following lines:
theoretical rationality is displayed in regulating beliefs, while practical rationality is displayed in regulating intentions, plans, or actions. It is a familiar idea
that evaluations of practical rationality take into account the goals of the agent.
It may seem dubious to suggest that evaluations of theoretical rationality are
also sensitive to the agent’s goals. One of the sources of theoretical irrationality
is wishful thinking: the agent believes what he wants to be true rather than
what the evidence available to him suggests is true. This might happen in the
context of enquiry, when the enquirer (non-deliberately) gives a faulty assessment of the evidence due to a desire to have results that will lead, say, to a career
advancement. Evaluations of theoretical rationality, it is tempting to conclude,
are essentially about whether the agent forms epistemically well-grounded
beliefs, so they should take into account the epistemic position of the agent and
not his goals.
The genius’s irrationality has nothing to do with forming beliefs for which
he does not have sufficient evidence or for which he has sufficient evidence to
reject. We can understand his irrationality only against the background of his
68
Sigrún Svavarsdóttir
specific cognitive goals, when by ‘specific cognitive goals’ I mean goals of
finding out whether something specific is the case or how it is to be understood. In
the genius’s case, the relevant cognitive goal is to find a design for a device that
makes it perform a specific function (sufficiently or maximally) well. The
genius’s irrationality has to do with how unsuitable his cognitive activity is to
this cognitive goal. However, if we stipulate that his cognitive quirk does not,
and he knows that it does not, risk the completion of his project, his job, his
friendships, etc., it is hard to see this as a case of practical irrationality. True, he
will not be as efficient as he could be in reaching his goals, but we can always
stipulate that the delay is of no concern to him. Nevertheless, I would certainly
deem the genius irrational if I observed him periodically interrupting his line
of enquiry in the way envisioned. Spinning his cognitive wheels in a way that
makes no sense, and he should know makes no sense, in light of the matter under
investigation detracts from the enquirer’s rationality. This, I submit, is better
regarded as a defect in theoretical rather than practical rationality, although that
requires that we broaden our conception of theoretical rationality.
In any case, it seems overly narrow to conceive of theoretical rationality as
being exclusively displayed in forming epistemically well-grounded beliefs.
Consider a scientist who makes, during the course of enquiry, a working
assumption solely for the purpose of gathering more evidence for or against it.
Even if the subjective probability of the proposition is only 50 per cent, the
scientist may be displaying great rationality in assuming it. For, other things
equal, it may be the most promising way of advancing her enquiry at this point.
This may be true even if, given her other concerns, it is doubtful whether it is
practically rational for her to proceed in this way. Say, she has a senior
co-worker who has insisted on a different, less promising, way of proceeding and,
predictably, will be irate at being outdone by her. Assume, moreover, that this
co-worker is in a good position to damage her research career and is evil enough
to be likely to do so out of spite, should she outsmart him in this way. Given
this, there is a real issue of whether it is practically rational for her to go ahead
and proceed in the way she thinks is most promising. However, this does not
seem to be a reason to retract the judgment that she, qua enquirer, would be
rational in making the working assumption. Other things being equal, it is an
intelligent move in an investigation, and it should be regarded as enhancing her
excellence qua RATIONAL enquirer. The way to think about this case, I propose,
is that the scientist would be displaying at least theoretical rationality in
making the working assumption, even if it is an open question whether she
would be practically rational.
It may be objected that the rationality of an agent qua enquirer should not be
construed as theoretical rationality, but rather as a restricted form of practical
Evaluations of Rationality
69
rationality. The agent is being practically rational insofar as his activities are
considered in light of his cognitive goals only; all things considered, he may still
be practically irrational. However, this way of construing rational-qua-enquirer
does not work. Add to the above story that our heroine rationally believes
that her co-worker is willing to ruin their research project rather than see the
major credit for its success go to her. In that case, it is not obvious that our heroine is, judged against the background of (only) her cognitive goals, practically
rational in making the assumption. Nevertheless, she is rational qua enquirer.
Rationality-qua-enquirer is not displayed by doing whatever one rationally
thinks is required to accomplish one’s cognitive goals but, rather, by conducting one’s enquiry in an intellectually resourceful and responsible manner. In
figuring out whether something is the case or how to understand it, it is not
only important to come to a sensible assessment of the evidence while forming
beliefs. It is also crucial that the enquirer exercise good judgment in narrowing
down his hypotheses, in selecting auxiliary hypotheses, in designing and running (empirical or thought) experiments, in constructing proofs, in designing
strategies for gathering more data, and so on. By exercising good judgment in
such endeavors, the enquirer is displaying rationality, and the rationality at
stake seems to deserve the tag ‘theoretical rationality’, though I have little interest in quibbling about the use of that term. Of course, ultimately it may not
matter how an agent came up with an interesting discovery. Even if he stumbled on it accidentally, relied on a wild guess, or prostituted himself, he is theoretically rational in accepting the resulting discovery if he can, retrospectively,
build a good case for its acceptance. But this does not mean that there is no
room for evaluations of theoretical rationality in the context of discovery. It
only reminds us that sometimes it pays to proceed in a manner that does not
demonstrate theoretical rationality.
Our conception of theoretical rationality should be broad enough to allow
that it can be displayed in any type of cognitive activity rather than merely in
forming beliefs. Moreover, the foregoing discussion suggests that assessments
of theoretical rationality need to take into account what it is that the agent is
trying to find out or understand, that is, they need to be sensitive to his cognitive goals as well as his epistemic position. An agent’s theoretical rationality is
reflected in the suitability of his cognitive activity to his cognitive goals as well
as in the quality of his representation and reasoning (subject to the epistemic
limit proviso).⁷
This way of conceiving of theoretical rationality may be thought to have the
undesirable consequence that most of us suffer from a high degree of theoretical
⁷ I add one more component to this conception of theoretical rationality in n. 12 below.
70
Sigrún Svavarsdóttir
irrationality just because, during any normal day, we notice a multitude of
things and draw all sorts of inferences based on the sounds and sights we take
in. It seems odd to think that such cognitive activity, even if it is epistemically
well grounded, undermines our theoretical rationality. But much of this activity appears not to be tailored to any cognitive goal and may, indeed, seem badly
suited to our cognitive goals.⁸ Thus, if evaluations of theoretical rationality
took place against the background of the cognitive goals of the agent evaluated
as suggested above, these routine cognitive activities would appear to undermine our theoretical rationality. They would, indeed, seem to be on a par with
the genius’s cognitive quirks with respect to how the subject’s theoretical
rationality is affected.
I doubt that much of our cognition serves absolutely no purpose. In routine
activities like reading, conversing, or paying attention to our environment,
there are usually cognitive goals at play: we are attempting to understand something or someone. It need not be an arduous attempt, made in a self-conscious
manner. Often, it is only when understanding becomes a challenge that we
resort to conscious strategic thinking and come to think of ourselves as trying
to find something out. In some cases, we may even resist acknowledging that
this is what we are up to. There is room for much self-deception and a lack of
self-awareness in this area. Nevertheless, there is undoubtedly some cognition
that does not serve any purpose ascribable to the subject. We are probably
primed by evolution to notice and draw inferences from certain features of our
environment, whether or not it serves any current purpose of ours. Training
may have similar results. A house painter, for example, may be apt to notice the
flaws in any painted surface and spontaneously draw inferences from her observations. Now, if such purposeless cognitive activity⁹ interferes with the agent’s
cognitive pursuits, the agent’s theoretical rationality is diminished much as the
genius’s theoretical rationality suffers due to his cognitive quirk. But my suggestion
that evaluations of theoretical rationality are sensitive to the cognitive goals of
the agent should not be construed to imply that any aimless cognitive activity
manifests theoretical irrationality on the part of the agent. Such activity may be
too insignificant in the agent’s overall mental life to detract from his theoretical rationality, much as aimlessly drumming his fingers on the desk may be too
insignificant in an agent’s life to detract from his practical rationality.
This leaves the question of whether the quality of the representation and
reasoning carried on in insignificant and aimless cognitive activity affects the
evaluation of the agent’s theoretical rationality. Is the painter irrational when
⁸ Thanks are due to Robert Audi, Don Garrett, and Thomas E. Hill, Jr., for pressing this point.
⁹ I have in mind cognitive activity that serves no purpose ascribable to the agent. It may still have some biological or psychological function.
Evaluations of Rationality
71
she spontaneously forms a poorly grounded belief about the paint used on a
wall while driving past it at high speed? Does she display rationality when she
spontaneously forms a well-grounded belief at a slower speed? My sense is that
the less we can see an individual as engaged in purposeful cognitive activity, the
more his cognitive activity resembles that of the machine imagined in the last
section and the less it is appropriate to think of that cognitive activity as reflecting
on the subject’s rationality.¹⁰ Even evaluations of theoretical rationality should,
I propose, be seen as applying primarily to agents engaged in purposeful
cognitive activity. Theoretical rationality primarily pertains to how well we use
our cognitive capacities in enquiry, when enquiry is construed broadly as any
attempt at discovery or understanding. Using our cognitive capacities well in
enquiry does not merely require that we reason and represent well but also that
our representation and reasoning is suitable to the cognitive goals that set our
line of enquiry.
VI
At this point, my guiding idea should be reformulated as follows: rationality is
the excellence of an agent qua RATIONAL being, assuming that purposeful activity
is the distinguishing mark of agency. I have suggested that theoretical rationality
is excellence in the use of one’s cognitive capacities for the purposes of enquiry.
I now propose that practical rationality be regarded as excellence in the use of
one’s cognitive capacities for whatever purposes one has.¹¹ Such excellence
consists partly in representing and reasoning as well as is feasible, given one’s
epistemic limitations. The quality of the representation and reasoning is
judged by standards that are not agent-relative and have much to do with truth
and truth-conduciveness. In the light of the foregoing discussion, I suggest that
this excellence also requires that the cognitive activity be suitable to the
purposes for which the agent wields his cognitive powers. The agent’s purposes
partly determine what is relevant to his reasoning as well as what successfully
concludes it. They help to set the parameters of relevance and success by determining what is appropriate to think through and investigate, but the nature of
the subject matter, of course, also affects what is relevant at any given stage of
reasoning as well as what makes for a successful conclusion to the reasoning.
¹⁰ That a belief is formed during, or even as the culmination of, purposeful cognitive activity should not
be taken to imply that the belief is voluntary. We do not believe at will, even if the cognitive activity that led
up to the belief was sustained by the desire to know or understand.
¹¹ I think of this as a first approximation of a defensible conception of practical rationality, but I will not
go beyond this approximation here. See, however, n. 15 below.
72
Sigrún Svavarsdóttir
There is one more component in the excellence under consideration:
appropriate responsiveness to the output of the cognitive activity.¹² An agent has
not displayed excellence in the use of his cognitive capacities for his purposes
(that is, in their service), unless he has both used his cognitive capacities well in
figuring out how to attain his ends and put the conclusion to use in advancing
these ends.¹³
I am not proposing that practical rationality is a matter of maximally
advancing one’s ends as neo-Humeans are wont to characterize practical rationality. It need not be irrational to act in such a way as to undermine one’s ends
overall. For the agent may not be in a position to anticipate this outcome. Such
an agent is unfortunate rather than irrational. He is not falling short qua RATIONAL being. However, my conception of practical rationality is very much in the
spirit of Hume. First, practical rationality requires only the kind of reasoning
familiar from the theoretical domain. There is no special type of reasoning—
like submitting one’s preference to the test of the categorical imperative—
required for using one’s cognitive capacities well in the pursuit of one’s ends.¹⁴
Second, assessments of an agent’s practical rationality are relativized to his
purposes such that it is at least conceivable that it would—to echo Hume—be
perfectly rational for a scoundrel to engineer a major disaster for others in order
to avoid a minor inconvenience to himself.¹⁵
¹² This is also a component in theoretical rationality. A theoretically rational agent uses his intermediate
conclusions to advance his enquiry towards its ultimate cognitive goal and his final conclusion to advance, if
feasible, his other cognitive goals (subject to the epistemic proviso). There are some complications, however.
In section V, I gave an example in which it is theoretically rational for an agent to use her intermediate
conclusion in ways that, other things being equal, would advance the enquiry, though they are not likely to
do so in this agent’s circumstances, due to factors extraneous to the enquiry.
¹³ Given the last condition, ‘true irrationality’ in Christine Korsgaard’s sense is possible on my conception of practical rationality. See Christine Korsgaard, ‘Skepticism about Practical Reason’, reprinted in her
Creating the Kingdom of Ends (Cambridge: Cambridge University Press, 1996), 318–25. I believe this opens
up a way to understanding the irrationality of the incontinent person as well as that of the seriously
depressed, although I will not spell out the account here.
¹⁴ When I speak of requirements of rationality, I am simply speaking of necessary conditions for counting
as rational. I reject imperatival and legal models for understanding such talk. There is no mental faculty or
more abstract normative authority—called reason or rationality—that dictates how to think, feel, or behave.
I conceive of practical rationality as a virtue. In failing to use our cognitive capacities in the service of our ends,
we fail to display this virtue rather than violate a command or a law, be it hypothetical or categorical.
¹⁵ Unlike many contemporary neo-Humeans, I think that the relativization has to be to the agent’s actual
ends, or ends that bear a close relation to the agent’s actual ends, rather than to highly idealized ends (e.g.
ends that the agent would have under full information). However, some idealization might be warranted. In
this context, I would also like to note that my conception of practical rationality is naturally extended to
allow for evaluations of an agent’s rationality in adopting goals. An agent does not display practical rationality if he shows ineptitude in using his cognitive powers when thinking through and setting (or revising) his
goals. Such evaluations are, I believe, holistic and are relativized to the agent’s attitudes and goals. I am using
‘goal’ very broadly. An agent’s goal is something that he is set to accomplish, promote, respect, or honor. There
is room here for self-deception and lack of self-awareness just like in the case of cognitive goals. I use ‘goal’,
‘aim’, ‘end’, and ‘purpose’ interchangeably.
Evaluations of Rationality
73
VII
The relativization of evaluations of rationality, both theoretical and practical,
to the agent’s goals is controversial. Although I have motivated it by reflections
on my guiding idea, it is admittedly not implied by that vague idea. Consider
the recognitional view of rationality, sometimes associated with Aristotle¹⁶ or
with Plato.¹⁷ This view construes rationality—the excellence of an agent qua
RATIONAL being—as a matter both of using one’s cognitive capacities well in
discovering reasons for belief and action and of basing one’s beliefs and actions
on these reasons. This view makes perfect sense of why evaluations of rationality apply primarily to agents engaged in purposeful activity. However, it does
not relativize these assessments to the purposes of the agent. For an essential
ingredient of the recognitional view is that reasons for action and belief are not
fixed by the agent’s purposes. Rather, whether a consideration is a reason for
belief or action is determined by whether and how it bears on the truth of the
belief or the value of the action (or its outcome), when both truth and value are
objective in the sense of not being fixed by anyone’s mental stance on the
matter at hand.
It is the commitment to objective values that has deterred many from
embracing the recognitional view of practical rationality. Metaphysical and
epistemological worries are frequently cited as motivations for rejecting this
view.¹⁸ These are not my motivations. By relativizing assessments of rationality
to the purposes of the agent evaluated, I am not attempting a naturalistic
reduction of rationality. In my characterization of rationality so far, I have
relied on the notion of excellence in the use of one’s cognitive capacities for one
purpose or another, and I have not made any promise that this notion can be
unpacked in naturalistic terms. I doubt that the limited unpacking already
done amounts to a naturalistic reduction of rationality. In any case, my
proposal is not motivated by reductive ambitions, even if it might take us a few
steps towards an understanding of rationality that can be incorporated into a
naturalistic conception of ourselves and our surroundings.
It is not that I am a fan of extravagant ontologies or a foe of naturalism.
Rather, the relativization issue is not a metaphysical issue and does not have any
obvious metaphysical implications. This can be best appreciated by seeing that
even emotivists about evaluative discourse can acknowledge both relativized
and non-relativized evaluations. According to them, evaluations that are
¹⁶ See e.g. Garrett Cullity and Barys Gaut (eds.), Ethics and Practical Reason (Oxford: Oxford University
Press, 1997), 13.
¹⁷ See e.g. Onora O’Neill, Bounds of Justice (Cambridge: Cambridge University Press, 2000), 13.
¹⁸ See e.g. Cullity and Gaut (eds.), Ethics and Practical Reason, 15–18; O’Neill, Bounds of Justice,13–14.
74
Sigrún Svavarsdóttir
relativized to the ends of the evaluated agent express attitudes, on the part of
the evaluating agent, conditioned on the assumption that the evaluated agent
has certain ends. In contrast, evaluations that are not thus relativized express
attitudes towards the evaluated agent that are in no way conditioned by what
sort of ends he has. I am not advocating the emotivist account of what is going
on when we make evaluations that are or are not relativized to the ends of the
evaluated agent. What I am highlighting is that this contrast between
relativized and non-relativized evaluations pertains, first and foremost, to the
nature of the mental act of evaluating or the concepts employed in that act. It
is a mistake to run that issue together with the metaphysical issue concerning
whether a value attaches to an object independently of our evaluation of the
object.¹⁹ Of course, once we start to ask about the nature of the act of evaluating, metaphysical issues about the mind-independence of values may arise. I
would have to be convinced that these issues are more pressing or less tractable
if the evaluations are non-relativized rather than relativized.
I reject the recognitional view because it does not relativize evaluations of
rationality to the purposes of the agent evaluated irrespective of whether this
will lead us into a metaphysical quagmire.²⁰ Earlier I considered and rejected
the proposal that the excellence of a being qua RATIONAL being is in part
constituted by not letting an enquiry into less important matters detract from
one’s enquiry into more important matters. This was not driven by metaphysical worries about objective measures of importance or value, but rather by an
intuition about when charges of irrationality are appropriate. Moreover, we
were able to understand the irrationality of going off on tangents and of teasing
out the trivial consequences of one’s beliefs only against the background of the
agent’s cognitive goals. Typically, our cognitive goals are relatively specific: our
goal is to discover or understand some specific phenomenon rather than to
discover any old truth or understand whatever there is to be understood. It is
only because of this that it is irrational to spend one’s energies on discovering or
understanding certain facts at the expense of discovering and understanding
some other facts. Thus, my motivations for relativizing evaluations of rationality
¹⁹ There are, of course, notorious questions about how to construe this metaphysical issue or whether
there is a genuine metaphysical issue here. See e.g. Simon Blackburn, Spreading the Word (Oxford: Oxford
University Press, 1984), 217–20; Ronald Dworkin, ‘Objectivity and Truth: You’d Better Believe It’,
Philosophy & Public Affairs, 25 (1996), 87–139; Gideon Rosen, ‘Objectivity and Modern Idealism: What Is
the Question?’, in Michaelis Michael and John O’Leary-Hawthorne (eds.), Philosophy in Mind (Dordrecht:
Kluwer Academic Publishers, 1994), 277–319; Sigrún Svavarsdóttir, ‘Objective Values: Does Metaethics
Rest on a Mistake?’, in Brian Leiter (ed.), Objectivity in Law and Morals (Cambridge: Cambridge University
Press, 2001), 144–93.
²⁰ Thus, I am committed to resist any other view that does not relativize evaluations of rationality to the
purposes of the agent evaluated, including Kantian views of rationality.
Evaluations of Rationality
75
to the purposes of the evaluated agent rest on intuitions about when charges of
irrationality are appropriate as well as an assessment of how these intuitions are
best explained. This way we arrived at the simple but compelling thought that
ineptitude in using one’s cognitive powers in ways suitable to one’s goals is a
shortcoming of an agent qua RATIONAL being. In its earlier statement, this
thought referred specifically to the agent’s cognitive goals, but I cannot see that
it matters that the goal is cognitive, except insofar as we are distinguishing
theoretical from practical rationality. Thus, I am hoping that the foregoing discussion makes my readers more receptive to a controversial idea that I happen
to find intuitively compelling, namely, that assessments of rationality are
anchored in the purposes of the agent evaluated.
Motivating my proposal by appealing to intuitions may seem dubious. Even
intuitions to which I have already appealed may be turned against me. Imagine
an agent whose sole aim in life is to have a deductively closed system of beliefs.
My proposal has the consequence that it is, both theoretically and practically,
rational for such an agent to work out all the implications of his beliefs. This
seems to go against the intuition about irrationality on which I have been relying, namely, the intuition that it is irrational to spend one’s days teasing out all
the implications of one’s beliefs. Nevertheless, I accept this consequence of my
view. I am not sure that the intuition on which I have been relying is triggered
by the purported counterexample when properly understood and, in any case,
I would not want to rely on an intuition triggered by this imagined scenario. It
is difficult to get an imaginative fix on an agent whose only aim is to have a
deductively closed system of beliefs. You have to imagine an agent who does
not aim to understand anything but the implications of his current beliefs—an
agent who has such a cognitive aim, but not the practical aim of staying alive
and well. Otherwise it is doubtful that we have an agent whose cognitive as well
as other ends are, on balance, served by his working out all the implications of
his beliefs. Well, perhaps we can imagine such a guy. Conjure up the image of
an independently rich guy sitting in his study teasing out all the implications of
his beliefs as a servant brings him sustenance. He passes his days in euphoria,
going through one trivial implication after another until sleep overtakes him.
Do we have an irrational agent or simply a weirdo? I would not let my understanding of rationality hinge on the intuition triggered by this case or similarly
far-fetched cases.
I would be satisfied if my conception of rationality matched and made
sense of intuitions about relatively realistic cases. However, the prospects
for this are grim, especially when it comes to intuitions about practical
rationality. Our intuitions conflict. For example, many find it counterintuitive
to think that an agent with awfully silly endeavors could be practically
76
Sigrún Svavarsdóttir
rational.²¹ Imagine a person who sets out to discover how many men named
John live in Durham, North Carolina. Admittedly, I would suspect this person
of irrationality, both theoretical and practical. For it is difficult to believe that
his life is so impoverished that spending his intellectual resources in this way
does not detract from cognitive goals and other endeavors closer to his heart.
But if I am wrong on this score, I don’t see that his shortcoming is that of
irrationality, especially if he pursues the matter in the most ingenious way. The
man is just silly and shallow—an awfully uninteresting person, indeed. My
intuition here is strong and plays a crucial role in motivating me to advance a
neo-Humean conception of rationality.
Alas, the intuition is not widely enough shared to provide solid grounds for
defending such a conception of rationality. As often in philosophy, we face
conflicting intuitions and a dispute about who bears the burden of argument.
Neo-Humeans tend to assume that what they have to say about the nature of
rationality has to be accepted by everyone as at least a partial account of
rationality, so the burden of proof falls on those who claim there is more to
rationality. However, anti-Humeans have, recently, gone on the offensive
and posed challenges to neo-Humeans.²² They have questioned whether
neo-Humeans can adequately account for the justificatory or critical force of
verdicts of rationality. This is, I believe, a serious issue that neo-Humeans have
to address, though not necessarily on the terms set by their opponents.²³ I aim
²¹ Perhaps the most difficult cases for me involve self-destructive agents. Consider a severely depressed
agent whose only aim is to find a way to end his life, even if he has been informed that with proper medical
treatment he will get out of the depressive state and enjoy life again. Is he displaying rationality when he uses
his cognitive capacities, in the most ingenious way, to fool the person who stands a suicide guard over him
and to find the means to end his own life? If he fails to do this, is he failing to display the virtue of rationality? My intuitions are shaky about this case and I am wary of resting my conception of rationality on the
intuitions that this example triggers (in me or others). This case involves a person suffering from mental illness. Although we often continue to evaluate the mentally ill in terms of rationality or irrationality, it is far
from clear that it is appropriate to apply such evaluations to them. Attempts to understand insanity in terms
of irrationality seem misguided. I would like to come back to this issue, once I have provided a deeper and
more theoretical understanding of why the virtue of rationality is best regarded as excellence in the use of
one’s cognitive capacities in the service of one’s ends—the sort of understanding for which I call below.
Thanks are due to Walter Sinnott-Armstrong for pressing the case of the suicidal depressive on me.
²² See Christine M. Korsgaard, ‘The Normativity of Instrumental Reason,’ in Garrett Cullity and Berys
Gaut (eds.), Ethics and Practical Reason, 215–54; Derek Parfit, ‘Reasons and Motivation,’ The Aristotelian
Society Supplementary Volume, 71 (1997),99–130; and Warren Quinn, ‘Rationality and the Human Good,’
reprinted in his Morality and Action (Cambridge: Cambridge University Press, 1993), 210–27.
²³ For example, Kantian formulations of the challenge tend to presuppose that neo-Humeans accept that
there is a hypothetical imperative of reason. But I reject imperatival and legal models of practical rationality.
See n. 14. Also, the claim is usually that neo-Humeans cannot do justice to the normative force of verdicts of
rationality. The notion of normativity is much thrown around, but it is not entirely clear what it designates. It
seems, however, to relate to the notion of obligation or at least of (deontic) ought. Since I am proposing that we
think of rationality as a virtue, I question whether there is anything like requirements of rationality expressed
by a sentence containing a deontic ought. See also n. 14. The challenge is, therefore, better formulated as questioning whether neo-Humeans can account for the justificatory and critical force of verdicts of rationality.
Evaluations of Rationality
77
to build a defense of my view of rationality that meets such challenges.
However, here is not the space to do so. A brief explanation of how I conceive
of that project will have to do for now.
VIII
There are two main questions that need to be addressed: First, are there
genuine evaluations that pertain to how well people use, given their epistemic
position, their cognitive capacities in the service of their ends? Second, assuming
that these are genuine evaluations, why is rationality best understood as the
virtue epitomized in this type of evaluation? Both questions need some clarification. It is far from obvious what would constitute adequate answers to them.
The first question, I submit, requires that we be given a good sense of why it
matters that people use their cognitive capacities well in the service of their
ends—matters in such a way that this can be considered a virtue. The second
question, I submit, requires that we be given a good sense of why an excellence
in the use of one’s cognitive capacities in the service of one’s ends is the virtue of
agents qua RATIONAL beings. Moreover, it must be elucidated why the verdict
that someone uses his cognitive capacities well in this way has a justificatory
force, while the opposite verdict has a critical force—the sort of justificatory
and critical force that verdicts of rationality and irrationality are commonly
understood as having.
It is far from trivial to meet these requirements. This can best be appreciated
by imagining an evil or a silly person who uses his cognitive capacities superbly
in advancing his evil or silly ends.²⁴ Can we really think of him as justified in
his evil or foolish ways just because they result from an excellent use of his
cognitive capacities in the service of his ends? Don’t his ends have to be worthy
as well? Can we really think of the excellence of this person’s use of his cognitive
capacities as mattering in the way that virtues matter? Can we think of it as the
virtue of agents qua RATIONAL beings? Does that virtue not also consist in
having worthy ends—or at least ends that are such that there is a reliable
connection between using one’s cognitive capacities well in the service of them
and ending up having true beliefs and doing valuable things? Isn’t that required
for verdicts of rationality to have the justificatory or critical force that they are
commonly understood as having?
These challenges assume that the justificatory force of verdicts of rationality
and the status of rationality as a virtue cannot be elucidated unless rationality
is shown to serve reliably truth and value. This is an assumption that I want to
²⁴ Compare Warren Quinn’s challenge to neo-Humeans in ‘Rationality and the Human Good’.
78
Sigrún Svavarsdóttir
challenge, even if I believe that, as a matter of fact, rationality tends to serve
truth and value. Understanding why rationality is a virtue and, in particular,
why verdicts of rationality have justificatory force requires that we probe
deeper into the nature of the type of agents—RATIONAL agents—that are
subject to evaluation of rationality. What sort of agents can be justified in or
criticized for forming a belief or an intention to act? It seems that the ability to
use one’s cognitive capacities purposefully does not suffice for being such an
agent. What is required is the ability to reflectively assess and revise one’s
beliefs, intentions, and possibly some other mental attitudes. Such reflections
can only take place against the background of other beliefs and concerns. There
is no way of stepping outside of one’s mental stance to figure out what to believe
or do. However, an agent may use his cognitive capacities better or worse in the
course of reassessing and regulating his beliefs, intentions, and other mental
attitudes. Evaluative notions such as responsibility, conscientiousness, and
integrity find their natural niche in assessing how such an agent regulates his
cognitive and practical endeavors. So does, I want to argue, the notion that an
agent can be more or less justified in forming a belief or an intention.
Such deepened understanding of RATIONAL agency is the key for understanding why evaluations of rationality have a credible justificatory or critical
force, even if they are relativized to the agent’s mental states in such a way that
there is no guarantee that rationality will serve truth and value. It will also cast
light on why rationality, as understood here, matters in such a way that full
justice can be done to my guiding idea: namely, that rationality is the virtue of
agents qua RATIONAL beings. At the same time, it will require that my conception
of practical rationality be further developed. An agent can display practical
rationality not only in acting one way rather than another but also in forming
and revising intentions, plans, and other commitments. Moreover, evaluations
of practical rationality need to take into account not only the agent’s goals but
also his various attitudes regardless of whether they are reflected in his current
aims. Alas, I have issued a lot of promissory notes. Let me stop before I sink
further into debt.
4
Intrinsic Value and Reasons for Action
Robert Audi
The concept of the good—or at least some concept of the good—has been a
central topic in ethical theory since Plato. Moral philosophers have generally
recognized a distinction between what is good only as a means and what is good
in itself. The latter, often called the intrinsically good, has proved difficult to
explicate and, by contrast with the former, seems to many philosophers to be
mysterious. G. E. Moore sought to clarify the notion of the intrinsically good,
and his theory of value was probably the most influential axiology developed
in the twentieth century. The theory is closely tied to his metaphysics, his
epistemology, and his normative ethics; but much of what he says about the
good can be maintained without commitment to all of his major claims in
those three domains. My project here is to set out a theory of intrinsic value that
incorporates the best elements of Moore’s account of the notion but avoids
commitment to his overall view in metaethics.
1. Some major elements in Moore’s theory of value
Since my aim is to produce a sustainable account of intrinsic value, I must be
brief in introducing the elements in Moore’s account that form a good basis for
critical discussion and for comparison with my view. I begin with his famous
claim
that ‘good’ is a simple notion, just as ‘yellow’ is a simple notion; that, just as you cannot,
by any manner of means, explain to any one who does not know it, what yellow is, so
This paper has benefited from comments by Stephen Barker, Panayot Butchvarov, Roger Crisp, Jonathan
Dancy, James Dreier, Brad Hooker, Judith Thomson, Derek Parfit, Robert Stecker, and Mark Timmons, and
from discussions at the Universities of Georgia, Memphis, Missouri, Notre Dame, and Oxford. In Southern
Journal of Philosophy, 61 (2003 supplement) where this paper first appeared, Barker published a commentary immediately following my paper. My reply is in ‘Intrinsic Value, Inherent Value, and Experience,’
Southern Journal of Philosophy, 61 (2003): 323–7.
80
Robert Audi
you cannot explain what good is. Definitions of the kind that I was asking for,
definitions which describe the real nature of the object or notion denoted by a word,
and which do not merely tell us what the word is used to mean, are only possible when
the object or notion in question is something complex.¹
Moore also says that he means by ‘good’ a property (17). His terminology and
his uses of inverted commas vary considerably. Tracking them would be a
major task, and rather than undertake that I shall simply be guided by the
context in interpreting him. For him, as for most writers on intrinsic value, the
phrases ‘intrinsically good’ and ‘intrinsically valuable’ are sometimes used
interchangeably, though with the understanding that the theory of intrinsic
value extends to the notion of what is intrinsically bad. (This is sometimes
called disvaluable.)
Moore is famous for his ‘principle of organic unities’: ‘The [intrinsic] value
of a whole must not be assumed to be the same as the sum of the [intrinsic] values of
its parts’ (28; cf. 29, 152). He illustrates this with an aesthetic case:
to be conscious of a beautiful object is a thing of great intrinsic value; whereas the same
object, if no one be conscious of it, has certainly comparatively little value, and is commonly held to have none at all. But the consciousness of a beautiful object is certainly
a whole of some sort in which we can distinguish as parts the object . . . and the being
conscious . . . Now this latter factor occurs as part of a different whole, whenever we are
conscious of anything; and it would seem that some of these wholes have at all events
very little value, and may even be indifferent or positively bad. . . . Since, therefore,
mere consciousness does not always confer value upon the whole of which it forms a
part, even though its object may have no great demerit, we cannot attribute the great
superiority of the consciousness of a beautiful thing over the beautiful thing itself to the
mere addition of the value of consciousness to that of the beautiful thing. (28)
I particularly want to bring out here two points whose significance Moore
himself (among many others) seems to have missed. First, he is attributing intrinsic
value to two radically different kinds of things: consciousness and, on the other
hand, external objects, where consciousness is conceived as internal to the mind,
or at least as having an element that is. Secondly, he is exhibiting an awareness
that the value, even of a beautiful object, is comparatively little in abstraction
from the consciousness of it and, thus abstracted, is ‘commonly held’ to be nil.²
¹ G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1903), 7 (references to this
book will hereinafter be included parenthetically in the text).
² Cf. Sidgwick, The Methods of Ethics, 7th edn. (London: Macmillan, 1907), 114 and Moore, Principia
Ethica, 221–5. Moore says in one place that although ‘the mere existence of what is beautiful does appear to
have some intrinsic value . . . Prof. Sidgwick was so far right in the view there discussed, that such mere existence of what is beautiful has value, so small as to be negligible, in comparison to that which attaches to the
consciousness of beauty’ (189).
Intrinsic Value and Reasons for Action
81
This duality regarding what sorts of things are intrinsically good significantly
contrasts with the view Moore held by the time he published Ethics nine years
later. He says there that ‘It may, in fact, be held, with great plausibility, that no
whole can ever have any intrinsic value unless it contains some pleasure.’³
Pleasure, by contrast with paintings, is or at least entails a form of consciousness. He rejects, however, the ostensibly implied view that ‘intrinsic value is
always in proportion to quantity of pleasure’ (Ethics, 152; cf. 153). For, on the
principle of organic unities, the addition of pleasure to a whole can reduce its
intrinsic value; and (in a quite different formulation) he reaffirms this principle
in the context (151–2). He also holds that
Whatever single kind of thing is proposed as a measure of intrinsic value, instead of
pleasure—whether knowledge, or virtue, or wisdom, or love—it is . . . not such a
measure; because it is quite plain that, however valuable any of these things may be, we
may always add to the value of a whole which contains any one of them . . . by adding
something else instead. (152)
His positive view concerning value and consciousness is that ‘it does seem as if
nothing can be of intrinsic value unless it contains both some feeling and also
some other form of consciousness’ (153). (A similar view is found in Ross’s The
Right and the Good.⁴)
There is one other element in Moore’s conception of intrinsic value that I
want to bring out before I present my own account. In responding to Frankena
(in 1942), Moore says that
if what Mr. Frankena means to assert is that the propositional function ‘x is
intrinsically good’ may be identical with the function ‘the fact that an action which
you can do would produce x is some reason for supposing that you ought to do that
action,’ then one condition necessary for the possibility of this being true is fulfilled . . . there is a two-way necessary connection between these functions . . . But
nevertheless I think there is a good reason, if not a conclusive one, for doubting
whether they are identical. . . . Is it not possible to think that a thing is intrinsically
good without thinking that the fact that an action within our power would produce
it would be a reason for supposing that we ought to do that action? It certainly seems
as if we can.⁵
³ See G. E. Moore, Ethics (Oxford: Oxford University Press, 1912), 148.
⁴ See W. D. Ross, The Right and the Good (Oxford: Oxford University Press, 1930), 86. In n. 14 of ch. 11
of Moral Knowledge and Ethical Character (Oxford: Oxford University Press, 1997), I comment on Ross’s
view of the element of consciousness in the intrinsically good.
⁵ G. E. Moore, ‘A Reply to my Critics,’ in Paul Arthur Schilpp (ed.), The Philosophy of G. E. Moore (New
York: Tudor Publishing Co., 1942), 151–2. Note that in treating reasons for action as entailing the production of intrinsic goodness, Moore either presupposes that reducing such evils as pain is producing such a
good or simply neglects to take into account that it is with intrinsic value rather than with positive intrinsic
value alone that reasons for action are essentially connected.
82
Robert Audi
The view Moore takes Frankena to attribute to him—roughly, an identification
of the intrinsic goodness of something with the existence of a reason for favorable action towards it—is important and represents one way (though not
Moore’s way) of explicating intrinsic goodness. A theory of intrinsic value
should help us appraise this view, and I shall return to the question.
2. Intrinsic, instrumental, and inherent value
I have emphasized that Moore attributes intrinsic value both to consciousness
of a beautiful object and to that object itself. He also attributes it to such things
as knowledge, virtue, wisdom, and love.⁶ Aware that it is ‘commonly held’ that
a beautiful thing by itself has no intrinsic value, and wanting to do justice to the
role of pleasure as an intrinsic good, he expresses sympathy for the view that
pleasure is intrinsically good, and (as noted) he maintains that nothing can be
of intrinsic value unless it contains both some feeling and also some other form
of consciousness, though (in Principia) without suggesting that pleasure must
be an element in the consciousness in question. It apparently did not occur to
him that a beautiful painting ‘contains’ neither form of consciousness. And
consider the other elements in question. Knowledge, virtue, wisdom, and love
need not contain any feeling, at least if this means that they can exist over an
interval of time only if their possessor has some feeling during at least part of
that time. Moreover, when they are not being manifested, they do not even
entail the occurrence of any form of consciousness. Granted, all of them may
have conceptual connections to consciousness. But nothing about them will
sustain Moore’s suggested view.
I believe that Moore has not done adequate justice to the data he himself
recognizes and that a more fine-grained theory is required to account for them.
I propose (as I have in earlier work⁷) that we take the bearers of intrinsic value to
be concrete experiences on the part of individuals (in this paper I consider only
experiences of persons, but I acknowledge the existence of valuable experiences
on the part of non-persons). I cannot give a full-scale analysis of the notion of an
experience (and will return to the notion), but what I want to say here may be
clear enough on the basis of the rough idea that an experience (1) is an occurrent
state of consciousness, such as paying careful attention to a painting, or a certain
kind of unified set of events in consciousness, such as daydreaming about a vacation trip; and (2) has an object, at least in the sense of a content, hence is not
⁶ See e.g. Moore’s at least implicit ascription of intrinsic value to knowledge, virtue, wisdom, and love in
Ethics, 152.
⁷ In ch. 11 of Moral Knowledge I have discussed bearers of intrinsic value in some detail.
Intrinsic Value and Reasons for Action
83
merely a matter of one’s being awake. An experience cannot be what we might
call a period of empty consciousness: something like an utterly blank moment in
which one has a mere receptivity to stimuli but is aware of nothing.
I am taking the notion of experience central here to be internal in a sense
implying that an experience need not have an external object, as it does in the
case of something that is actually seen and is thus, in an external sense, visually
experienced and also experienceable by others. The experience as of seeing the
same object when one is only hallucinating would be internal and in a sense
non-relational, since its occurrence is not even partly constituted by a relation
to anything other than the person having it.⁸ But suppose one is actually
experiencing a symphony. Call this second case experience in the relational
sense. If one has the latter, one has the former as what might be called its
psychological base, but not conversely. It is apparently experiences in the
internal sense that are the basic bearers of intrinsic value. This does not imply
that enjoying a symphony cannot have intrinsic value. It surely does; the proposal
is only that it has that value in virtue of its internal, experiential qualities, hence
that what might be called basic intrinsic value is wholly experiential. Thus,
silently enjoying reciting a poem to oneself would be intrinsically good;
being pained at the thought of having cancer would be intrinsically bad.
Hallucinatory experiences might also have intrinsic value on this view, and that
possibility calls for comment. Before I address difficulties for the view I am
proposing, however, let me develop it further.
How can the experientialist account of intrinsic value do justice to the value
of a beautiful painting, which is not an experience? Isn’t it good ‘in itself,’ hence
intrinsically good? The phrase ‘in itself ’ has doubtless been misleading. I find
it too crude to serve centrally in understanding intrinsic value. It plainly applies
to the goodness of an intrinsically good experience: such an experience is not
only non-instrumentally good, it need have no relational kind of goodness. But
‘good in itself ’ also applies to things whose goodness is quite different, in ways
I will bring out, from intrinsic goodness. Granted, since the non-instrumental
goodness of a beautiful painting resides in its beauty and that, in turn, is
consequential (supervenient, in one sense of that term) upon its intrinsic—in
the broad sense of ‘non-relational’—properties, it is very natural to think of
the painting as good in itself. Moore and many others have apparently so
conceived such objects and partly for that reason considered them intrinsically
⁸ I have in mind ‘concrete’ objects; on some views, an experience of, for example, thinking about a problem or even hallucinating a painting requires being in a relation to propositions or properties. It might also
be noted that on a content externalist view, certain experiences are not possible apart from the existence of
external objects; but supposing this is so, those objects need not be, as it were, constituents in the experience,
rather than suitable factors in the life of, the subject.
84
Robert Audi
good. But here we do well to recall Aristotle. Among the pertinent points
he made are these: that one good is more ‘final’ than another if we seek the
latter for the sake of the former, and that the good is that which makes life
‘choiceworthy.’⁹ Surely we (properly) value (e.g. go to see, preserve, celebrate,
etc.) beautiful paintings in order to view them, and indeed not merely to view
them but to do so in a way that is aesthetically rewarding. Viewing them in that
way yields having an aesthetically valuable experience (by which I mean one that
is good from the aesthetic point of view, not one that is a good object of aesthetic
appreciation, though that status is not ruled out for special cases). I take
experiences having such value to be intrinsically good. I also take them to be
‘more final’ in Aristotle’s sense than their objects.
A related point emerges when we ask how one should conceive beautiful
paintings in relation to Aristotle’s thesis about the good. Do they contribute to
the ultimate choiceworthiness of a life simply by their physical presence in our
environment or through our viewing them—hence visually experiencing
them—in a way that is aesthetically good? Plainly they would not so contribute
if we never viewed them or, on viewing them, never had a good experience in
doing this or in experiencing anything else in which they figure.
Are beautiful paintings, then, just instrumentally good after all, say means to
aesthetic enjoyment? One could call them constitutive means; for they are
constituents in, and not mere instruments for attaining, such enjoyment. But
the term ‘means,’ used by itself, is highly misleading. They are not means in the
ordinary, instrumental sense implying a non-constitutive, contingent role in
bringing about the intrinsic good in question. They are, in Moorean terms,
part of it. We cannot have the pleasures of reciting Shakespeare’s sonnets
without reciting them. We can produce heat in our homes by various alternative means.
A useful term to employ in reference to objects that are good in themselves
is ‘inherent value’. C. I. Lewis called inherent values those
which are resident in objects in such wise that they are realizable in experience through
presentation of the object . . . The value in question is one which is found or findable in
the object itself . . . in the sense of being one which is disclosed or disclosable by
observation of this object and not by examining any other object.¹⁰
⁹ Aristotle says, ‘We call that which is pursued as an end in itself more final than an end which is pursued
for the sake of something else’ (Nicomachean Ethics, trans. Martin Ostwald (Indianapolis: Hackett, 1962) ).
Cf. Terence Irwin’s translation, 2nd edn. (Indianapolis: Hackett, 1999), which uses ‘complete’ in place of
‘final’. I take Aristotle to be referring not to mere pursuit but to a kind that is ‘proper.’ The idea of the ‘selfsufficiency’ of happiness (the good), in virtue of which it makes life choiceworthy, is explicit in 1097b.
¹⁰ C. I. Lewis, Analysis of Knowledge and Valuation (La Salle, Ill.: Open Court, 1946), 391 (Lewis makes
no reference to Moore in the context).
Intrinsic Value and Reasons for Action
85
By contrast, he saw ‘intrinsic value [as] attaching exclusively to realizations of
some possible value-quality in experience itself ’ (389; cf. 390–1). He called
this realization the only ‘ultimately good thing’ (390) and went on to advance
the experientialist view that ‘What is ultimately desirable is not merely that this
object and this property of it called its beauty should exist, but that this beauty
of it should illuminate the experience of some beholder’ (391).
Lewis does not give a detailed account of intrinsic value or, especially,
inherent value. I make no claim to be giving an interpretation of his view; but
I find the term ‘inherent value’ useful, and I want to clarify the notion I have
in mind.
First, I seek a wider notion than the one sketched by Lewis. We should allow
that an experience of silently reciting a poem can be intrinsically good; but this
is not a case of observation. I suggest that we conceive an inherently good
thing more broadly, as roughly one such that an appropriate experience of it is
intrinsically good. There are many kinds of appropriateness. A mere visual
experience of seeing a beautiful painting need not be intrinsically good; some
aesthetic appreciation is normally required. A mere walk on a beautiful beach
may be dull; a zesty one graced by mild temperature and a fresh breeze may be
delightful. And an aesthetic appreciation of a silent poetic recitation can be
intrinsically good even if there is no external object thereof.
More must be said to explicate appropriateness (more, unfortunately, than
can be said here), but a necessary condition is that an appropriate experience of
an inherently good thing be an experience of a suitable subset of its intrinsic
properties (though not necessarily of all or only the intrinsic properties of it).
This conception suggests that inherent value has at least a great deal in
common with what are usually called secondary qualities. But if we take it to
be such a quality, we must note that there is no implication that inherent value
properties are not ‘in’ the object. Indeed, they are non-relational, and in that
sense intrinsic, properties; but their constitutive manifestations, like those of
colors, entail relations to living things.
I also want to allow, as Lewis might not have, that what is intrinsically good
can also be inherently good. How can this be if the bearers of intrinsic value are
experiences? Surely there can be intrinsically good second-order experiences.
Suppose I view with pleasure my own enjoyable silent recitation of a poem.
May I not be taking pleasure in the contemplation of the enjoyable experience,
happily noting that I seem to be getting intonations that are difficult to
capture? Granted, this could be argued to be not an experience of reflecting on
another experience, but simply a complicated first-order experience. It is, however, a reflective experience in which one thinks about a present experience,
whereas the recitation (for most poems) involves no such thoughts. This
86
Robert Audi
difference in content would seem to warrant countenancing two different
experiences; and the content of the former apparently warrants considering
that experience second-order.
Perhaps, moreover, we may go further and allow that I can experience, for
instance, your pleasure in singing a song you audibly delight in singing. If I can,
then the intrinsically valuable first-order experiences of others can also have
inherent value for me: my experience of your pleasure can be intrinsically good.
To be sure, I must experience your pleasure through its expression in your
behavior; but there is a sense in which I see your pleasure in that conduct as
well. We may thus take ‘inherently good’ to apply to certain intrinsic goods
constituted by various kinds of first-order experiences. The case also illustrates
a kind of relativity possessed by inherent, as opposed to intrinsic, goodness.
The latter is experientially realized in the life in which it occurs; the former,
though good ‘in itself,’ is realizable in intrinsically good experiences only on the
part of beings who can appropriately experience it. Not everyone can enjoy
atonal music or abstract art.
The general idea suggested by these reflections is that something is inherently good provided that an appropriate experience of it is intrinsically good,
and it is inherently bad provided an appropriate experience of it is intrinsically
bad, where the range of appropriate experiences is limited to those that are
responses to certain of its intrinsic properties. (These will include at least some
of the properties of the thing in question on the basis of which it is plausibly
considered good or bad in itself.) Which properties are crucial is, in a certain
way, relative to the kind of value in question. If the value is aesthetic, then the
appropriate experience must include some sense of the object’s aesthetic properties (or at least those they are grounded in). If the value is moral, the experience
must similarly include a sense of morally relevant properties, such as harming
or helping. If the value is philosophical, the experience must include considering conceptual or other ‘intellectual’ properties. And so forth. In the light of
the kinds of properties in question, an experience may be aesthetically good or
bad, morally good or bad, and so on for other kinds of value. An experience
of certain harmonious sounds, for instance, can be aesthetically good; an
experience of being tempted to cheat can be morally bad, say shameful.
There is another part of my characterization of inherent value that needs
comment. Suppose A, a bully, maliciously gives B, who is timid and much
smaller, an unprovoked sharp slap in the face. I want to call this deed inherently
(morally) bad. One appropriate experience in response to it is indignation. Is
the experience of indignation in such a case intrinsically bad? It is likely to be
both unpleasant and morally distressing. Equally important, perhaps, one
might quite reasonably prefer that it not have been called for. On the other
Intrinsic Value and Reasons for Action
87
hand, it is a response that we are glad an observer has in such a case. Our
positive attitude toward it is a clue to its status. Perhaps we should say that the
response may be, overall, both inherently good and intrinsically bad. As an
experience, it may be intrinsically bad overall (though it need not be, depending
on the mix of distress and, say, felt righteousness); but in the circumstances, it
is good that we have it: our having it is a state of affairs that is inherently good
and can be viewed, and in that sense experienced, with moral satisfaction.
Similarly, suppose that A sadistically teases B and greatly enjoys this. A’s
experience is intrinsically good, at least hedonically; B’s experience, let us
assume, is one of suffering and is intrinsically bad. Still, the overall concrete
event, A’s enjoying sadistically teasing B and thereby causing B’s suffering,
seems inherently bad. We properly tend to feel at least displeasure, and perhaps
a kind of moral repugnance, in the contemplation of it; and its inherent
badness can provide sufficient reason to intervene. To say that it provides a
reason is not to say it constitutes a reason; I take providing a reason to entail the
existence of a reason, usually either by having the reason as its content, as with
desires and intentions, or by virtue of something quite evident in the context,
as where it is obvious that intervening would prevent the bad thing from
coming to be. Here what constitutes the reason is (generically) the fact that the
action would prevent the suffering in question.
3. Intrinsic value, inherent value, and organic unity
This case of taking pleasure in someone else’s suffering can be used to help in
motivating another principle concerning value, inherent as well as intrinsic.
Why is displeasure in experiencing a sadist’s action and its effect appropriate?
And why is it so even if the sadist’s pleasure is very great and the suffering in
which it is taken is only slight? It seems to me very plausible to say that this
pleasure ill-befits its object; there is a basic inappropriateness in taking pleasure in
someone else’s suffering—the more so if one is causing it. It is thus appropriate
to be displeased upon experiencing someone’s taking pleasure in someone else’s
suffering. On my view, even if the pleasurable experience of the sadist is intrinsically good, it may be inherently bad. I think that (depending on the details)
it almost certainly will be inherently bad, even if the pleasure is quantitatively
greater than the suffering the sadist causes. For the pleasure has an inappropriate kind of object; and accordingly, an appropriate second-order experience of
that pleasure is intrinsically bad in the relevant way.
Why is it that suffering is an inappropriate kind of object of pleasure? There
is much room for theory here. One answer available to those who countenance
88
Robert Audi
intrinsic value and connect it with objective reasons for action in the way
Moore and many others do is as follows. Consider the point that there is a basic
reason (though not necessarily an undefeated reason) for action to realize or
continue the pleasurable thing, whereas there is a basic reason (though not
necessarily an undefeated one) to avoid or extinguish a painful or unpleasant
thing. Taking pleasure in causing or even observing someone else’s suffering,
then, provides a reason for producing or sustaining something that there is
reason to avoid. The kind of experience in question has properties with opposing
valences: one kind, we might say, calls for realization, the other for avoidance.
There is no contradiction here, but there is a kind of axiological incoherence,
at least on the assumption of the objectivity of the kinds of reasons in
question.¹¹ This in part accounts for the intuition that such pleasure ill
befits its object. If we extend this idea, it can be applied to taking pleasure in
thinking about causing someone suffering. Envisaging or imagining suffering
is still a case of having before the mind something of negative valence and
taking a positive (non-instrumental) attitude toward it. It mirrors, in a sense,
the enjoyment of actual sadistic conduct.¹²
I am suggesting that sadistic pleasure is (perhaps with special exceptions)
inherently bad in part on the basis of its intentional object, the kind such pleasure has even if the sadist only mistakenly thinks the other person is suffering or,
somewhat similarly, is taking pleasure in the mere thought of someone’s
suffering. I am not assuming, but am allowing, that, overall, the pleasure is also
intrinsically bad. It may seem that it must be so at least where the suffering
caused is greater than the pleasure taken in causing it. But perhaps it need not
be so in special circumstances, such as causing suffering in administering
deserved punishment, as where the jailer—within limits—takes pleasure in
causing the pain that an unrepentant violent criminal feels in being locked up.
In any event, the intrinsic value of the pleasure cannot be decided on the basis
of how much suffering the pleasurable action causes; for in that case we have an
experience with an external effect: the having of that object is a relational
property of the experience, whereas I am assuming, as Moore and others have,
that the bearers of intrinsic goodness have it in virtue of at least one of their
¹¹ That pleasure and pain (whether actual or in prospect) provide, as Moore thought, objective reasons
for action is of course controversial; I assume it in this paper, but have argued for it in detail in The
Architecture of Reason (Oxford: Oxford University Press, 2001), esp. chs. 4–6. Cf. Principia, where Moore
says, ‘when I talk of a thing as “my own good” all that I can mean is that something which will be exclusively
mine, as my own pleasure is mine (whatever be the various senses of this relation denoted by “possession”) is
also good absolutely; or rather that my possession of it is good absolutely . . . everyone else has as much reason
for aiming at it as I have myself ’ (99).
¹² The attitude must be non-instrumental to rule out cases in which one takes pleasure in someone’s suffering because it is a means to, for example, the person’s recovery from depression or a cleansing penance for a wrong.
Intrinsic Value and Reasons for Action
89
non-relational properties. The complex consisting of an experience and its
external effects can have inherent but not intrinsic value.¹³
The sorts of cases we are considering make it natural to posit Moore’s
principle of organic unities. I prefer, however, a broader formulation that specifies
parts or aspects of the bearer of value. I also think that the resulting principle
should be applied to inherent as well as intrinsic value. Recall the sadist’s taking
pleasure in someone’s suffering. Even if the pleasure is much greater than the
suffering, the inherent value of the complex whole can be negative, illustrating
that the inherent value of the whole is not the ‘sum’ of the inherent values of the
‘parts’ (I assume that these are the only parts, or the only relevant ones, and that
the relevant notion of a sum is not strictly quantitative).¹⁴
As to the case of intrinsic value, we can speak of one aspect of the experience
of pleasure in someone’s suffering as the pleasant quality of the experience and
of another as its content, in the sense of its being directed at the suffering. We
can now say that the pleasant quality of the experience is inappropriate to this
content and that even if the former is intense, the overall intrinsic value of the
experience is either negative or at any rate much less positive than that of an
experience differing from it only in the pleasure’s being taken in something
neutral. The same point applies even if we suppose the person only mistakenly
thinks the other is suffering and we hold all else fixed. It is important to see,
however, that even if inherent badness is the only kind of badness possessed by
these experiences, it may have sufficient magnitude to warrant the kinds of
negative appraisals we require, for instance judging that such experiences are
non-instrumentally bad and that one should avoid bringing them about.
Aesthetic cases more readily illustrate the organicity of inherent and intrinsic
value. Consider paintings and poems. Both can have parts and aspects that
have no inherent value, such as a blank white space in a painting, a certain
rhythmic shift in a musical composition, and ellipsis marks in a poem. But the
overall inherent value of these artworks may be very positively affected by such
elements, so that the value of the whole is greater than the sum of the values of
¹³ If one wonders how a relational experience, such as pleasure in a symphony, can have intrinsic value at
all, given that the symphony is not intrinsic to its internal side, two points should be stressed: first, the
symphony is intrinsic to the experience qua relational; but, second, the internal side of that experience is also
intrinsic to it, and it is in virtue of this (which includes intentional content) that the overall experience has
intrinsic value.
¹⁴ The non-additivity claim in the principle of organic unities does not entail that there cannot be a case
in which intrinsic or inherent value is additive. Consider two qualitatively identical experiences. Might we
say that—provided one has a determinate value—the pair has twice that value? This would be important for
the theory of rational action as well as for the theory of value. Extensive discussion of the principle and a
defense of it are provided by Noah Lemos in Intrinsic Value: Concept and Warrant (Cambridge: Cambridge
University Press, 1994) and ‘Organic Unities,’ Journal of Ethics, 2 (1998). For criticism of the principle, see
Michael A. Zimmerman, The Nature of Intrinsic Value (Lanham, Md.: Rowman & Littlefield, 2001).
90
Robert Audi
the parts or aspects. This can hold even if all of the parts and aspects are
inherently good. Similar points apply to intrinsically valuable experiences.
Consider a pause in musical work: as experienced in itself, it may be
aesthetically empty but, as part of the overall musical experience, valuationally
important.
4. Some problems of specificatory incompleteness
An apparent difficulty for my view may seem to arise in the light of some
important points made by Aristotle in the Nicomachean Ethics:
Good is spoken of in as many ways as being . . . as God and mind; in quality, as the
virtues; in quantity, as the measured amount; in relation, as the useful . . . Hence it is
clear that the good cannot be some common [nature of good things] that is universal
and single; for if it were, it would be spoken of in only one of the categories, not in them
all. (1096a 24 ff.)
Following Aristotle, one might point out, as Peter Geach and Judith Jarvis
Thomson have, that, contrary to the impression Moore sometimes invites, a
thing cannot be just good.¹⁵ It must be a good F, for some appropriate F, or
good in a way, or good as a thing of a certain sort, and so on. Is Moore
committed to denying this? I do not see that he is; in any case, he need not be
if, like many others (including me), he takes goodness to be a consequential
property, as he seems to.¹⁶ By this I mean roughly that things are good in virtue
of other properties they have, in a sense of ‘in virtue of ’ that implies (but is not
equivalent to) the impossibility of two things differing in their goodness
though they are alike in all their other properties. For intrinsic and inherent
value, the relevant base properties are, in my view, intrinsic and, in some
intuitive sense to which I shall return, ‘descriptive.’
If this view is correct, then we might expect the context of a predication of
goodness, especially inherent on intrinsic goodness, to indicate the relevant
kind of thing.¹⁷Call this a value-anchoring kind, since specifying it anchors the
¹⁵ See Peter Geach, ‘Good and Evil,’ Analysis, 17 (1956): 33–42; and Judith Jarvis Thomson, ‘The Right
and the Good,’ Journal of Philosophy, 94 (1997): 273–98.
¹⁶ Moore’s term for ‘consequential upon,’ at least in what I believe is his last published essay on intrinsic
value, is ‘depends on.’ He says, for example, that whether and to what degree anything possesses intrinsic
value ‘depends solely on the intrinsic nature of the thing . . . anything exactly like it, must, in all circumstances,
possess it in exactly the same degree.’ See ‘The Conception of Intrinsic Value,’ in Moore’s Philosophical
Studies (Patterson: Littlefield, Adams, 1959), 265 (originally published by Routledge & Kegan Paul in
1922).
¹⁷ Indeed, if the consequentiality of goodness is a priori and necessary, a thing absolutely cannot be just
good, rather than good by virtue of being something else, whereas it is not a priori that a thing can’t be (say)
Intrinsic Value and Reasons for Action
91
use of ‘good’ in question. The ascription ‘She is good’ implies goodness as a
person in one context, goodness as an athlete in another; ‘Beautiful paintings
are among the good things in life’ implies that viewing them is (intrinsically)
good; and so forth. When we know what kind of thing is being called good and
what kind of goodness is attributed to it, we can begin to figure out what sorts
of experiences of it are intrinsically good and what sorts of qualities it must have
to yield those experiences.
If, however, the bearers of intrinsic value are experiences, don’t we need to
make sense of the locutions ‘a good experience’ and ‘a bad experience’ even apart
from any context? Not entirely apart from any context, for there are at least two
commonly accepted constraints on a theory of intrinsic goodness—and perhaps
on any normative discourse embodying notions of intrinsic goodness—that can
help us here. First, it must be clear how what is properly thought good in itself
(whether intrinsically or inherently) can contribute to making life choiceworthy. Second, and related to this, it must be clear that what is intrinsically good
can provide reasons for action. This is not to say that the goodness itself constitutes a reason; what constitutes a value-based reason for action is typically, and
perhaps will always be, something connecting the action with that in virtue of
which the relevant experience is good (or bad), say the fact that doing the thing
in question will be enjoyable.¹⁸ I take a choiceworthy life to be (roughly) one
constituted by a favorable balance of good experiences relative to bad ones.
Moreover, I take it to be a non-contingent and probably a priori truth that the
fact that an experience would be intrinsically good provides a reason—some
reason, at least—to seek it (in a very wide sense that can include trying to bring
it about for someone else) and that the fact that an experience would be intrinsically bad provides a reason for avoiding it. More specifically, the kinds of qualities of experiences in virtue of which they are good or bad experiences—most
yellow but not in virtue of anything else true of it (this does not hold for all descriptive properties, such as
being colored, since it is a priori that a thing cannot have this property except in virtue of having a specific
color property). I am not sure whether Moore noticed this contrast.
¹⁸ I am again allowing that an intrinsic good might ‘provide’ a reason without constituting one. Suppose,
for example, that we adopt T. M. Scanlon’s ‘buck-passing’ view of goodness and value (hence presumably of
basic reasons for action as well), on which reasons are constituted by the specific things in virtue of which
something is good—say, being enjoyable—not by its goodness as such. See What We Owe to Each Other
(Cambridge, Mass.: Harvard University Press, 1998), esp. 95–100. We can still speak, however, of intrinsically good experiences as providing the reasons, though in virtue of their grounds, such as pleasurable qualities of experience. We could also distinguish between elements’ directly and indirectly constituting reasons
and between specific and general reasons. I might do a thing because it is good even if I think it is good only
because of some particular property of it. Suppose, moreover, that I do it only because, entirely on your testimonial authority, I believe it will be good. It would appear that even if I would pass the buck to you given
an enquiry into the status or basis of my reason, my generic reason—to do something good—is where the
buck stops for me. To be sure, it stops there on your authority, but that does not make my reason doing what
you suggest.
92
Robert Audi
obviously being enjoyable or painful—constitute grounds for desire and
thereby for action. Is there, then, reason for the sadist who would enjoy torturing a terrorist to death to do so? If we keep in mind that there can be a reason
that is defeated by the presence of better reasons for something else, I believe we
should answer affirmatively. How such a question is to be decided will be
addressed in section 6.
Countenancing things that are inherently good but not intrinsically good
can be seen to cohere with this partial account of intrinsic goodness. There is,
for instance, non-instrumental reason to preserve a beautiful painting because
it is a potential constituent in an intrinsically good aesthetic experience of it.
The account is also sufficiently pluralistic to accommodate the variety of inherent and intrinsic goods we should countenance. For each inherent good, we
can imagine intrinsically good experiences in which it is a constituent; and each
such experience grounds reasons for positive action regarding the inherently
good thing.
To be sure, if, in requiring a specification of the kind of object said to be
good—the specification of a value-anchoring kind—as a condition of the
intelligibility of predicating intrinsic value, one seeks to reduce the predication
to a ‘descriptive’ statement, then my view may not be accommodating. A good
knife may be simply one that cuts certain things easily and retains its edge
through multiple uses, but it is not at all clear that a good person is simply (say)
one who performs certain tasks efficiently and tries to help, and never to harm,
others (this gains some plausibility, however, if we take ‘help’ and ‘harm’ to be
richly normative terms, as where helping people is understood to imply
contributing to their moral well-being). Perhaps ‘good knife’ admits of a
descriptive, if vague, analysis; but ‘good person’ surely does not. If we distinguish between reductive and identificational uses of phrases in which ‘good’ is
anchored by a specification of the relevant kind, then, for ascriptions of
intrinsic and inherent goodness, my account welcomes the latter uses but
makes no commitment to the former.
If predications of goodness are intelligible only relative to some valueanchoring kind, are they also implicitly comparative, and might they be
analyzable in terms of comparative, perhaps even ‘descriptive,’ statements?
Granted, knowledge of intrinsic goodness may require comparing the thing in
question with other things of the relevant kind, or at least a readiness to make
such comparisons. But it does not follow, and I think is not true, that the
attribution of intrinsic goodness is equivalent to a comparison of the thing
with others or, especially, to a ranking of it in relation to them on the basis of
some descriptively specifiable property such as efficiency in producing a
certain effect. One can know an experience of a beautiful painting to be
Intrinsic Value and Reasons for Action
93
intrinsically good without being committed to this experience being, say, at
least as enjoyable as most of one’s pleasurable experiences of viewing paintings.
There is another kind of incompleteness we must consider: ‘evaluative incompleteness.’¹⁹ Consider someone’s experiencing pleasure. How should we evaluate
such an experience? If, as hedonists and others hold, pleasure is intrinsically good,
it would seem to follow that we must call this experience intrinsically good; but,
as we saw, the pleasure may have an object which pleasure ill befits.
It is useful here to view intrinsic goodness on the model of prima facie duty
as understood by Ross. Consider promising. If all I know is that you promised
to do something, I am not entitled to infer that you ought (in the sense of final
obligation) to do it. But surely I can know that you have a prima facie obligation, in a sense that implies final duty if no set of other duties overrides this
one.²⁰ In this sense, we might say, promising is intrinsically obligating.
Similarly, suppose I can tell that you are taking pleasure in something, but I do
not know what it is. May I not infer that there is something intrinsically good
(good-making) in your experience, indeed something such that, if nothing in
or regarding your experience defeats this good, the experience is, in an overall
way, intrinsically good? I think so; and if this is so, then just as promising is
intrinsically, though not necessarily finally, morally obligating, pleasure is
intrinsically normative—since it is good-making—and, where its value is not
overbalanced by some other element(s) in or regarding the experience to which
it belongs, renders that experience intrinsically good overall. Moreover,
pleasure, like moral obligation, is ‘practical’ in implying a reason for action.
We now face another question. Why should we not say that pleasure is only
prima facie good? Let me make two points about this.
First, we may say that pleasure is prima facie good, so long as we take prima
facie goodness to be a kind of intrinsic goodness: the pleasure cannot fail to be
good qua pleasure, even though the whole of which it is an aspect can fail to be
intrinsically good, just as an action cannot fail to be obligatory qua promised
even if other facts about it prevent it from being obligatory on the whole. This
kind of defeasibility does not imply a relational status; that pleasure does not
necessarily make the experience it characterizes intrinsically good overall does
¹⁹ See Michael Zimmerman, ‘Evaluatively Incomplete States of Affairs,’ Philosophical Studies, 43 (1983):
211–24, and The Nature of Intrinsic Value. He says in the latter, regarding the state of affairs, John’s being
pleased at time t, ‘Is there any reason to think that contemplation of a state of this form as such requires that
you favor it? . . . If it occurs in virtue of a state of the form [John’s being pleased at Mary’s pain] . . . What
reason is there to think that anything good has happened?’ (145) Here intrinsic goodness is ‘defeated.’
²⁰ I have argued for this in ‘Moderate Intuitionism and the Epistemology of Moral Judgment,’ Ethical
Theory and Moral Practice, 1 (1998): 14–34. For a treatment of Ross’s analogy between goodness and prima
facie duty, see Lemos, ‘Organic Unities’ (I take Lemos’s defense of intrinsic value here to be at least largely
compatible with my response to the analogy).
94
Robert Audi
not imply that the pleasure itself is good only in relation to something else, or
only when certain conditions are met.
One way to see this point about pleasure is to note that we recognize, as a
basic reason for performing an action—even if a defeated reason—that one
will enjoy it. Another way is to note why it is that we do not look favorably on
pleasure on the part of a malicious person who in no way merits it (here I vary
a well-known example Kant used in the Groundwork). The best explanation of
this is that we take the pleasure to be a good the person should not have.
The explanation of this, in turn, is not, or not only, that the pleasure itself is
inherently bad; we might think the person should not have it even if we believe
that the pleasure is not the kind that ill befits its object. Virtually regardless of
the object, we take the pleasure to be a good that the person should not have.²¹
More specifically, the state of affairs, that malicious person’s having it, is (as
indicated earlier) virtually always inherently bad. Pleasure can ill befit its possessor, just as it can ill befit its content.²²
The second point is that the experientialist theory of value I am proposing is
not as such committed to taking any particular kind of experience to be
intrinsically good overall. It may be combined with the view that the overall
intrinsic goodness of an experience cannot be determined until its relevant
content, particularly its phenomenal and intentional properties, is taken into
account. Overall intrinsic value would then be what Ross calls a ‘toti-resultant
attribute.’²³
A complication here is that individuation of experiences is no easy matter. If
I read a beautiful poem while hearing crude, ugly music, am I having a single
mixed experience or two experiences, a good literary one and a bad auditory
²¹ I say ‘virtually’ rather than always to allow for the possibility that such people of bad character might
take pleasure in the thought of doing things to make amends for their past wrongs and to improve their character. But if we are happy to see them have this kind of pleasure, it seems in part because we take it to show
at least that they have enough decency to look with pleasure on what morality calls for. Cf. Moore’s view that
‘When I talk of a thing as “my own good” all that I can mean is that something which will be exclusively
mine . . . is also good absolutely. The good of it can in no sense be “private” . . . The only reason I can have for
aiming at “my own good” is that it is good absolutely that what I so call should belong to me . . . which, if I have
it, others cannot have’ (Principia, 99). On the surface, this precludes Moore’s holding that an experience of
pleasure might be intrinsically good, yet I should not have it. But perhaps he could here invoke the principle
of organic unity and claim that enjoying causing someone pain would not normally constitute my own good.
In any case, my theory allows us to say both that sadistic pleasure may be intrinsically good, qua pleasure, but
not good, considered overall, and that it may in any case be inherently bad. The latter point alone might
imply that there is adequate reason to oppose it.
²² In ‘Organic Unities’ Lemos uses this kind of example to show that evaluative incompleteness does not
undermine the view that intrinsic value (construed as encompassing inherent value) is organic. He notes that
(on highly plausible assumptions) S’s being wicked, deserving no pleasure, and having none is, in itself, better than S’s being wicked, deserving no pleasure, and having it, even apart from the content of the pleasure.
See pp. 330–1.
²³ Ross, The Right and the Good, 28.
Intrinsic Value and Reasons for Action
95
one? There is no simple way to determine plausible criteria here, but that is not
crippling. We can say that there is one experience with good and bad qualities
and seek to determine its overall value, or we can countenance two simultaneous
experiences, one good and one bad. On the latter alternative, we can ask
whether the person’s experience is overall (intrinsically) good or bad. This is not
to posit additivity of intrinsic value and abandon organicity but to recognize a
kind of combinability of the separate intrinsic values of experiences over
moments of life. Such combination is something we must reckon with in any
case to assess how good a life is.²⁴
One further point is appropriate here. My positive intuitive notion of a good
experience has been partially explicated by appeal to the concept of the kind of
experience that makes life choiceworthy. Is my account circular because I
define ‘intrinsically good’ in terms of, in effect, ‘intrinsically good life’? I am
not giving a definition or analysis of ‘good life’ in any sense, as opposed to
an explication (and I am willing to leave open whether Moore was right in
thinking that there is a significant sense of ‘indefinable’ in which ‘good’ is indefinable). But one point to note is that if only experiences are intrinsically good,
then one’s life, not being an experience, cannot be intrinsically good, as opposed
to inherently good. Still, this is only because a succession of experiences is not
necessarily an experience. It is, however, in a broad sense experiential. We need
a term for a good life or indeed for any desirable whole made up of intrinsically
good experiences, or of a kind and proportion of them such that it is overall
desirable to realize that whole. ‘Choiceworthy’ might do for lives, and perhaps
also for sequences of experiences, say those constituting an evening. But it must be
understood in terms of the things we intuitively take to be good, such as enjoyable
social and aesthetic experiences. Another possibility is the term ‘compositely good’
(or ‘compositionally good’); not all goods are such, and basic ones cannot be.²⁵
It may help to draw an analogy to epistemology. The general idea is to
explicate intrinsic goodness intuitively, in terms of experiences, as we do
grounds of justification, and then to provide a way to use the notion to guide
choice as we use the notion of justification to guide the formation and criticism
of belief. Certain social, hedonic, and aesthetic qualities of experiences are
the sorts of things for which we (rationally) choose to have experiences; the
choiceworthy experiences are the kinds of thing for which we ultimately
choose lives themselves.
²⁴ For a defense of the additivity of intrinsic value (once we have ‘basic’ intrinsic values to work with), see
Zimmerman, The Nature of Intrinsic Value, esp. ch. 5.
²⁵ For support of the idea that we must rely on intuitive starting points in the theory of value and discussion of how overall value is to be understood in the light of those he chooses for his base clauses in a recursive
value theory, see Thomas Hurka, ‘Two Kinds of Organic Unity,’ Journal of Ethics, 2 (1998).
96
Robert Audi
5. The ontology of intrinsic value
Nothing I have said entails that intrinsic or inherent goodness is a non-natural
property. If, however, they are consequential in the way I have suggested, then
they are indeed properties, and thus a kind of realism is implied. This realism is
of a special sort, however; for experiences of the kind that have intrinsic value
are mind-dependent, or at least life-dependent. Hence there cannot be anything
of intrinsic value in a world in which there are no persons—or minds or at least
living things capable of experience.
By contrast, an inherently valuable thing can exist in an otherwise empty
world. But there is a sense in which such a thing is axiologically unfulfilled. It
is, to be sure, good ‘in itself,’ as non-relationally and non-instrumentally
good.²⁶ Inherent goodness is a significant status, but inherently good things do
not directly provide reasons for action nor, simply as unexperienced inherent
goods, contribute to the experiential goodness of any life. This is not to deny
that a life in which one’s friends flourish is better than a life experientially just
like it in which one has the same evidence that they do, but is wrong. The
former life is even better in itself—inherently. We could also call the former
more choiceworthy if we are thinking of choosing in the light of knowing the
important facts about each, including the error in one of them. Its inherent
superiority makes it preferable in the contemplation of it; its entailing the
realization of additional intrinsic value—which it does since the friends’ flourishing entails that realization—provides an other-regarding reason to choose it
over the alternative.²⁷ But for the person in question, neither life is better in the
living of it. We can and indeed should insist that the point of view of the subject
living a life is not the only one from which to make assessments of its goodness;
but this does not require treating all non-instrumental goodness as intrinsic or
taking the inherent to be as basic as the intrinsic.
What about Moore’s idea that (the property of being) good is simple?
Butchvarov has suggested that it is a generic property, presumably a determinable
²⁶ Arguably, we may say non-relationally hence non-instrumentally good. But suppose one makes a device
for capturing emissions that do not and may never exist. Could we not say it is a good pollution control device,
where the goodness is plainly instrumental? We might reasonably rule that this is potential instrumental goodness and that actual instrumental goodness requires the existence (past, present, or future) of something to
which the instrumentally good thing is a means. I cannot here explore the merits of the main alternatives on
this matter; the contrast between inherent and instrumental value that I make use of is that the former, but not
the latter, is strictly (presumably even conceptually) necessary for that to which it is a means.
²⁷ Suppose we know that creating something of a kind that has inherently good instances could not lead
to anyone’s having a valuable experience of it. Perhaps a painting could be created such that the attempted
viewing of it immediately destroys it (nor does it have any other way of figuring in a valuable experience).
Does it (as opposed to, for instance, the intrigue of producing it) provide any reason to create it? (If the
impossibility is logical, of course, the thing in question is not inherently good.)
Intrinsic Value and Reasons for Action
97
such as shape.²⁸ There are indefinitely many shapes, as there are colors. Are
these determinable properties simple? There is a sense in which they may be:
one does not define them (or the determinates under them) by genus and
difference. Something at least highly analogous seems to hold for the property
of being a good experience.
Shapes and colors, however, are apparently natural properties; why not,
then, conceive value properties as natural? Moore said one thing on this score
that impressed C. D. Broad and needs examination.²⁹ It is that there is a sense
in which what we intuitively call natural properties are descriptive, and goodness is not. Insofar as ‘fully’ describing anything makes sense, can we not fully
describe something that is intrinsically good without saying whether it is
intrinsically good? This would seem so if, from our description, as with a
description of a painting, it does not ‘analytically’ follow (as Moore could
plausibly claim it does not) that it is good or that it is not good. For in that case
we cannot be plausibly thought to have ‘implicitly’ described it as good or as
not good. That will remain, in one sense, an ‘open question.’
To look at the matter from the other direction, if I tell you that a painting
(for instance) is good, have I even partially described it? One plausible reaction
is that I have not; but another, perhaps as natural, is ‘You’ve described it as
good, but now tell me more about it.’ Still another is ‘I know what sorts of
things characterize a good painting; which do you have in mind?’
Pushing this line further, we might get something it is natural to call normative
disjunctivism, the view that an attribution of a normative property is equivalent
to an ascription of a disjunction of relevant grounding properties. If we
suppose that for each normative property, there are certain natural properties
on which, as a matter of conceptual truth, it is consequential, this view gains
some support. The supposition also provides for a contrast with natural properties, since, so far as the notion of a natural property is clear, the counterpart
point apparently does not hold for them. But must there be a definite list of
base properties for each normative one? And, if not, do we, for any context of
normative attribution, have a way to determine what base properties are
implicitly ascribed? We need at least a theory of normative properties and a
theory of kinds of description, and apart from such theories it is not clear how
²⁸ For a detailed account of goodness along these lines see Panayot Butchvarov, Skepticism in Ethics
(Bloomington: Indiana University Press, 1986), esp. 63–6.
²⁹ Explicating Moore, Broad said, ‘a description of a thing can be complete even if it omits those characteristics of it which, though determined solely by its intrinsic nature, are not themselves intrinsic. E.g., a
pleasant experience, which is also good, could not be completely described unless its pleasantness was mentioned. But it could be completely described without its goodness being mentioned. I find it most difficult
to follow or accept this.’ But he does admit that goodness is a ‘derivative characteristic’ (‘Moore’s Ethical
Doctrines,’ in Schilpp (ed.), The Philosophy of Moore, 60).
98
Robert Audi
much weight to give to Moore’s initially plausible claim that goodness is a
non-natural property.
There is something further we should note, something perhaps more controversial. It is not clear that the properties of being intrinsically or inherently
valuable or, for that matter, any broadly normative properties, have causal
power. Their non-normative base properties do. If this is correct—if, for
instance, injustice’s causing a revolt is a matter of the elements underlying the
injustice doing the causal work, then the point may help to explain why
normative properties are not plausibly considered natural or, perhaps, even
descriptive.³⁰ (One might also take it to imply that they are not really properties at all.) But, one might reply, does shape, as such, have causal power, as
opposed to the circularity, say, of a wheel’s having it? One answer is that a
circular thing may have causal power qua circular, where being circular is a
shape property. Does injustice, for instance, have causal power qua normative
property, say as a moral evil? The reality of moral intrinsic value, like that of
other kinds of intrinsic value, does not self-evidently entail that as intrinsically
valuable, such things have causal power.
If intrinsic value properties (of which the most generic are being intrinsically
good, being intrinsically bad, and being intrinsically neutral) are natural
properties, they are then certainly real. But surely they can be real without
being natural, and they can perhaps be natural without being causal. I leave
open whether they are causal, but I am taking them to be real properties.
Are they, however, as Moore and others have thought, essential to their
possessors?³¹
If intrinsic value properties belong to experiences, and if the (intrinsic) value
properties of experiences are (necessarily) consequential on their phenomenal
properties, then there is reason to take the former to be essential. For the phenomenal properties of an experience are plausibly considered essential to it.
What about thought properties, such as silently reciting a certain Shakespearian
sonnet? If these are not phenomenal, they in any event seem essential to the
experiences that have them. But we again find difficulty regarding criteria of
individuation. An experience of a painting in which I am conscious of a yellow
patch in that painting is not of the same experience-type as one in which I am
conscious of red in the same spot. Asked which color I like better, I might
respond by saying I can answer only on the basis of experiences of each.
Suppose, however, that a painting is designed to have that spot go from yellow
³⁰ In ‘Ethical Naturalism and the Explanatory Power of Moral Properties’ (in Moral Knowledge) I have
explored in detail the question whether moral properties are causal.
³¹ For discussion of the apparent essentiality of intrinsic value see Zimmerman, The Nature of Intrinsic
Value, e.g. 47–9; and Lemos, Intrinsic Value, esp. ch. 3.
Intrinsic Value and Reasons for Action
99
to red with each viewing. Then one might be inclined to speak of a single,
changing experience of the painting rather than of two experiences differing in
respect of the difference between yellow and red.
We might, then, speak of a qualitative, fine-grained mode of individuation
and of a coarse-grained, temporal unification mode, the kind illustrated by the
second case. In the second case, pragmatic factors may also enter in. If, for
instance, we are concerned with viewing paintings, then even two viewings in
rapid succession with only a movement of the head between them will yield
two experiences; if we are concerned with how a person likes the experience of
entering a newly decorated room, then a view of paintings, furniture, and rugs
may count as a single experience.
In replying to his critics Moore said something quite pertinent to this question: that there is a sense in which pleasure is not intrinsic to an experience. For
you and I can have the same experience of the taste of caviar and one of us like
it, the other not. He does not, however, take up criteria of individuation at this
point. Presumably pleasure is not intrinsic to any experience non-hedonically,
say gustatorily, characterized, since any such experience could be enjoyed
or not; but on a qualitative standard of individuation, pleasure is surely an
intrinsic experiential property.³² Indeed, he goes on to say that if we speak of an
experience of being pleased by the taste of caviar, then the pleasure in question
is intrinsic. Calling an experience one of being pleased is giving it a hedonic
characterization. (Perhaps he is thinking of experience-types. Doubtless pleasure is not intrinsic to an experience-type unless it is essential to it, and pleasure
is not intrinsic or essential to the type, tasting caviar. But Broad, to whom
Moore was responding here, was presumably thinking of experience-tokens, as
I am when I speak of experiences as the bearers of intrinsic value.)
A further point is that, plainly, to call an experience one of the taste of caviar
is not to give a phenomenally ‘complete’ description (whether we speak of types
or of tokens). Perhaps we can say that if a description of an experience (token)
is phenomenally complete, it will include any hedonic properties the experience has, and if one of them is pleasure, then the token necessarily has that
property, presumably in virtue of (another set of ) its intrinsic properties. In
that case, if a pleasurable experience is necessarily good, then its goodness is
also essential to it.
Moore must have seen something we need not deny: we can specify a type of
experience in a coarse way by specifying what its object is; and when we do this,
³² This would be denied by a proponent of the conative (intrinsic desire) theory of pleasure suggested by
Richard B. Brandt. I have noted serious difficulties for that theory in ‘Prospects for a Naturalization of
Practical Reason: Humean Instrumentalism and the Normative Authority of Desire,’ International Journal
of Philosophical Studies, 10 (2002).
100
Robert Audi
we do not foreclose whether it is pleasant or not, since an experience even of,
say, a backrub may not be pleasant. But this is quite compatible with saying
that when an experience is pleasant, its pleasantness is an intrinsic property of
it. For calling the experience one of a backrub is a coarse specification in terms
of a single object (note that one can have an experience that is of both the taste
of caviar and the sound of crude music, but we may refer only to the former if
that is the focal object or is salient), whereas calling an experience pleasant is
(in one way) a more fine-grained—and is a phenomenally more specific—
characterization.
Inherent value properties might seem to admit of the same treatment as their
intrinsic value counterparts. But there is at least one important difference.
An inherently valuable thing can retain its identity across changes that alter its
inherent value. A beautiful painting can be damaged without ceasing to be the
painting it is; this is clear, at least, where it is fully restorable. If, however, we
specify inherent value given the exact (non-relational) condition of the thing in
question, then there is the same kind of reason for saying that the inherent value
properties of a thing are essential to it as for saying that intrinsic value properties of a thing are. I do not, however, claim that it is clear that we must consider
either kind of property to be essential.
6. The epistemology of intrinsic value
I will be brief in discussing the question of how we know propositions about
intrinsic value. This is in part because I have addressed the topic in some detail
elsewhere³³ and in part because my most important points are neutral with
respect to the various plausible positions on the epistemology of value.
In Principia Moore made the sweeping claim that propositions about the
good are all synthetic, and he clearly took some of them to be a priori as well.³⁴
Here again the concept of goodness as a determinable is a useful focus. There is
some reason to think that the domain of determinables and their determinates
is one where synthetic a priori truths occur. But do they? I doubt that anyone
has shown that the proposition that nothing is red and green all over at once is
either analytic or empirical, and the proposition that if something is square,
then it is not round seems quite similar on this score.³⁵ Do we, however, know
a priori that, for instance, pleasure is (intrinsically) good and that burning
³³ In e.g. ch. 11 of Moral Knowledge and in ‘Moderate Intuitionism.’
³⁴ He said, for example, that ‘propositions about the good are all of them synthetic and never analytic’ (7).
³⁵ This position is defended in ch. 4 of my Epistemology, 2nd edn. (London: Routledge, 2003).
Intrinsic Value and Reasons for Action
101
people to death is morally bad? We need a lot of experience to be in a position
to know these things; the background knowledge needed, involving such
notions as that of the intrinsically good and the morally bad, requires considerable conceptual sophistication. But arguably the required experience is only
that needed to acquire the concepts and is not evidential. Could we know
general propositions about value, say that pain is intrinsically bad, in the way
we know empirical truths, such as laws of nature? I doubt this, largely on
account of the different kinds of grounds appropriate to each; but I will not
pursue the matter here.
There is another way to approach the epistemological issue. Suppose it is not
only necessary but a priori that, as Moore held, between ‘the propositional
function “x is intrinsically good” . . . and the function “the fact that an action
which you can do would produce x is some reason for supposing that you ought
to do that action” . . . there is a two-way necessary connection.’ Then, if we
know a priori that producing pleasure is a basic reason for action, we may
conclude by a priori reasoning that it is intrinsically good. This does not
presuppose that the notion of a reason for action is more basic than that of the
intrinsically good, though it is compatible with that view.³⁶ Indeed, we can
treat both goodness and reasonhood as common consequences of the same
supervenience bases, including pleasure and pain. For Moore, however,
reasonhood implies intrinsic goodness; and unless he countenances moral
goodness, he could not (as I would) take the fact that an act would be, say, a
breaking of a promise or a deprivation of liberty as a basic reason for action.
Suppose, on the other hand, that basic axiological knowledge is empirical. If
value properties are causal, or indeed natural in any sense, this would seem
possible. Might one not still have a kind of intuitive knowledge of such
propositions as that enjoying viewing a beautiful painting is intrinsically good?
And might one not still be able to have non-inferential knowledge of some
propositions attributing intrinsic value to experiences? I cannot see why these
possibilities must be ruled out. It is not as if only a priori knowledge could
be intuitive and non-inferential.
Moreover, even apart from whether principles of practical reason can be
known non-inferentially, the most important point here is that much the same
ones may be retained as on a rationalist epistemology. I refer to principles
indicating that there is reason to avoid pain, to seek pleasure, to weigh future
³⁶ This view is suggested by some (but only some) of what Thomas Nagel says in The View from Nowhere
(Oxford: Oxford University Press, 1986). See e.g. ch. 8, sections 5 and 6. A position further in the direction
of taking rationality to be more basic than value is Elizabeth Anderson’s view that ‘to be intrinsically valuable is to
be the immediate object of such a rational attitude’ (she refers to ‘love, respect, consideration, affection, honor
and so forth’). See Value in Ethics and Economics (Cambridge, Mass.: Harvard University Press, 1993), 20.
102
Robert Audi
goods on a par with present ones other things equal, to take account of
probabilities in deliberation, and so forth. Thus, although I favor a moderate
rationalist conception of such principles, I do not think that the kind of theory
of intrinsic value I am developing requires it in order to sustain highly similar
principles for guiding and appraising conduct. Those principles could also be
retained if, instead of distinguishing intrinsic from inherent value, we took
inherent value to be a kind of intrinsic value. Neither my preferred epistemology nor my proposed taxonomy interferes with the selection and defense of
plausible normative principles.
7. The intrinsically valuable as a basis of reasons for action
Suppose that there is a close connection between intrinsic value and reasons for
action. More must be said to indicate how my account of intrinsic value may
contribute to the general theory of practical reason. Let me first bring out the
advantage of conceiving the bearers of intrinsic value as experiences rather than
concrete objects or even such dispositional elements as knowledge and virtue.
I assume that reasons for action are always expressible in (though not only
in) the kind of infinitive clause that indicates the content of an intention or
follows ‘in order to’ in an explanation of action, as in ‘I agreed to read the book
in order to satisfy a long-standing curiosity.’³⁷ The contents of these clauses are
apparently types of events, especially act-types, all of which are things that can
be realized (and in that sense brought about). More important, if their being
realized (tokened) in a certain way has intrinsic value, then—as indicated
earlier—there is some reason for the agent to act to realize them, though the
ascription of intrinsic value does not specify just what it is that has that value.
Now experiences, like actions, admit of a type–token distinction, and to bring
either one about is to token it. Might we not say, then, that experiences of performing actions are among the bearers of intrinsic value? Surely the experience
of conversing, of singing, of swimming, and of many other actions can be
intrinsically good. These often prominently include the experience of producing
something inherently good, such as a fine poem or a beautiful sculpture.
To see how this idea is best understood, consider what it is to do something
‘for its own sake.’ This is to do it for qualities of performing it that attract us on
their own account. These are always experienceable qualities; and there is much
plausibility in taking the experience we have in acting, particularly in respect of
³⁷ I argue for this view in ch. 5 of The Architecture of Reason, where I also suggest some contrasts between
the propositional and infinitival forms of expression of reasons for action.
Intrinsic Value and Reasons for Action
103
these qualities, to be the bearer of any intrinsic good we seek in performing the
action. If I want to satisfy my curiosity for its own sake, I want to experience the
acquisition of the relevant knowledge, not merely to acquire it. To have a
machine implant the relevant knowledge in me by brain manipulation while
I sleep would not fulfill the particular desire I have. It is the experience of
gaining the knowledge that I specially want, and if that experience is rewarding
in a certain non-instrumental way, it will be intrinsically good.
In part, the general idea is this. If there is anything intrinsically good (or
intrinsically bad), it provides a basic (if defeasible) reason for action to bring it
about (or continue its existence, which is roughly equivalent to bringing about
a later stage of the thing in question). Intrinsically good things should thus be
the kinds of things that can be realized in the way the act-types (or at least
experience-types), which are (or are specified by) the contents of reasons for
action, can be realized. Experiences can be realized in that way; paintings, as
opposed to the actions of producing and viewing them, cannot. If intrinsic
goodness belongs to certain kinds of experiences, it can accordingly figure in
the content of reasons for action, and intrinsically good experiences will figure
centrally in the content of the most basic kind of reasons for action: to enjoy a
symphony, for instance, or to swim in cool waters, to talk with a loved one, or
reduce the pain of a headache.
The suggested view of reasons is connected with the idea that the intrinsically
good is that in virtue of which a life is choiceworthy. Surely a life is choiceworthy on the basis of the experiences that—as I would put it—constitute that life.
Knowledge and virtue are, to be sure, good ‘in themselves’—where that means
inherently good. But would they make a life good apart from experience? In
principle, a person could be created with a great deal of both, but in a deep
sleep; if such a person ceased to exist before experiencing anything, would the
person’s life be good? Inherent goods provide non-instrumental reasons for
action because of their constitutive place in intrinsically good experiences. This
gives them incalculable normative importance. But their existence in itself, as
opposed to the experiences we bring about in which they essentially figure,
does not make anyone’s life, as lived, better.
Might the only reasons for action be grounded in considerations of intrinsic
value, for instance of pleasure and pain? Suppose this is so. It would not follow
that the notion of intrinsic value, as opposed to that of something which has it,
must come into the concept of a reason for action. Still, not only instrumental
reasons, but also non-instrumental reasons deriving from inherently good
things, might ultimately depend on a relation to what is intrinsically good or
intrinsically bad. We can say this without treating the concept of value as a constituent in that of a reason for action. I do not want to presuppose, however,
104
Robert Audi
that all reasons for action are grounded in considerations of value. I leave open,
for instance, that a deontological reason such as that a deed would be dishonest
can provide a reason independently of intrinsic value (at least of non-moral
intrinsic value).
I am supposing, however, that some experiences are moral and have moral
intrinsic value. Moral reasons, then, need not derive from non-moral grounds
even if they do ultimately depend on considerations of intrinsic value (including
disvalue), and they can be a kind of basic reason for action. I have mentioned
indignation; an experience of this emotion can be intrinsically good qua moral,
as well as inherently good on the basis of its overall moral character, yet
intrinsically bad hedonically. The experience of being done an injustice can be
not only painful, and in that way intrinsically bad, but also (intrinsically)
morally bad; and in my view even the felt attraction to doing an injustice is an
intrinsically morally bad experience.³⁸
To be sure, to say that some basic reasons for action derive from considerations of moral value leaves open whether all moral reasons for action so derive.
I have said nothing implying that the good is more fundamental than the right,
as it in some sense would be if all moral reasons derive from considerations of
intrinsic moral value. But my view certainly makes that position intelligible
and indicates some of the questions that must be pursued in appraising it.
One further matter needs attention to fill out my sketch of the connection
between intrinsic value and reasons for action. Is there, in the domain of such
reasons, an analogue of the organicity of intrinsic value? Recall the case of
Schadenfreude. If I enjoy teasing someone, I thereby have a (normative) reason
to do it; but if I am pained by (the experience of ) causing someone else to suffer
and believe that the person in question will suffer from my teasing, I thereby
have a (twofold) reason not to do it. Now suppose that my pleasure will (quantitatively and qualitatively) outweigh her suffering. Do I thereby have better
reason to tease her than not to? Not if, as is quite rational, I believe that pleasure
taken in something that one knows causes pain as this does is unbefitting to its
object in a way that renders the pleasure inherently bad. Experiencing such
pleasure even in prospect is intrinsically aversive, say repugnant; this kind of
experience provides reason to avoid realizing that prospect. Indeed, I do not
have to believe anything this theoretical in order to have non-instrumental
reason to avoid the teasing. As I have described the case, then, I apparently have
better reason not to tease. Suppose this is so. Is it something I know by
summation of the values in question? I think not.
³⁸ It is of course not self-evident that experience (or anything else) can have moral intrinsic value; that it
can is argued in my ‘The Axiology of Moral Experience,’ Journal of Ethics, 2 (1998).
Intrinsic Value and Reasons for Action
105
Indeed, summation, so far as it is possible in a rough-and-ready way, would
not work to give me a decision even apart from considerations of fittingness.
We must use practical judgment to decide how to weight the good against the
bad or, more generally, sets of positive against sets of negative intrinsic values.
This is something Ross insisted on in connection with conflicts of prima facie
duties. It applies to what we have overall reason to do, as well as to ascertaining
what is intrinsically good overall.
If basic moral reasons for action depend entirely on things of intrinsic value,
then the kind of organic composition of prima facie duties Ross saw as yielding
final duty is to be expected. It might in fact be expected even if values play a
lesser role in grounding duties, say because some duties are purely deontological.
For instance, if there are duties of beneficence and also of self-improvement,
each set concerns promotion of good things and elimination of bad ones in the
lives of others or in one’s own life. The best overall result will often be determinable only by taking into account the organicity of the values in question.
Reading all afternoon might best promote my knowledge; helping a friend
might best reduce his suffering; and there may be many other considerations
pertinent to my decision regarding the same period of time.
Conclusion
The theory of intrinsic value sketched here is experientialist, pluralistic, internalist but not subjectivistic, realistic about values, and, in the way the most
plausible kinds of intuitionism are, moderately rationalist about knowledge of
certain propositions about value. It takes what is intrinsically good to provide
basic reasons for action, but does not imply that every basic reason for action is
grounded in intrinsic value. The theory is consonant with the idea that goodness—whether intrinsic or inherent—is unanalyzable, but it is not committed
to that view. It is also consonant with the idea that value properties are not natural properties—so far as that idea is clear—but is not committed to that
either. It is committed to the view that value properties are consequential on
non-value properties. If these are properly characterized as natural or, as seems
plausible, ‘descriptive,’ value properties can be consequential on them whether
or not they themselves are natural. Although a property cannot be identical
with the set of properties on which it is consequential, it may or may not be of
the same ontological kind.
How Moorean does all this make the position? This depends in part on how
much significance attaches to distinguishing between intrinsic and inherent
value and attributing, to many things Moore considered intrinsically valuable,
106
Robert Audi
inherent value instead. It also depends on how much significance attaches to
avoiding Moore’s commitment to non-natural properties. Another difference
is that moral value is specifically posited by the theory I have introduced, but
not by Moore’s;³⁹ nor does Moore’s view provide for basic deontological
reasons for action. I have also broadened the principle of organic unities both
in considering aspects of the valuable as well as parts of it and in applying the
principle to the inherently as well as the intrinsically valuable.
Still another difference between my view and Moore’s concerns the connection between the right and the good. He holds a kind of maximizing consequentialism. No such position follows from the theory of value I have
sketched here (and I do not endorse it). It may indeed not follow from Moore’s
theory of value either, but the idea that the right is, in a quasi-quantitative way,
derivable from the good is a major strain in his ethical thought.⁴⁰
There is a great deal more that must be said to appraise the theory I have
presented. The sketches of the relevant notion of experience and of criteria for
individuating experiences need extension; the revised principle of organic
unities and its application need further explication; and the nature of value
properties and of their grounding in what seem to be natural properties is an
unending philosophical problem. But I believe that at least if a cognitivist view
is sound, then the kind of experientialist, pluralistic axiological intuitionism
presented here is a major option for ethical theory.
³⁹ I do not take this to be obvious. But given Moore’s presentation of his utilitarian view of goodness and
such remarks as that ‘so far as definition goes, to call a thing a virtue is merely to declare that it is a means to
good’ (Principia, 173), this is a plausible reading.
⁴⁰ In e.g. Ethics, Moore says that ‘the total consequences of right actions must always be as good, intrinsically, as any which it was possible for the agent to produce under the circumstances’ (98).
5
Personal Good
Connie S. Rosati
When we survey the variety of lives people lead, we discover a great many ways of
living in which a person’s life can go well. We find a correspondingly diverse array
of goods, things the presence of which can help to make a life go well. A person
can engage in many avocations, such as gardening, running, woodworking, or
music-making. She can undertake various jobs or careers, becoming a doctor
or a lawyer, a preacher or a parent, a sales agent or a soldier. She can live a life
centered on spiritual, intellectual, material, artistic, or social values, a life alone,
in a convent, or in love. But while there are many goods and many ways of
living good lives, it isn’t true of any arbitrary individual that the presence of just
any of these goods or the choice of just any good way of living would make her
life go well. Perhaps she cannot care about or take an interest in certain activities or pursuits; perhaps she lacks the talent or aptitude for others. What is the
exact nature of the relation between a person and certain good things or good
ways of living that constitutes their being good for her?¹
Any attempt to answer this question and explain what it is for something to
be good for a person must come to grips with our dual nature.² We are
I want to thank Sarah Buss, Richard Dees, David Sobel, Mark Timmons, and members of the philosophy department at the University of California, San Diego, for their helpful comments on earlier versions of this paper.
¹ My discussion assumes that a person’s intrinsic, nonmoral good is comprised of things that are themselves good, or at least not bad. I proceed in this way in order to bracket certain difficult questions that could
not be explored adequately in the space of this essay, among them, the question of whether bad or even evil
things can be noninstrumental parts of an individual’s welfare. I suspect that if they cannot be parts of our
welfare, it is not because a correct analysis of personal good precludes them. Rather they are wholly or largely
precluded because of our nature as rational agents and our nature as a particular type of biological creature.
Again, I won’t be able to explore these suspicions here, though I will very briefly address the problem of evil
activity as an apparent counterexample to my analysis. My discussion also makes the simplifying assumption
that a good life is a life in which a person largely obtains things good for her. In fact, though, any more global
assessment of a person’s life will have to attend, among other things, to how ‘narrative relations’ among life
events can affect the value of that life. See Velleman (1991). Velleman has suggested in conversation that the
impact of narrative structure on the value of lives is just an instance of the phenomenon of organic unities.
² Amartya Sen makes a seemingly similar point. See Sen (1985: 185–7, 1987: 41). But he argues for a
duality in the sense that a person’s success isn’t to be judged purely in terms of his well-being, but also in terms
of his agency, since persons may value goals and commitments not a part of their well-being. Sen’s point thus
108
Connie S. Rosati
humans—a certain type of biological creature. As such, we have determinate
properties which, as Peter Railton has stressed, are not a matter of how we
conceive ourselves (1986a: 14–16). Yet we are also agents—creatures with a
capacity for autonomy that enables us, at least to some degree, to engage in selftransformation and, thus, to be as we conceive ourselves. Whether something
is good for us will surely depend on both our humanity and our agency, on our
determinate features and our capacity to determine our futures.
Any attempt to specify what it is for something to be good for a person must
also come to grips with the paradoxical experience of our good. We commonly
encounter our good as a discovery, indeed, a revelation. We uncover our
untapped talent for comedy; we marvel at our sudden interest in the history of
polar exploration; we unexpectedly find a new love in an old friend. Yet as we
plan and live out our lives, our good often seems to be up to us—it seems for us
to decide what to pursue and so for us to decide where our interest shall lie.³
We decide to be parents rather than priests, soldiers rather than salesmen,
understanding full well that we thereby shape our current and future interests.
This duality in how we experience our good no doubt reflects our dual nature.
My aim in this essay is to sketch an analysis of personal good that squares
with our humanity and our agency and with our experience of our good as both
discovered and invented.⁴ I mean for this sketch to be preliminary, and so while
I attempt to make an intuitive case for the view and to show how it can begin
to address certain theoretical challenges to accounts of welfare, I will not
attempt to deal fully with the many questions it will no doubt raise.
I will generally use the expression ‘personal good’ in place of more common
terms, such as ‘flourishing’, ‘welfare’, ‘well-being’, and ‘self-interest’. I do so
because it encapsulates three points that these other terms may not so readily
call to mind.⁵ First, an account of personal good is an account of goodness for
bears on the relative significance of a person’s well-being and his other values, whereas my point concerns
how the duality of our nature affects the proper characterization of well-being itself. For discussion related
to Sen’s, see Sobel (1998).
³ R. M. Hare (1963) attempted to accommodate both freedom and reason in his universal prescriptivism.
But Hare’s notion of freedom differs from the notion of autonomy that I employ, and he means something
different when he talks about being constrained by reasons than I mean when I talk about discovering one’s
good. See Rosati (2003: n. 46).
⁴ This essay offers a first stab at an analysis of good for a person. A fuller treatment of the nature of a person’s good would offer not only a characterization of the good for relation but of what it is for things to be better or worse for a person and of what makes for a good life for a person. I will have little to say in this essay
about alternative theories of personal good. See Rosati (1995a, 1995b, 2000, and manuscript a) for critical
discussion of certain important views.
⁵ Stephen Darwall has also used the expression ‘personal good’. See Darwall (1983: chs. 8–9), where he
appears to endorse a view like Rawls’s. For his more recent ideas, see Darwall (2002). Darwall has indicated
that he finds no great difference among the expressions ‘a person’s interest’, ‘welfare’, ‘benefit’, ‘well-being’,
‘good’, ‘happiness’, or what ‘would make her life go best’. Joseph Raz has distinguished between the notions
Personal Good
109
persons—beings who, as just noted, are not only humans but also autonomous
agents. Second, an adequate theory of goodness for a person must allow that
what is good for one person may not be good for another.⁶ Finally, because we
are persons, our good is personal in the more mundane sense that, at least
under favorable conditions, individual choice and effort figure crucially in how
some things come to be a part of our good.
An account of personal good has implications, of course, for what things
comprise an individual’s good, but it does not itself concern substantive
welfare. Instead, like G. E. Moore’s account of the nonrelational property good,
an analysis of the relational property good for a person concerns a particular
normative feature of things in the world, most notably, of choices, activities,
pursuits, commitments, relationships, and, derivatively, states of affairs. The
goodness at issue is nonmoral, nonaesthetic, and nonperfectionist, although an
individual’s good may include as components moral, aesthetic, and perfectionist
values.⁷ As is well known, Moore himself was skeptical of the idea that things
could be ‘good for’ someone; I briefly address this skepticism later.
According to the analysis offered here, goodness for a person must be understood on the model of a healthy love relationship. Such relationships have
characteristic features that explain both their importance in our lives and their
reason-giving force. Achieving our good at a time is a matter of bringing
ourselves into relations with various persons and pursuits, activities and undertakings, that exhibit these same features. Faring well is, we might say, a form of
successful loving.
1. ‘Fit’
The property of being good for a person is instantiated when a particular
relation holds between a person and some activity, pursuit, individual, or
undertaking. We would therefore expect theories of welfare to specify the
nature of this relation. Yet some of the more prominent theories of welfare in
fact say little, if anything, about it, and this suggests that they must be offering
analyses or accounts of some property other than personal good. It may be true,
for example, that what is good for a person, P, is something she would want herself
of well-being and self-interest, treating ‘well-being’ as ‘the general term for evaluating success in life’ and
treating self-interest as a largely biological notion. But he treats the notions of well-being and welfare as the
same. See Raz (1986: 294–5).
⁶ For a clear discussion of the notion of relational value, see Railton (1986a: 10–11).
⁷ I here follow Wayne Sumner’s way of expressing the point. See Sumner (1996).
110
Connie S. Rosati
to want were she fully informed and rational—were she ‘P⫹’.⁸ Presumably P⫹
would see what relations obtain between P and X, and so what P⫹ would want
for P in light of what she sees may reliably indicate P’s good—at least if her
desire springs in part from concern for P.⁹ Even so, the fact that P⫹ would want
X for P does not tell us the nature of the relation between X and P. Likewise,
what is good for P may be what anyone (including P) ought to want for P out
of concern for her or for her sake.¹⁰ But surely whether one ought to want X for
P for her sake depends on the relation between X and P. And so to say that what
is good for P is what one ought to want for her sake is not yet to tell us what that
relation is.¹¹
As a starting point in our effort to understand personal good, we might say
that it consists in a relation of ‘fit’ or ‘suitability’ between persons and things:
those activities and pursuits, commitments and relationships, that are intrinsically, nonmorally good for a person are all and only those that fit or suit her.
Taking this metaphor as our lead, we might describe the central problem for
theorizing about personal good as the problem of characterizing the relation of
fit or suitability between a person P and any X that constitutes X’s being
(intrinsically, nonmorally) good for P. The notion of goodness for a person is
itself a normative notion. Accordingly, any characterization of the relation of
fit or suitability must be normative: when that relation obtains, the fact of its
obtaining must give rise to normative reasons.
Now whether activities, pursuits, commitments, or relationships fit or suit a
particular individual depends on a variety of factors, among them, her physical
makeup, personality, capacities, and circumstances, including her culture. And
those goods that do fit or suit her may come to do so in various ways. Some
goods come ready to wear—they fit right off the rack. The clearest cases
concern those instrumental goods that we need, physiologically, in order to
develop properly and exist in good health, such as appropriate nutrition and
medical care, and those ‘primary goods’ that a person will want no matter what
⁸ See Railton (1986b). See also Railton (1986a). Railton defends a view like this, though it is not altogether
clear whether he means it as an account of welfare, as some of his writings suggest, or of what is good from a
person’s point of view. See Darwall (2002: 109 n. 6, 111 n. 21).
⁹ Darwall (2002: 31) has observed that ‘if any informed-desire standard can serve as a plausible criterion
of welfare, . . . it will be something like the following. Something is for someone’s good if it is what that person would want for herself, as she actually is, insofar as she is fully knowledgeable and experienced and
unreservedly concerned for herself.’ Darwall does not himself embrace such an analysis. I believe that it would
fail, in any case, since it still tells us nothing about the good for relation. My point here is that a fully informed
person who cares for someone and, thus, is disposed to want what will benefit her, may well respond to the
appropriate relation when she sees it.
¹⁰ Darwall (2002) has recently defended this analysis.
¹¹ I develop this objection to Darwall’s analysis of personal good in Rosati (forthcoming) and Rosati
(manuscript a).
Personal Good
111
her conception of the good.¹² But ‘ready-made’ goods may include things
intrinsically good for persons as well, as when an individual has a natural gift
and affinity for an activity. Other goods come to fit us only with effort, with
some alteration or tailoring, deliberate or accidental, undertaken by ourselves
or (as is typical in childrearing) by others on our behalf. The needed adjustments may sometimes include that we alter ourselves; some things come to be
a part of our good not because they fit us just as we are but because we bring
ourselves (or circumstances bring us) to fit them. We may need to revise our
relationships, change jobs, or adjust our routines and pastimes. But we may
also need to develop our interests and capacities or to modify our temperaments and motivational tendencies. These considerations suggest that we can
realize our good in two chief ways: we can find or produce those things
that already fit us, or we can bring ourselves to fit those things we can find,
including those we can dream up.
2. The good for relation
So what relation must an individual stand in to goods that come to fit her in
these different ways such that some and not others suit her—are good for her—
at least at a particular time? I see no way to make progress in identifying this
relation without assuming that some of our judgments about a person’s good
are roughly correct and then investigating the relationship that seems to hold
between a person and those things judged to be a part of her good.¹³ I shall proceed accordingly, though my path will be a bit circuitous. I first look outward
in an effort to identify important aspects of a person’s achievement of her good.
I then explore how a person must sometimes work to bring herself into the
good for relation with things. Finally, I turn inward to our experience of our
activities and undertakings in order to uncover the nature of that relation itself.
An example
Let’s begin with an example of a life in which an individual attains things we
would judge to be good for him and, indeed, attains them despite difficult
¹² For discussion of the place of biological wants and needs, see Raz (1986). Rawls first presented the
notion of primary goods in (1971: § 15). I suggest here that primary goods are instrumental goods, but careful examination may show some, in fact, to be intrinsic.
¹³ To proceed in this way is not to privilege our intuitions about personal good. We seek a plausible interpretation of our experience and a corresponding metaphysical picture of personal good, and the results of our
enquiry might well upset some of our antecedent assessments of how an individual is faring.
112
Connie S. Rosati
circumstances. As he tells the tale in his Autobiography, Malcolm X was born to
a dark-skinned Baptist minister and a light-skinned mother (Haley and
Malcolm X 1965). Perhaps because of the difference in skin tone, Malcolm
speculates, his parents fought bitterly, his father sometimes beating his mother.
As a young boy, his family’s home in Michigan was burned by members of a
white racist group, and when Malcolm was 6, his father was killed, presumably
by members of this same group. The effects of his father’s death, compounded
by the effects of the depression, were devastating. Ultimately, the pressures of
poverty, of caring for eight children, and of fending off insensitive state welfare
workers drove Malcolm’s mother into a mental illness from which she never
recovered. Malcolm and his siblings were dispersed to various foster homes.
His misbehavior led to his placement briefly in a detention center at 13.
Malcolm describes as the ‘first major turning point’ of his life an encounter
with a teacher the year before he was to enter high school. When Malcolm, who
was among the top students in the class, expressed interest in becoming a
lawyer, the teacher responded that this ‘was no realistic goal for a nigger’ and
suggested that he become a carpenter instead (ibid.: 38).
Malcolm found himself withdrawing from white people, he tells us. He
moved to Boston to live with a half-sister from his father’s first marriage, and
his years there marked a steady decline. He obtained his first job, shining shoes
at the Roseland State Ballroom, and moved on to positions as a busboy and
then a dishwasher and a sandwich man on a train. As his own drug use progressively escalated, he turned to drug dealing, pimping, and finally to burglary. He hit bottom when he was ultimately busted for the last in a string of
robberies, receiving a ten-year prison sentence. He was then not quite 21.
During his imprisonment, Malcolm’s siblings began to write to him about
the teachings of the black Muslim leader Elijah Muhammad. He went from a
state of nonbelief—his fellow prisoners had nicknamed him ‘Satan’—to a state
of devout faith: ‘I had sunk to the very bottom of the American white man’s
society when—soon now, in prison—I found Allah and the religion of Islam
and it completely transformed my life’ (ibid.: 153). He began the transformation by following his brother’s instructions to stop eating pork and smoking
cigarettes, which he initially thought his brother offered as part of a ploy to procure his release. Over many years, he developed through deliberate exercises
(such as copying the dictionary) his reading and writing ability. He expanded
his knowledge through voracious reading of books in the prison library. He
engaged in a lengthy correspondence with Elijah Muhammad. And he learned
how to pray. ‘I still marvel at how swiftly my previous life’s thinking pattern slid
away from me, like snow off a roof. It is as though someone else I knew had
lived by hustling and crime. I would be startled to catch myself thinking in a
Personal Good
113
remote way of my earlier self as another person’ (ibid.: 173). Malcolm went on
to become, as we all know, a prominent Muslim leader; he married, and had a
loving relationship with his wife and children. He had begun to moderate his
views about the evil of whites as a result of a trip to Mecca shortly before his
assassination.
Now we can well imagine many ways in which Malcolm’s life might have
gone better for him. It would have been better for Malcolm had his father not
died, had his mother not become ill, had he been able to pursue his dream of
becoming a lawyer, and, most fundamentally, had he not lived in a racist
society.¹⁴ Still, it is a remarkable fact that, against many odds, Malcolm X
ultimately achieved a good life. Indeed, he salvaged what could have been a
lost life.
So how was Malcolm able to succeed under such adverse circumstances?
How is it that despite horrible experiences as a child, despite meager options
and misguided choices as a teen and a young adult, Malcolm managed to
achieve a good life rather than a life ruined for good? The short answer, in my
view, is that he was an autonomous agent—the sort of being whose good is
achieved, in intriguing ways, largely through autonomous activity. As I will
suggest, Malcolm X’s life illustrates the fact that we cannot be passive in the
realization of our good and not simply because we must search out our good or
because our good includes activities of various kinds. Rather, we cannot be passive because our activity is essential to making certain things a part of our good.
Like a successful relationship, achieving and maintaining our good requires
effort. Some goods do, as already noted, come to a person naturally. But many
of the more fundamental parts of a person’s good—including many of her
deepest commitments, longer-term projects, and central avocations—require
that she work herself into a relationship with a particular good.
In speaking of working one’s way into a relationship with something, I mean
to point to a familiar experience. We have all, at one time or another, undertaken a new activity or pursuit—say, climbing or dancing, practicing law or
accounting—that we did not initially find intrinsically motivating. Perhaps we
struggled with it, perhaps we even disliked it, perhaps we pursued it solely for
instrumental reasons. Over time, however, engaging in it came to seem quite
natural to us, even if not effortless. Our question, again, concerns, the relation
that holds in these and other cases.
¹⁴ Dworkin (1990: 73) has argued that the circumstances that make a life better and are too often unavailable include that a society be just. Obviously, we don’t know for certain that Malcolm’s life would have been
better had it been different in the enumerated ways. But our ordinary value judgments favor the prediction
that it would have been better.
114
Connie S. Rosati
Making something our good
It may help us to get a grip on that relation if we first look at how a person can
get herself into it. How can she bring it about that something she may not,
heretofore, have valued is now a part of her good? We can gain some hints of
what the process might involve from Pascal’s advice to the person who finds
himself unable to believe in God even though he accepts the conclusion of the
wager argument. The would-be-believer regards belief in God as unsupported
by the evidence and as foreign to himself, though he recognizes the apparent
value of believing. Pascal instructs him as follows:
Endeavour then to convince yourself, not by increase of proofs of God, but by the
abatement of your passions. You would like to attain faith, and do not know the way;
you would like to cure yourself of unbelief, and ask the remedy for it. Learn of those
who have been bound like you, and who now stake all their possessions . . . Follow the
way by which they began; by acting as if they believe, taking the holy water, having masses
said, etc. Even this will naturally make you believe. . . . (1910: 233, emphasis added)
Pascal’s advice is akin to Aristotle’s familiar ideas about the role of habit in
the acquisition of moral virtue, though as I will explain in a moment, his discussion contains an additional point of importance to personal good. Pascal
prescribes that the would-be-believer ‘go through the motions’, imitating the
person who already believes. We might likewise suggest that to make one’s own
what was formerly alien a person must imitate the person who already values
that thing.¹⁵ She must act toward it as does the person who already regards it as
a part of his good, treating that person’s relation to that thing as an ideal.¹⁶ Of
course, learning to value in this way requires work and practice. We become
better friends, spouses, parents, and colleagues, better practitioners of our
hobbies and crafts, with effort and experience.
But just what is one trying to achieve in imitating the believer or the valuer? As
Pascal describes it, these outer imitations have an internal aim: the alteration of a
person’s motives and, often concomitantly, the alteration of her self-conception.
The would-be believer enacts the role of the believer in order to induce genuine
belief—to become a believer—not merely to put on a good performance. Pascal
might also have said that he seeks to induce genuine love of God: he tries to
bring it about that his believer-like actions come from the inside out, from the
natural motive to such action. The would-be valuer is similarly trying to induce
¹⁵ If, as Elizabeth Anderson has argued, there are different modes of valuing appropriate to different types
of goods, then this will partly involve coming to feel as natural the mode of valuing appropriate to a given
thing. See Anderson (1993).
¹⁶ Velleman (2002) describes a process akin to the one described by Pascal and like the one discussed in
this essay.
Personal Good
115
an attitude toward a thing that will govern her actions, feelings, and intentions,
so that her valuing actions likewise come from the inside out. Having induced
that attitude, she comes to occupy a perspective from which she sees herself as
having various noninstrumental reasons for action.
We see this basic process at work in the life of Malcolm X. Malcolm came to
engage, initially for instrumental reasons, in the activities of the person for
whom practicing Islam is a central good. Over time, he underwent motivational change and acquired a radically new self-conception, with the result that
he acted and valued from the inside out, living the life of a genuine believer, a
genuine valuer of his wife and children, a genuine honorer of his commitments, a genuine contributor to his community.
To emphasize, as I have thus far, Malcolm’s own role in bringing about the
goodness for him of the life he had is not to overlook the constraints within
which he operated. After all, he didn’t choose his upbringing or circumstances.
He didn’t choose to be in jail or select the religion adopted by his siblings.
Moreover, given the facts about Malcolm, it is surely no accident that his
decision to become a follower of Islam and to structure the life he did from
within that religion worked out so well for him. He clearly had the intellectual
aptitude, personality, and capacity for self-discipline to have become a lawyer,
had circumstances permitted, and to become, as circumstances did permit, a
spokesperson and leader in his church and community.¹⁷
Still, it would be a mistake to say that the life of a devotee of Islam was good
for Malcolm just as he was when he first learned of the religion. Of course, it was
true even at that time that it was possible for him to achieve a good life as a
member of Islam. It was true of him, even at that time, that were he to choose
this life and were he to pull it off, then it would be good for him. What wasn’t
true of him at that time was that being a devotee of Islam was good for him, that
the tenets of Islam gave him noninstrumental reasons for action. The problem
is not the epistemic difficulty that he didn’t realize that it was good for him or
that it gave him noninstrumental reasons for action. Rather, he did not stand
in a relation to the practice of Islam such that it was true of Malcolm that it was
good for him. He had to make it true through successful transformation of
himself. Of course, to say that being a devotee of Islam did not give him
noninstrumental reasons to act is not to say that he had no reason to act as he
did. It is only to say that, with some qualifications, neither the reasons he had
for undertaking his religious practice nor the reasons he had for acting as one
would from within that practice concerned what was good for him at that time.
¹⁷ In addition, Malcolm’s father evidently advocated black-race purity and the return of black Americans
to Africa, and so he must have had early exposure to ideas not unlike the teachings he ultimately embraced.
116
Connie S. Rosati
One might suggest that, nevertheless, undertaking to become a devotee of
Islam was good for Malcolm even before being a devotee of Islam was good for
him. I think we can acknowledge at least two senses in which this claim is true.
First, perhaps any self-discipline intrinsically benefits a person, even when it
will not produce a successful conversion; in that sense, the undertaking was
good for him. Second, in Malcolm’s case, where the conversion succeeded, the
undertaking put him on the track to a transformation that enabled him to
achieve a life good for him; it made (instrumental) sense for him to try to realize
that life insofar as his concern was to achieve a good life. Yet to concede these
two points is not to allow that in advance of his self-transformation, it was true
of Malcolm that being a devotee of Islam or even becoming one was intrinsically good for him. We can say in advance only that it would be good for him if
he chose it and made it good for himself, and so when we say, in advance of an
agent’s efforts, ‘you should do X, it will be good for you’, or ‘if you did X, it
would be good for you’, the implicit suggestion of a benefit yet to be brought
into existence must often be taken seriously.
There is the promise of a benefit, as it were, but a promise premised on a
gamble. This is the extra element, I suggested earlier, that we find in Pascal’s
advice to the would-be believer and his appeal to habituation. As Pascal
describes it, the would-be believer does not seek to believe because theoretical
reason convinces him of God’s existence. Rather, he draws a practical conclusion in favor of making the wager. He does not—indeed, cannot, as Pascal
describes the balance of arguments—know in advance that God exists or that
he will benefit from belief in God. Instead, he gambles on the benefits of belief
and his efforts at habituation are in the service of making the wager. The person who seeks her good often finds herself in the position of making a gamble.
She need not be convinced antecedently that something is good for her, and
indeed, often cannot know in advance that it either is or will be good for her.
But she might nevertheless see how it could be rationally supportable for
her, yielding a life that would make sense to her while she was living it and
affording her reasons for action. She may, as a result, draw the practical
conclusion that it makes sense for her to regard it as her good ‘in prospect’.¹⁸
¹⁸ Let me here emphasize two qualifications to my description of the process of making something a part
of one’s good. First, we ought not to think of it as implying rigidly ordered and temporally distinct steps. It
isn’t, for example, that a person must first grasp how a certain good could be rationally supportable for
herself, then engage in activities in order to induce motivational change. In the case of Malcolm X, for
instance, he began to follow some of the instructions urged on him by his siblings before coming to believe
in the truth of Islam and coming to accord it a central place in his life. A person may come to see how something might be rationally supportable for herself, for instance, only after she has come to engage in certain
activities and begun to induce various other cognitive and motivational changes. Second, we shouldn’t overintellectualize the work done by an agent in seeing how a certain good might be rationally supportable for
herself. She needn’t have an argumentative scheme before her. Nor need she recognize something as
Personal Good
117
When a person, P, undertakes an activity, adopts an aim, or begins a
relationship, claiming that she does so in order to achieve her good, she
becomes subject to various agent-relative, hypothetical requirements. That is,
having undertaken to X, certain hypothetical requirements come to apply to
her, such as the requirement to undertake A, if A is a necessary means to X.
Someone can, of course, be under a hypothetical requirement while lacking
any reason to act at all.¹⁹ The requirement on her is simply this: having undertaken to X and believing A is a necessary means to X, P must not continue to
pursue X and refuse to undertake A. P might satisfy this requirement either by
giving up X or by undertaking A. The requirement is hypothetical in that it is
conditional on a hypothesis P accepts in adopting X as an end, namely, that in
achieving X she will achieve something that is good for her.²⁰ The person who
self-interestedly undertakes a pursuit thus makes a certain gamble—in choosing a particular career, for instance, she may gamble on a successful transformation of herself. No evidence she has at the outset can force her to—or assure her
of—the conclusion that what she has undertaken will in fact be good for her.²¹
If her hypothesis is correct, then she will have had reason to undertake or
achieve X. But its being correct is contingent upon the success of that very
undertaking. She will be justified in going after X as a part of her good only if the
outcome of that very endeavor is success in making X good for her—success in
entering into the good for relation with X.²² Malcolm’s undertakings, then, were
not good for him at the time of his undertaking them, but in acting successfully
on his choices, he brought it about that certain things were good for him and
that he therefore had reason for—was justified in—having acted as he did.²³
rationally supportable or good for her under that description. She must, however, have some sense of
considerations in light of which a good makes sense for her.
¹⁹ I have benefited here from Darwall’s recent defense of his ‘rational care’ theory of welfare, adapting a
line of argument he makes in discussing the normativity of welfare. See Darwall (1983: 15–17, 43–50). See
also Broome (1999).
²⁰ Darwall (2002) argues that what is good for a person is what one ought to want for her for her sake,
that is, insofar as one cares for her. One who cares for another, he maintains, is under certain hypothetical
requirements, requirements conditional on a hypothesis the person accepts in caring for another, namely
that the object of his concern is worthy of care. I here make a structurally similar point about those things
one undertakes or seeks out as a part of one’s good.
²¹ The point I make here is akin to the point Velleman (1989b) has made in his discussion of epistemic
freedom.
²² The idea conveyed in this sentence may seem to resemble Bernard Williams’s claims about ‘moral luck’,
but our ideas differ in a critical respect. On Williams’s view, the success of certain choices, which depends on
matters outside of one’s control, determines whether one was justified in what one did. The underlying picture seems to be that one fails or succeeds, in large measure due to luck, in achieving something the actual
value of which is fixed quite apart from one’s efforts. If I am right, in the case of personal good, whether something of value exists to be realized at all often depends on one’s successful efforts. See Williams (1979).
²³ Of course, a person’s choice can be justified even if her endeavor fails, at least in the sense that the choice
was, in light of the information available to her at the time, a rational or reasonable one to make. It involved,
we might say, a reasonable risk.
118
Connie S. Rosati
One might be tempted to say more. Given the truth of determinism, there
will be a fact of the matter about what Malcolm would choose and about his
prospects for success. His gamble, one might argue, is thus purely a function of
his epistemic position and has no metaphysical import. It is, of course, true,
that in knowing he would succeed, were he to undertake certain pursuits, he no
longer makes a gamble. Still, what he would have known about had the facts
become available to him is not a benefit independent of his efforts, but, rather,
one conditional on his efforts, and conditional in a particular way; he will stand
in a certain normative relation to some things only as a result of his efforts.
Because his standing in the good for relation to a thing is conditional on his
efforts, it remains true that it is up to him whether it will be a part of his good,
and nothing forces it upon him as his good. We can thus acknowledge the
deterministic truths behind the claim that any gamble in Malcolm’s efforts was
epistemic, while insisting that being a devotee of Islam became good for
him only by his having chosen it and having succeeded in his efforts at selftransformation and that undertaking to become a devotee was good for him
only in those senses already allowed.
An analysis of good for a person
So what relation did Malcolm have to Islam and to his other pursuits and
commitments after his transformation, when they were good for him, that he
didn’t have before? Thus far, I have assumed an intuitive grasp of that relation.
We must now examine it directly. Our ordinary judgments that something is
good for us or good for another rest on our experience. If we are to understand
the feature that we seem to detect, we must attend, then, both to our observations of others and to the phenomenology, to the experience of our own
involvement in various activities and pursuits. Of course, I can speak directly
only of my own experiences, but I hope that the two I will recount prove
sufficiently familiar to yield some useful insights.
Here is the first experience. About two years ago I took up running. While
I prefer swimming, running (I have read) better helps to maintain bone density.
I have managed to keep up this exercise regimen with moderate success,
but I still don’t enjoy it. I like how I feel after running but not the running itself.
It’s a big, sweaty bore. And so these two years later, I still have to drag myself to
the gym. I do think that running is good for me but not intrinsically, not in the
way it was, say, for Florence Griffith Joyner, and not simply because, as a
runner, FloJo was spectacular whereas I am spectacularly bad. For a person’s
good surely can include things for which she has no real talent and, conversely,
exclude some things for which she has great talent.
Personal Good
119
Here is the second. A number of years ago, I took a brief hiatus from
philosophy to study law. During the academic year, my law courses were all consuming. But in between semesters, I would sometimes steal away to a café and try
to do a bit of philosophy. I would sit, sipping tea and contemplating some deeply
perplexing problem, and I would experience a sense of calm. Kinesthetically, a
warmth spread throughout my body; my muscles and mind were relaxed.
Psychologically, I felt at home with myself and my activity; I understood and felt
comfortable with who I was and what I was doing. It was not always like this.
I can remember as an undergraduate finding philosophy awfully dull reading. I
found the issues interesting and challenging, to be sure; I saw philosophy as
genuinely valuable. But in contemplating whether to go to graduate school, I
wondered whether I would ever develop the requisite internal motivation and,
more deeply, whether I would ever come to think of myself as a philosopher.
Now I am inclined to say that doing philosophy—or more precisely,
engaging in the kind of intellectual exploration characteristic of, if not peculiar
to, philosophy—is good for me. Not instrumentally, as running is, but in
something like the way that running was good for Joyner and practicing Islam
was good for Malcolm X. Assuming that this is correct, the question is how to
understand it. I doubt that we should simply say that in those moments of
doing philosophy, I was happy or contented. While true, this oversimplifies the
experience and disguises more subtle elements of it. For it was nothing like the
happiness I feel on encountering an old friend, or the satisfaction I feel when
successfully completing a task, or the pleasure I feel watching trashy television.
I said earlier that one must work oneself into a relationship with many of
those things that comprise one’s good, that like a successful relationship,
achieving and maintaining one’s good requires effort. I want to take this comparison to interpersonal relationships quite seriously. Ordinary talk suggests
that we should. People often speak of their work, for instance, as they would of
a lover. They are passionate about it, they tell us; they find it exciting or stimulating or fascinating. They talk of finding their niche much as they talk of finding their ‘soul mate’. Of course, interpersonal relationships can be good or bad,
healthy or unhealthy. I propose to explore common characteristics of healthy
love relationships. The good for relation is especially salient in such relationships and so exploring them will prove useful to understanding personal good.
Consider, then, typical features of such relationships. First, they tend to support their participants’ sense of their own value. Having a sense of one’s own
value should not be confused with having those feelings of self-esteem that
come from worthwhile activities or accomplishments.²⁴ In fact, to have a sense
²⁴ I say more about this distinction in Rosati (manuscript b).
120
Connie S. Rosati
of one’s own value is not really to have a feeling at all. Rather, it is to have
internalized a piece of knowledge that informs how one thinks and feels more
generally. Partners in a relationship obviously will not feel good about themselves in all ways at every moment; their self-esteem will be affected by events
both external and internal to the relationship. But healthy relationships tend to
support the partners’ stable sense of themselves as basically lovable, that is, as
worthy of love. Second, such relationships tend to be enlivening rather than
enervating. Relationships require work; they can be frustrating and tiring. But
healthy relationships tend to increase or at least support people’s receptivity
and to expand, deepen, or support their interests, whereas unhealthy ones can
lead them to be withdrawn, angry, depressed, and apathetic. Third, they provide
an important component of an individual’s identity and a sense of direction in
life; in these ways, they contribute to an individual’s self-understanding.
Partners think of the ‘we’ as a part of who they each are, and the relationship
serves as one organizing principle in their lives. Finally, healthy relationships
tend to be self-perpetuating (which is not to say eternal), sources of internal
motivation. People who are in a healthy love relationship have an appreciation
of the relationship and their partner and tend to behave in ways that both
manifest and maintain their investment in it.
When love relationships deteriorate or fall apart, especially when one party
is betrayed, the wronged or abandoned parties often feel uncertain of who they
are and what their lives are about. They may find themselves feeling aimless and
uprooted, insecure and depressed, and in doubt about whether they are
lovable. In these ways, the breakup of an important relationship tends to be
especially devastating.
Healthy love relationships obviously differ in many ways from the other
things that are good for us, but I believe that those things that are intrinsically,
nonmorally good for a person all exhibit essentially the same features sketched
above. They tend to support or not undermine an individual’s sense of her own
value, to enliven rather than enervate, to provide identity and direction, and to
furnish self-supporting sources of internal motivation. When a person is
engaged with an activity, a pursuit, or another person in a way that exhibits
these features, the item in question is indeed good for that person.
Must each intrinsically good thing for a person exhibit every feature listed
above? Perhaps, instead, goodness for a person is, as Richard Boyd has
suggested of other normative properties, a homeostatic cluster property (1988:
196). Though far more discussion of this question is needed than I can supply
here, I think it is reasonable to assume that all of these features must be present
at least to some degree. For one thing, a fuller characterization of these features
would arguably show them to be connected in quite fundamental ways.
Personal Good
121
It would hardly be surprising, for instance, that those things that sustain a sense
of our value can provide self-supporting sources of internal motivation, or that
those things that provide a sense of identity and direction tend to be motivating and to sustain a sense of our value. And many apparent counterexamples
that the assumption may invite can be adequately addressed, or could be, given
further refinements. I cannot defend this last claim fully here, of course, but let
me quickly address a few seeming difficulties, since doing so may help to clarify
the claims I have been making.
One might be tempted to suggest that, on my account, well-being could be
supplied by a drug.²⁵ Even if this were true, it does not strike me as a problem.
Remember, first, that this possibility would be realized only by a drug that exhibited all of the features I have described; on my view, being good for someone is
not to be confused with a bare subjective state like pleasure or ‘desirable consciousness’. I don’t know whether imbibing any currently existing drug works like
this, but if it did—if, for instance, this were the way that anti-depressants work—
then it seems to me that taking that drug could be good for a person. But
second, I have addressed only the question of what makes something good for
someone. That leaves entirely unsettled questions about what is better or worse
for a person or about what makes for a good life for a person. A drug could be
good for a person, but if it blocked achievement of other goods or diminished
a person’s life considered as a whole, she would have reason not to take it
anyway. In such cases, a person would confront, with respect to the drug, issues
she confronts with respect to many other things in her life.
Similar responses might be offered to the worry that, although I have talked
about the relation we stand in to goods of various kinds, it seems an individual
could stand in the good for relation to quite wicked things. I have been told
that John Lee Malvo, the younger of the Washington, DC snipers, reportedly
spoke with enthusiasm and pride about the details of the killings. He expressed
no regrets and in fact indicated that he found the killings personally fulfilling.²⁶ Again, being good for a person is not, on my view, a bare subjective state,
and so Malvo’s feelings, even assuming we can trust his report, are not dispositive. Still, quite complex questions exist about whether distasteful, even evil,
things are precluded from being good for a person conceptually or only
contingently.²⁷ Whatever we might finally say about these difficult questions,
even if Malvo’s activity did satisfy the analysis offered here, which I doubt, an
²⁵ Thanks to Sarah Buss for raising the question about drug use.
²⁶ I am grateful to Mark Timmons for asking about this example.
²⁷ While I cannot address these questions here, we should remember that people do seem to flourish in
cultures whose practices we would regard as unethical.
122
Connie S. Rosati
overwhelming case would exist for saying that a good life for him would not
include being a killer.
Finally, one might be tempted to suggest that on my account, we cannot say
that chocolate, for example, is good for a person, because as wonderful as
chocolate is, chocolate consumption does not have the features I have
described.²⁸ This does not seem to me to be a problem either, because I am
inclined to deny that chocolate is, in the usual case, good for a person. This
example simply illustrates the broader point that many of the things we may
ordinarily consider good for a person are only instrumentally good for her. My
analysis concerns what it is for something to be intrinsically, nonmorally good
for a person.
I have assumed that when something is good for a person all of the aforementioned features are exhibited. But my claim is not quite that goodness for
a person is identical with these features. It is that being good for a person is the
second-order relational property of being productive of this set of features.
P stands in the good for relation to X when X is productive of this set of features
for P. Since the features are ones exhibited in a healthy love relationship—a
relationship of reciprocal love—we might say that the things that are good for
a person are those she can successfully love. When I say she can successfully love
them, I mean that her love of them is, in the way already described, rewarded;
her activities, pursuits, and so on ‘love her back’, so that her relation to them
exhibits the features described. Unfortunately, the word ‘love’ may conjure up
ideas of romance and sexual passion that form no necessary part of the relation
I have in mind, so we might say instead that goodness for a person is a relation
of rewarded appreciation.
I believe that my analysis captures our experience better than alternative
views.²⁹ This matters because the goodness of something for a person at least
seems to be detectable in experience—in our observations of others lives as well
as our own—and so our experience should provide clues as to the nature of the
property.³⁰ I believe that it is precisely when a person stands in the relation
²⁸ Thanks to David Sobel for raising this problem.
²⁹ Even sophisticated forms of analytical hedonism, I believe, must oversimplify, either ignoring some of
the features I have described or adopting an implausibly broad notion of pleasure in an effort to encompass
all of them. And some important theories of personal good, such as Darwall’s recent analysis of welfare as
what one ought to want for a person for her sake, do not address our experience at all or do so only at the level
of normative theory. See Darwall (2002: ch. 4). In my view, the nature of personal good is to be found in
those features of our experience that explain why one ought to want them for someone for her sake. Full
information informed-desire theories take account of our experiences in the sense that, through the requirement of full information, an individual considers what her experience of different choices would be like. See
Railton (1986b). But that is not the kind of experience I have been discussing.
³⁰ The claim that goodness for a person is a property detectable in experience should not, of course, be
confused with the claim that goodness for a person just is an experience or subjective state.
Personal Good
123
described above to her principal activities and undertakings, and to the central
people in her life, that we are inclined to describe her as thriving. It is precisely
when she stands in this relation to a particular aim, activity, or person that we
are inclined to say that it ‘fits’ or ‘suits’ her. When we judge X good for P we are,
accordingly, doing one of three things. First, we may be acknowledging or
remarking on the actual existence of this relation between X and P. Consider
how we respond when a friend has finally found the ‘right’ person and we see
him happy, open, and planning his future, or how parents respond when an
adult child settles into a career she finds exhilarating. Second, we may be
expressing our belief that P could stand in this relation to X, as when we remark
that X would be good for her. We sometimes think it would be a good for a
friend if he devoted himself to an undertaking for which he has some talent—
he would feel better about himself, would stop floundering, would be less
insecure. Finally, we may be noticing and reporting that such a relationship is
developing. A parent watches her child’s face light up as he progressively
masters more difficult pieces of music; we watch a friend become increasingly
invested in and excited about a project he started as a lark.
3. Theoretical challenges
Thus far, I have offered intuitive support for my analysis. I now want to explore
more theoretical considerations that lend it support. I believe that the account
of personal good sketched here is well positioned not only to account for the
dualities with which we began but to address certain longstanding problems
that any plausible account of personal good must be able to address.
The dualities
I suggested at the outset that an account of personal good must square with our
nature as both humans and agents and with our experience of our good as both
discovered and invented. I believe that my analysis succeeds in both respects.
Return for a moment to interpersonal relationships. Whether such relationships will be healthy, whether they succeed in exhibiting the features described
earlier, depends on a number of factors. In general, they require for success a
basic compatibility of the parties—attraction and affection, some mutuality of
interests and fundamental values, and complementary toleration for differences and idiosyncrasies. Whether these general requirements are met clearly
depends on facts about the individuals involved, and, to that extent, on matters
partly beyond their control.
124
Connie S. Rosati
But whether they are met also depends very much on what those individuals
do. A person can, for example, behave in ways that communicate attraction and
affection, and behaving in these ways can induce, strengthen, or perpetuate the
relevant feelings. He can treat the other party in a manner consonant with his
or her value as a person, that is, in a manner that expresses recognition of the
other’s value. He can change his own bad habits and behaviors, cultivate interests that spark another person’s interest, and so on. Perhaps a person cannot
choose his basic attractions, but he can choose to behave in ways that help to
begin, nurture, and maintain a healthy relationship.
In the same way, whether any pursuit, undertaking, or relationship fits or suits
a particular individual, whether it can ‘love her back’ so that her engagement
rewards her in the way described, depends on facts about what she is like and
about the culture in which she operates. Indeed, it depends on such a complex
array of facts that we can likely have, at best, only more or less useful generalizations, and those generalizations rest on facts about our human nature. We
tend, for instance, to have some preference for undertakings and pursuits that
we can understand and appreciate, or can begin to understand and appreciate,
and for which we have some facility. Usually, an activity can reward us only
when we can meet the basic physical and intellectual demands of the activity.
But the success involved in the good for relation, while generally requiring some
facility at an activity, shouldn’t be confused with success at that activity.
Someone who is tone deaf, for example, may stick with those singing lessons if,
given other facts about him, his efforts are nevertheless rewarded.
At the same time, whether a pursuit, undertaking, or relationship fits or suits
a person also depends very much on what he does, as the life of Malcolm X
illustrates. Now I have not offered any particular theory of autonomy, but it
seems plausible to think that a great many of the actions we undertake in pursuing our goals, developing our skills, and changing ourselves count as
autonomous, even if done with no grand view of our lives in mind. It seems
plausible at least to the extent that these actions do not display those features
that lead us to doubt that an action is autonomous. Personal good, as I conceive
it, captures the way in which our good depends on our nature as autonomous
agents by recognizing our active role in determining our futures and making
our good. But it also squares with our agency in a deeper way, which I explore
momentarily.
Having shown how the analysis of personal good agrees with our dual
nature, it is easy enough to see how we might explain the duality of our experience of our good. A person will experience her good as a discovery in a number
of typical instances. The most obvious will be in cases of ready-made goods.
But in addition, in the process of living her life, a person will ordinarily learn a
Personal Good
125
great deal about herself, her capacities, likes, and dislikes. To learn that she has
a capacity for an activity and enjoys it, even if the capacity requires development and the enjoyment only comes over time, is to learn something about her
good. She learns not, as noted earlier, that the activity was good for her all
along, but that her good could, and now does, include that activity.
Although we often experience our good as a discovery, it is not surprising, on
my analysis, that our good often seems to be up to us, for it is. Insofar as something becomes good for us as a result of our efforts, it is for us to decide where
our interest shall lie. A person’s actions do not, of course, create the property of
being good for a person, but they do often determine whether it is instantiated
and, thereby, whether it is true that something is good for her. Facts about our
nature as humans and individuals will partly determine the success of our
experiments—our success in bringing ourselves into the good for relation with
things. But whether a good for us exists to be realized often depends on our
undertaking the experiment. Our good is invented as well as discovered.
Two Moorean challenges
Moore famously argued that proposed analyses of good, whether naturalistic
or metaphysical, leave an ‘open question’ and, therefore, fail, because ‘good’
expresses a simple, nonnatural, unanalyzable property. Though Moore
targeted analyses of good, his argument presents a more general challenge that
would apply as well to proposed analyses of good for a person. Moore did not
himself raise the open question against analyses of welfare perhaps chiefly
because he recognized no such species of value. On the contrary, he challenged
as nonsensical the idea that something could be ‘good for’ me or ‘my own
good’. I believe that my analysis of personal good may be able to go some way
to meeting each of Moore’s challenges.
The open question argument
Commentators disagree, of course, about whether the open question argument
has any force, and those who think it does disagree about what problem it
exposes and about whether and how the problem can be met. I limit my
remarks here to suggesting how my analysis might begin to address two related
diagnoses of the force of Moore’s argument.
Noncognitivists (and some others) have argued that the open question
argument exposes the inherent or automatic recommending and expressive
force of ‘good’, an apparent conceptual connection with the guidance of action
(Darwall, Gibbard, and Railton 1992). I am persuaded by the noncognitivists’
claims and believe that extant forms of naturalism fail to capture the
126
Connie S. Rosati
recommending force of ethical judgments. But I believe, and have argued
elsewhere, that this failing rests on a deeper one, namely, that existing naturalistic accounts of personal good do not bear the right relationship to our agency
(Rosati 2003). Moore’s argument exposes the challenge of providing an
account that can capture how something’s being good for us engages those
capacities and motives in virtue of which we are autonomous agents. Although
I will not be able to make the case fully here, let me say a bit about why my
analysis may be able to meet this challenge. As a consequence, to call something
good for someone will serve to recommend it and to guide attitude and action.
Suppose, as some have suggested, that autonomous action would not be
possible if persons lacked certain cognitive capacities and motives that specially
equip them to function as autonomous agents. These might be, as has variously
been suggested, intrinsic desires for self-awareness and self-understanding
(Velleman 1989a) or an inclination toward autonomy (Velleman 1996), a
disposition toward coherence (Smith 1997), or standing desires for desires consonant with reality and for one’s own long-term happiness (Brandt 1979).³¹
Without these ‘autonomy-making’ capacities and motives, individuals would
not be able to evaluate and act, in any reliable way, contrary to the press of their
presently strongest lower-order motives.
I have taken no position on the nature of the autonomy-making motives and
capacities, but it is plausible to believe that whatever they might turn out to be,
a person can function effectively as an autonomous agent only if she is able to
trust and value herself. Indeed, when self-trust and a sense of one’s own worth
are seriously undermined, autonomous functioning tends to be impaired.
Without self-trust a person will not be disposed to treat her choices and assessments as authoritative and to govern herself in accordance with them, and
without self-regard she won’t care to do so. Those things that engender our
trust in ourselves, that enable us to understand and feel at home with ourselves
and to appreciate our own value, will naturally engage our autonomy-making
motives. To enter into the good for relation with things is, at one and the same
time, to enhance our well-being and our autonomous functioning. It is because
personal good, as characterized here, is a feature of things that engages our
autonomy-making motives that the judgment that something is good for a
person, or will or would be good for her, has recommending force and guides
³¹ See also Rawls (1980). Velleman seems to mean for his apparently different characterizations to come
to the same thing. See Velleman (1997: 41 n. 20). Smith (1997: 318) more precisely claims that freedom is
a matter of ‘orthonomy’ rather than autonomy. Brandt’s discussion of standing motives is not offered as part
of a full account of autonomy but, rather, to explain how we might act now against a presently strongest
desire. His view, however, exhibits the basic structure of interest here. For discussion of these ideas in relation
to personal good, see Rosati (2000, 2003).
Personal Good
127
attitude and action. The fundamental point, of course, is not that one’s good
makes one more autonomous—that our faring well and functioning well are
importantly related—although that is true.³² Rather, the point is that as
autonomous agents we have motives that are naturally engaged by those things
that support our sense of our own value, and no error lies in this. There is no
antecedent reason to think that the motives that help to make us autonomous
will turn out to be irrational; we have no reason to believe that reflection would
undermine them. And our sense of ourselves as valuable is not mistaken—we
are valuable.
Good v. good for
In his discussion of egoism as a doctrine of ends, Moore expressed incomprehension of the idea that something could be ‘good for’ an individual.³³ He
writes, ‘In what sense can a thing be good for me? It is obvious, if we reflect, that
the only thing which can belong to me, which can be mine, is something which
is good, and not the fact that it is good. When, therefore I talk of anything I get
as “my own good”, I must mean either that the thing I get is good, or that my
possessing it is good’ (1993/1903: 98). Perhaps the chief difficulty Moore
alleges for the idea of relational good is that it would be a strange kind of
value—one that could give the individual but no one else a reason for action.
‘The only reason I can have for aiming at “my own good” is that it is good
absolutely that what I so call should belong to me . . . But if it is good absolutely
that I should have it, then everyone else has as much reason for aiming at my
having it, as I have myself ’ (1993/1903: 99). Something can be normative—
can be valuable or good at all, Moore appears to have thought—only if it gives
everyone a reason for action: normativity is essentially agent-neutral. There
may be grounds for rejecting this idea, since it amounts to rejecting any agentrelative values or reasons for action (Smith 2003). But I think that an individual’s good does afford more than agent-relative reasons, and it would be hard
to sustain the apparent moral importance of personal good were this not so.³⁴
The appropriate answer to Moore, I believe, lies in certain Kantian ideas
about the structure of value, which a number of philosophers have recently
explored and which I mean to exploit as well. According to what I will call the
³² See Rosati (manuscript b).
³³ See Moore (1993/1903: 96–105). Donald Regan, who agrees with Moore, has argued that the notion
of relational value is incoherent. See Regan (1983), but compare Regan (2002), where he suggests that
Mooreans can accept a picture that comes close to what ‘good for’ theorists have in mind. For critical examination of the idea of good as a distinct property, see Geach (1956) and Thomson (2001).
³⁴ Welfare theorists certainly seem to suppose that individual welfare has more than agent-relative
importance. See e.g. Sumner (1996). For a skeptical take on the need for a theory of welfare, see Scanlon
(1998: ch. 3).
128
Connie S. Rosati
‘Kantian structure’, the normativity of welfare, or of states of affairs involving
persons, depends on the value of persons; a person’s good matters because she
matters.³⁵ The importance of her good thus depends on her agent-neutral
value. If persons indeed have agent-neutral value, then they are worthy of our
concern and anyone has reason to value them. Valuing persons, insofar as they
are beings whose lives can go better or worse, will involve acting in ways that
advance, or at least in ways that do not impede, the realization of their good,
and anyone has reason to so act. If the normativity of personal good derives in
this way from the value of persons, then personal good, contrary to what
Moore thought, gives agent-neutral and not merely agent-relative reasons for
action.
The Kantian structure is obviously controversial, and I cannot attempt to
defend it here. I merely mean to note how it permits my analysis to address the
difficulty. Of course, once we accept this view of the structure of value, then the
basic strategy is available no matter what the correct analysis of welfare might
be, and so proponents of any theory of personal good could appeal to it in order
to defend the agent-neutral normativity of welfare as they conceive it.³⁶
Nevertheless, I believe that the analysis of personal good offered here fits
especially well with this structure.³⁷
If a person’s welfare matters because she matters, it makes sense that her good
would consist in those things that support her as the valuable being she is, that
respond to her value by both supporting her sense of herself as valuable and
enhancing her functioning. What is good for a person responds to the goodness of a person. Notice that on this picture, value exhibits something like the
recursive structure we find in Moore’s theory of organic unities. A person is
valuable and her welfare consists in those things that sustain a sense of her own
value. The acts of another in furtherance of her welfare amount not only to an
acknowledgment or appreciation of her value but, given the nature of personal
good, of her appreciation of her own value.³⁸
³⁵ Velleman (1999) has suggested this in his discussion of the right to die. Most recently, Darwall (2002)
has argued that the normativity of welfare is agent-neutral and depends on the agent-neutral value of
persons.
³⁶ See Rosati (manuscript a).
³⁷ In making this claim, I avail myself of a move that Darwall (2002) has made in defending his account
of welfare. He suggests that his analysis fits well with the agent-neutral normativity of welfare. I have argued
elsewhere that Darwall’s theory does not fare better than alternative theories of welfare in accounting for the
agent-neutral normativity of welfare, both because appeal to the Kantian structure is available no matter
what the analysis of welfare and because his theory does not actually provide an analysis of welfare but, rather,
identifies a property necessarily coextensive with goodness for a person. See Rosati (manuscript a, forthcoming).
I accept Darwall’s suggestion, however, that some analyses of welfare may seem a more intuitive fit with the
Kantian structure.
³⁸ For helpful discussion of Moore’s theory of organic unities and their recursive structure, see Hurka
(1998).
Personal Good
129
Conclusion
I have suggested that when something is good for a person she stands in a
certain relation to it, that goodness for a person is a form of successful loving or
rewarded appreciation. This proposal fits well, I have argued, with our dual
nature and accounts well for our experience. Much more work will certainly be
needed to address the many questions the account raises. But if the sketch
offered herein is roughly correct, we can easily understand the reason-giving
force of considerations of our welfare, as well as our efforts as parents, teachers,
and advisers, to raise children who can fare well because they can function well.
We see to our own good by bringing ourselves into the requisite relation, as
circumstances and our own capacities permit, to a variety of persons, activities,
and pursuits. We see to the good of children by raising them to be autonomous
beings—that is, beings who live good lives, at least in large part, because they
can, through their own successful functioning, make their own good.
References
Anderson, E. (1993). Value in Ethics and Economics. Cambridge, Mass.: Harvard
University Press.
Aristotle (1990). The Nichomachean Ethics, trans. David Ross. Oxford: Oxford
University Press.
Boyd, R. N. (1988). ‘How to be a Moral Realist’, in G. Sayre-McCord (ed.), Essays on
Moral Realism. Ithaca, NY: Cornell University Press: 181–228.
Brandt, R. B. (1979). A Theory of the Good and the Right. Oxford: Clarendon Press.
Broome, J. (1999). ‘Normative Requirements’, Ratio, 12: 398–419.
Darwall, S. L. (1983). Impartial Reason. Ithaca, NY: Cornell University Press.
—— (2002). Welfare and Rational Care. Princeton: Princeton University Press.
—— Gibbard, Allan, and Railton, Peter (1992). ‘Toward Fin de Siècle Ethics: Some
Trends’, Philosophical Review, 101: 115–89.
Dworkin, R. (1990). ‘Foundations of Liberal Equality’, in Grethe Peterson (ed.),
Tanner Lectures on Human Value, vol. xi. Salt Lake City: University of Utah.
Geach, P. (1956). ‘Good and Evil’, Analysis, 17: 33–42.
Gibbard, A. (1990). Wise Choices, Apt Feelings: A Theory of Normative Judgment.
Cambridge, Mass.: Harvard University Press.
Haley, A., and X Malcolm (1965). The Autobiography of Malcolm X. New York:
Ballantine Books.
Hare, R. M. (1963). Freedom and Reason. Oxford: Oxford University Press.
130
Connie S. Rosati
Hurka, T. (1998). ‘Two Kinds of Organic Unity’, Journal of Ethics, 2: 299–320.
Moore, G. E. (1993/1903). Principia Ethica, rev. edn. Cambridge: Cambridge
University Press.
Overvold, M. (1980). ‘Self-Interest and the Concept of Self-Sacrifice’, Canadian
Journal of Philosophy, 10: 105–18.
Pascal, B. (1910). Thoughts, trans. W. F. Trotter. New York: P. F. Collier & Son.
Railton, P. (1986a). ‘Facts and Values’, Philosophical Topics, 14: 5–31.
—— (1986b). ‘Moral Realism’, Philosophical Review, 95: 163–207.
Rawls, J. (1971). A Theory of Justice. Cambridge, Mass.: Harvard University Press.
—— (1980). ‘Kantian Constructivism in Moral Theory: The Dewey Lectures 1980’,
Journal of Philosophy, 9: 515–72.
Raz, J. (1986). The Morality of Freedom. Oxford: Clarendon Press.
Regan, D. H. (1983). ‘Against Evaluator Relativity: A Response to Sen’, Philosophy &
Public Affairs, 12: 89–132.
—— (2002). ‘The Value of Rational Nature’, Ethics, 112: 267–91.
Rosati, C. S. (1995a). ‘Persons, Perspectives, and Full-Information Accounts of the
Good’, Ethics, 105: 296–325.
—— (1995b). ‘Naturalism, Normativity, and the Open Question Argument’, Nous,
29: 46–70.
—— (2000). ‘Brandt’s Notion of Therapeutic Agency’, Ethics, 110: 780–811.
—— (2003). ‘Agency and the Open Question Argument’, Ethics, 113: 490–527.
—— (forthcoming). ‘Darwall on Welfare and Rational Care’, Philosophical Studies.
—— (manuscript a). ‘Darwall’s Rational Care Theory of Welfare’.
—— (manuscript b). ‘Autonomy and Personal Good: Lessons from Frankenstein’s
Monster’.
Scanlon, T. (1998). What We Owe to Each Other. Cambridge, Mass.: Harvard
University Press.
Sen, A. (1985). ‘Well-Being, Agency and Freedom’, Journal of Philosophy, 82:
169–221.
—— (1987). On Ethics and Economics. Oxford: Blackwell.
Smith, M. (1997). ‘A Theory of Freedom and Responsibility’, in Garrett Cullity and
Berys Gaut (eds.), Ethics and Practical Reason. Oxford: Clarendon Press: 293–320.
—— (2003). ‘Neutral and Relative Value after Moore’, Ethics, 113: 576–98.
Sobel, D. (1998). ‘Well-Being as the Object of Moral Consideration’, Economics and
Philosophy, 14: 249–81.
Sumner, W. (1996). Welfare, Happiness, and Ethics. Oxford: Clarendon Press.
Thomson, J. (2001). Goodness and Advice. Princeton: Princeton University Press.
Velleman, J. D. (1989a). Practical Reflection. Princeton: Princeton University Press.
—— (1989b). ‘Epistemic Freedom’, Pacific Philosophical Quarterly, 70: 73–97.
Personal Good
131
—— (1991). ‘Well-Being and Time’, Pacific Philosophical Quarterly, 72: 48–77.
—— (1996). ‘The Possibility of Practical Reason’, Ethics, 106: 694–726.
—— (1997). ‘Deciding How to Decide’, in Garrett Cullity and Berys Gaut (eds.),
Ethics and Practical Reason. Oxford: Clarendon Press.
—— (1999). ‘A Right of Self-Termination?’ Ethics, 109: 606–20.
—— (2002). ‘Motivation by Ideal’, Philosophical Explorations, 5: 89–104.
Williams, B. (1979). ‘Moral Luck’, in Mortal Questions. New York: Cambridge
University Press, ch. 3.
This page intentionally left blank
6
Moore on the Right, the Good, and
Uncertainty
Michael Smith
In Principia Ethica G. E. Moore famously argues that there is an analytic
connection between facts about which actions are right and wrong and facts
about the goodness and badness—that is, the value—of actions’ outcomes.
All moral laws, I wish to shew, are merely statements that certain kinds of actions will
have good effects. The very opposite of this view has been generally prevalent in Ethics.
‘The right’ and ‘the useful’ have been supposed to be at least capable of conflicting with
one another, and, at all events, to be essentially distinct. It has been characteristic of a
certain school of moralists, as of moral common sense, to declare that the end will never
justify the means. What I wish first to point out is that ‘right’ does and can mean nothing but ‘cause of a good result,’ and is thus identical with ‘useful’; whence it follows that
the end always will justify the means, and that no action which is not justified by its
results can be right. That there may be a true proposition, meant to be conveyed by the
assertion ‘The end will not justify the means,’ I fully admit: but that, in another sense,
and a sense far more fundamental for ethical theory, it is utterly false, must first be
shewn. (Moore 1903: 146–7)
Though Moore here claims that it is analytic that right acts maximize value, he
later came to amend this under the influence of Russell (Russell 1910, Moore
1942). He subsequently suggested that it is a priori, but not analytic, that right
acts maximize value. But though even this weaker claim can be and has been
challenged (Rawls 1971, Scanlon 1998), I will take the stronger claim that it is
analytic for granted in what follows. For what interests me is not the status of
the claim that right acts maximize value, but rather what those who accept it,
whether as analytic or a priori, should say when they are reminded that we
cannot be certain about the consequences of our actions. This is the question
that Moore goes on to address in the relevant section of Principia Ethica. Before
proceeding, however, let me anticipate a couple of misgivings.
134
Michael Smith
First, as we will see, Moore moves seamlessly between talk of defining
rightness in terms of the maximization of value and talk of defining duty in
such terms. But, it might be objected, to say that an act is right is to say that it
is permissible, and hence that it is not the case that one has a duty not to do it. It
is not to say that it is one’s duty to do it. I will, however, overlook this difference
in what follows. If it is one’s duty to maximize value then acts that are permissible, but not one’s duty, are presumably those that produce at least as much
value as some other act that maximizes value. Right acts maximize value just
the same as duties. As between the options that one faces, one’s duties are
simply those acts that are uniquely right.
Second, far from being analytic, or even a priori, some might think that
Moore’s consequentialist account of right action isn’t so much as true. As such,
they might say that they have no interest in what one who accepts it should say
in the light of uncertainty. But this would be a serious mistake. For the kind of
consequentialism that is true, even if what Moore says is analytic, is a kind of
consequentialism that deontologists can and should happily accept (Dreier
1993, Louise 2004). While Moore’s own view was that value is a simple
property, and hence would presumably have favoured a formalization of his
definition of right action along the following lines:
(x)(x’s -ing at time t is right iff x’s -ing at t maximizes value)
we can allow that values might be relativized to persons and times in a way congenial to deontology (Smith 2003). A formalization of the following kind
might therefore be more accurate:
(x)(x’s -ing at time t is right iff x’s -ing at t maximizes valuex,t)
My keeping my promise now might maximize valueme,now, and so be the right
thing for me to do, even though my now breaking my promise would
maximize valueme,later and perhaps even maximize valueme,over time. Moore’s
puzzle about the impact of uncertainty on the definition of right action is thus
a puzzle for everyone.
The paper is in four main sections. In the first I spell out Moore’s view of the
way in which uncertainty affects the proposed definition of rightness in terms
of the maximization of value. In the second section I compare Moore’s view
with an alternative put forward more recently by Frank Jackson (1991). In the
third and fourth sections I offer my own account and say why it should be
preferred to both Moore’s and Jackson’s views. To anticipate, it turns out that
Moore and Jackson are both right about something and wrong about
something. The correct view combines elements from both.
The Right, the Good and Uncertainty
135
1. Why Moore thinks that uncertainty is a problem
Why is uncertainty an issue, given the proposed definition of ‘rightness’ in
terms of ‘maximization of value’? Moore explains the problem this way.
In order to shew that any action is a duty, it is necessary to know both what are the other
conditions, which will, conjointly with it, determine its effects; to know exactly what
will be the effects of these conditions; and to know all the events which will be in any
way affected by our action throughout an infinite future. We must have all this causal
knowledge, and further we must know accurately the degree of value both of the action
itself and of all these effects; and must be able to determine how, in conjunction with
the other things in the Universe, they will affect its value as an organic whole. And not
only this: we must also possess all this knowledge with regard to the effects of every possible alternative; and must then be able to see by comparison that the total value due to
the existence of the action in question will be greater than that which would be produced by any of these alternatives. But it is obvious that our causal knowledge alone is
far too incomplete for us ever to assure ourselves of this result. Accordingly it follows
that we never have any reason to suppose that an action is our duty: we can never be
sure that any action will produce the greatest value possible. (Moore 1903: 149)
The problem is thus supposed to be that, since we can never be certain of all of
the effects of the actions we perform, still less all of the effects of the actions
which we don’t perform but which were options for us, it follows that ‘we never
have any reason to suppose that an action is our duty.’ But if this is the problem, then it seems to be rather overblown. If we cannot be certain of the effects
of our actions then, let’s agree, there can be no conclusive reason to suppose that
any particular action is our duty. But that doesn’t entail that there are no
reasons at all to suppose that any particular action is our duty—or rather, that
would follow only if reasons for belief had to be conclusive reasons, and there
seems to be no good reason to suppose that this is so.
As subsequently becomes clear, however, the claim that there is no reason at
all to suppose that any particular action is our duty is not crucial to the problem
Moore wishes to raise. He goes on as follows.
Ethics, therefore, is quite unable to give us a list of duties: but there still remains a
humbler task which may be possible for Practical Ethics. Although we cannot hope to
discover which, in a given situation, is the best of all possible alternative actions, there
may be some possibility of shewing which among the alternatives, likely to occur to any
one, will produce the greatest sum of good. This second task is certainly all that Ethics
can ever have accomplished: and it is certainly all that it has ever collected materials for
proving; since no one has ever attempted to exhaust the possible alternative actions in
any particular case. Ethical philosophers have in fact confined their attention to a very
136
Michael Smith
limited class of actions, which have been selected because they are those which most
commonly occur to mankind as possible alternatives. With regard to these they may
possibly have shewn that one alternative is better, i.e. produces a greater total of value,
than others. But it seems desirable to insist, that though they have represented this
result as a determination of duties, it can never really have been so. For the term duty is
certainly so used that, if we are subsequently persuaded that any possible action would
have produced more good than the one we adopted, we admit that we failed to do our
duty. It will, however, be a useful task if Ethics can determine which among alternatives
likely to occur will produce the greatest total value. For, though this alternative cannot
be proved to be the best possible, yet it may be better than any course of action which
we should otherwise adopt. (Moore 1903: 149–50)
The real problem that concerns Moore is that there is no plausible task for
Practical Ethics in the absence of certainty about which actions are right. There
is no way of getting from the abstract knowledge of what it is about an act that
is right that makes it right to a decision about what to do in concrete situations
in which we have only limited knowledge of consequences.
There are two parts to Moore’s response to this problem. In the first he
details a ‘humbler’ task for Practical Ethics. Even though we can never be
certain which acts maximize value, there is, he tells us, still something ‘useful’
we can establish, namely, ‘which among the alternatives, likely to occur to any
one, will produce the greatest sum of good.’ In other words, in the absence of
certainty about which actions maximize value, we can still say which acts are
likely to maximize value, given our limited knowledge. And then in the second
part he argues that, since what we would conclude if it were to emerge that an
act that we classified as ‘right’ on the basis of our limited knowledge didn’t in
fact maximize value is that our classification of the act as ‘right’ was in error, it
follows that acceptance of this humbler task for Practical Ethics gives us no
reason to question our original definition. When we call an act ‘right’ what we
mean to be saying of it, even when we engage in the humbler task, is that that
act maximizes value, not merely that (say) to the best of our knowledge at the
time of speaking it maximizes value.
Moore appears to be on strong ground in giving this two-part response. For
a similar set of worries arises in the case of less controversial definitions, and the
solution, in such cases, is the very solution Moore proposes. Consider the definition of ‘bachelor’ as ‘unmarried male.’ The worry in this case is that, since
we have only limited knowledge of the things that people have done in the past,
we can never be certain whether any particular person is an unmarried male.
There is, however, still something ‘useful’ we can do in our search for bachelors
in the absence of such certainty, for we can classify people on the basis of the
evidence available to us. In compiling a list of bachelors we thus decide whom to
The Right, the Good and Uncertainty
137
put on the list on the basis of the likelihood that they are unmarried males. But if
it subsequently comes to light that someone we classified as a bachelor was
married at the time that we made our list, then we conclude that it is our list
that is in error. To say of someone that he is a ‘bachelor,’ even when we engage
in the humbler task of classifying people on the basis of the evidence available
to us, is thus to say of him that he is an unmarried male, not merely that (say)
to the best of our knowledge at the time of speaking he is an unmarried male.
For all the appearance that Moore is on strong ground, however, both parts
of his response have been challenged. In the following section I will compare
Moore’s definition of right action, and his resultant account of the ‘useful’ task
to be undertaken in Practical Ethics, with a competing definition of right
action and account of the task of Practical Ethics proposed by Frank Jackson.
Though Jackson does not explicitly mention Moore, the arguments he gives
look like they could well have been formulated with the aim of refuting both
parts of the Moorean view.
2. Jackson versus Moore
As we have seen, Moore’s view has two parts. The first is the claim that right acts
are those that maximize value. The second is his account of the ‘humbler’ task
of Practical Ethics, given that we cannot be certain which acts maximize value.
We decide what to do by figuring out which acts are likely to maximize value.
Let’s begin by focusing on the second component.
Frank Jackson provides what seems to me to be a quite decisive refutation of
Moore’s account of the task of Practical Ethics. The refutation takes the form of
a counter-example. He asks us to consider Jill, a physician, who has to decide
on the correct treatment for her patient, John, who has a minor but not trivial
skin complaint.
Jill has only two drugs, drug X and drug Y, at her disposal which have any chance of
effecting a cure. Drug X has a 90% chance of curing the patient but also has a 10%
chance of killing him; drug Y has a 50% chance of curing the patient but has no bad
side effects. Jill’s choice is between prescribing X or prescribing Y. It is clear that she
should prescribe Y, and yet that course of action is not the course of action most likely
to have the best results. ( Jackson 1991: 467)
What Jackson’s counter-example suggests is that, at least when we do Practical
Ethics, we have no special interest in which actions are most likely to maximize
value. Though Jill’s prescribing drug X has a 90 per cent chance of maximizing
value—that is, of bringing about a complete cure—and prescribing drug Y
138
Michael Smith
only has a 50 per cent chance of having this result, our intuitive response is
that Jill should prescribe drug Y, not drug X. Jill should prescribe drug Y,
notwithstanding the fact that it is not most likely to maximize value.
Jackson makes explicit his alternative conception of the task of Practical
Ethics in his discussion of another variation on the example of Jill, the
physician, and her patient, John. This time
Jill . . . has three drugs to choose from: drug A, drug B, and drug C. Careful consideration of the literature has led her to the following opinions. Drug A is very likely to
relieve the condition but will not completely cure it. One of drugs B and C will completely cure the skin condition; the other though will kill the patient, and there is no
way she can tell which of the two is the perfect cure and which the killer drug. What
should Jill do?
The possible outcomes we need to consider are: a complete cure for John, a partial
cure, and death. It is clear how to rank them: a complete cure is best, followed by a
partial cure, and worst is John’s death . . . But how do we move from that ranking to a
resolution concerning what Jill ought to do? The obvious answer is to take a leaf out of
decision theory’s book and take the results of multiplying the value of each possible
outcome given that the action is performed, summing these for each action, and then
designating the action with the greatest sum as what ought to be done. In our example
there will be three sums to consider, namely:
Pr(partial cure/drug A taken) ⫻ V(partial cure) ⫹ Pr(no change/drug A
taken) ⫻ V(no change)
Pr(complete cure/drug B taken) ⫻ V(complete cure) ⫹ Pr(death/drug B
taken) ⫻ V(death)
and
Pr(complete cure/drug C taken) ⫻ V(complete cure) ⫹ Pr(death/drug C
taken) ⫻ V(death)
Obviously, in the situation as described, the first will take the highest value, and so we
get the answer that Jill should prescribe drug A. (Jackson 1991: 462–3)
What this example shows is that there is a clear alternative to Moore’s suggestion
that when we engage in Practical Ethics, we should try to figure out which
action is most likely to maximize value. For, much as decision theory tells us
that the right action to choose is the one that maximizes expected utility, so we
might suppose that when we engage in Practical Ethics the right action to
choose is the one that maximizes expected value. This allows us to make good
sense of our reaction to the two drugs example. For though prescribing drug X
is more likely than prescribing drug Y to maximize value—90 per cent versus
50 per cent—the 10 per cent chance that prescribing drug X will have a very
The Right, the Good and Uncertainty
139
very bad outcome—it will kill John—as opposed to the 50 per cent chance that
prescribing drug Y will merely leave him with his minor but not trivial skin
complaint, means that the expected value of prescribing drug X is much lower
than the expected value of prescribing drug Y. This well explains why we think
that Jill should prescribe drug Y rather than drug X.
So far I have portrayed Jackson as offering an alternative task for Practical
Ethics. But in fact Jackson thinks that—with a qualification to be mentioned
presently—the two- and three-drugs examples show not just that the second
component of Moore’s view is mistaken, but that the first component, his
definition of right action, is mistaken as well. Here is the relevant passage.
The other possible account of how to recover what a person ought to do from consequentialism’s value function that we need to consider holds that a person’s beliefs . . .
do not come into the picture. What is crucial is simply which action in fact has, or
would have, the best consequences. Many consequentialists write as if this was their
view . . .
There are two problems with this proposal. First, it gives the intuitively wrong
answer in the [three] drugs case. In the [three] drugs case, either it is prescribing drug
B or it is prescribing drug C which is the course of action which would in fact have the
best consequences—and Jill knows this, although she does not know which of the two
it is—but neither prescribing drug B nor prescribing drug C is the right course of
action for Jill. As we observed earlier, it is prescribing drug A which is the intuitively
correct course of action for Jill despite the fact that she knows that it will not have the
best consequences. We would be horrified if she prescribed drug B, and horrified if she
prescribed drug C.
The second problem arises from the fact that we are dealing with an ethical theory
when we deal with consequentialism, a theory about action, about what to do. In
consequence we have to see consequentialism as containing as a constitutive part
prescriptions for action. Now, the fact that an action has in fact the best consequences
may be a matter which is obscure to the agent. In the drugs example, Jill has some idea
but not enough of an idea about which course of action would have the best
results . . . We need, if you like, a story from the inside of an agent to be part of any
theory which is properly a theory in ethics, and having the best consequences is a story
from the outside. It is fine for a theory in physics to tell us about its central notions in
a way which leaves it obscure how to move from those notions to action, for that
passage can be left to something which is not physics; but the passage to action is the
very business of ethics. (Jackson 1991: 465–7)
The two related problems with Moore’s proposed definition of right action are
first, that it gives the wrong answer in the three-drugs example, and second,
that since it defines a concept that is not and cannot be action-guiding, it
cannot be a definition of our ordinary concept of right action.
140
Michael Smith
However, as subsequently becomes clear, the second objection is much more
important than the first—here we come to the qualification I mentioned
earlier.
. . . I need to note an annoying complication. I have been arguing for an interpretation
of consequentialism which makes what an agent ought to do the act which has the
greatest expected moral utility, and so is a function of the consequentialist value function and the agent’s probability function at the time. But an agent’s probability function at the time of action may differ from her function at other times, and from the
probability function of other persons at the same or other times. What happens if we
substitute one of these other functions in place of the agent’s probability function at the
time of action? The answer is that we get an annoying profusion of ‘oughts’ . . .
I think that we have no alternative but to recognize a whole range of oughts—what
she ought to do by the lights of her beliefs at the time of action, what she ought to do
by the lights of what she later establishes . . ., what she ought to do by the lights of one
or another onlooker who has different information on the subject, and, what is more,
what she ought to do by God’s lights, that is, by the lights of one who knows what will
and would happen for each and every course of action. . . . I hereby stipulate that what
I mean from here-on by ‘ought,’ and what I meant, and hope and expect you implicitly
took me to mean when we were discussing the examples, was the ought most immediately relevant to action, the ought which I urged to be the primary business of ethical
theory to deliver. When we act we must perforce use what is available to us at the time,
not what may be available to us in the future or what is available to someone else, and
least of all what is available to a God-like being who knows everything about what
would, will and did happen.(Jackson 1991: 471–2)
Jackson’s concession here is both subtle and important.
Though our concept of right action is the concept of an action that maximizes expected value, Jackson suggests that this concept is, as such, incomplete.
In order to classify actions we need to know whose expectations are at issue.
There are therefore as many concepts of right action as there are creatures with
expectations: the agent at the time of acting, an observer, the agent in retrospect, and even, at the limit, God with his perfect knowledge of everything.
The latter is, in effect, Moore’s view. But though it follows that Moore is right
when he says that ‘the term duty is . . . so used that, if we are subsequently persuaded that any possible action would have produced more good than the one
we adopted, we admit that we failed to do our duty’—there is indeed a concept
of right action relative to which this is so—we must not conclude, on this basis,
that our primary or most central concept of right action is the concept of an
action that (say) maximizes value.
In support of Jackson’s ‘profusion of “oughts” ’, it is perhaps worth emphasizing that, in the sorts of situations Moore imagines in which we evaluate
The Right, the Good and Uncertainty
141
actions with hindsight, expectations still matter. An action performed by an
agent at t1, but evaluated subsequently at t2, is still appropriately judged to be
right just in case that action maximizes expected value, it is just that the expectations in question are those the evaluator has subsequently at t2. Imagine the
original three-drugs example evaluated in retrospect when (say) it is believed
that the original assignment of probabilities was mistaken. Instead of drug A
being very likely to relieve John’s skin condition but not completely cure it, and
one of drugs B and C being almost certain to effect a complete cure, while the
other will kill the patient, the new view is that it is drug B that is very likely to
relieve the skin condition, but not completely cure it, and one of drugs A and C
which is almost certain to completely cure the condition, while the other will
kill the patient, with there being no way to tell which will do what. In this case
the subsequent evaluator should surely suppose that the right thing for Jill to
have done was to give John drug B, not drug A. In other words, by the subsequent evaluator’s lights, it is still virtually certain that the right thing for Jill to do
is not what will maximize value. It is just that it is her subsequent expectations,
not Jill’s original expectations, that are crucial for evaluating rightness.
But notwithstanding the ‘profusion of “oughts”,’ Jackson argues that one
‘ought’ stands out from the others as the ‘ought’ that is ‘most immediately relevant to action.’ What I take it he means by this is that only one of the ‘oughts’
grounds genuine criticism of the agent. Only one links up, in the right kind of
way, with a story about what the agent can appropriately be held responsible for
doing, and this should therefore be identified as our primary concept of right
action. Imagine again the revised three-drugs example. The subsequent evaluator might well think that Jill ought to have given John drug B, not drug A, as
this is what maximizes expected value where the expectations in question are
the subsequent evaluator’s own. But if the information about the drugs wasn’t
available to Jill, then there is no sense in which she failed to live up to her
responsibilities as an agent in giving him drug A. Indeed, if Jill had given John
drug B, then there is surely a sense in which the subsequent evaluator would
still be totally horrified. Jill would have done what the evaluator thinks is the
right thing to do, but her right conduct could at best have been a complete
fluke, relative to her own reasons for acting. But if this is right—if Jill can only
be held responsible for doing the best she can, given the information available
to her—then, Jackson argues, the ‘ought’ that it is ‘the primary business of ethical theory to deliver’ is the ‘ought’ defined in terms of maximization of expected
value where the expectations are those of the agent at the time of acting.
Let me sum up. Contrary to Moore, Jackson argues that the task of Practical
Ethics is to establish which acts maximize expected value, not which acts are
most likely to maximize value. Moreover and much more importantly, again
142
Michael Smith
contrary to Moore, Jackson argues that though there are alternative concepts of
right action—there are as many concepts of right action as there are expectations relative to which we could assess the maximization of expected value—
right action in the sense of action which maximizes expected value, where the
expectations are the agent’s own at the time of acting, is the primary ethical
concept, the one that hooks up in the right kind of way with what we can hold
an agent responsible for doing. Moore is thus doubly wrong when he says
‘ “right” does and can mean nothing but “cause of a good result.” ’ ‘Right’ can
mean ‘maximizes expected value’ and this meaning is primary.
3. The definition of ‘right’
Moore and Jackson give us competing definitions of right. But who is right and
who is wrong?
According to Jackson, the primary meaning of ‘right’ is the meaning that is
‘most immediately relevant to action.’ As I have said, I take it that his idea is
that right acts, in this primary sense, are those we can legitimately expect agents
to do and criticize them for failing to do. As Jackson sees things it follows that
the primary definition of right action must be given in terms of the maximization of expected value, not, as Moore thinks, in terms of the maximization of
value tout court, and it must be given in terms of the maximization of expected
value where the expectations are the agent’s at the time of action, not those of
the agent at some other time or someone else. Unfortunately, however, there is
a gap between Jackson’s premise and conclusion. We can no more legitimately
expect agents to maximize expected value than we can expect them to maximize value.
A striking feature of Jackson’s definition of ‘right’ is an asymmetry in the
treatment of evaluative and non-evaluative facts as determinants of right
action. Suppose I observe an agent behaving and wonder whether she is failing
to live up to her responsibilities as an agent, so leaving herself open to criticism.
What exactly does this involve? Jackson rightly points out that, on the
non-evaluative facts side of things, all we can reasonably expect is that an agent
does the very best she can, given the information available to her: the full exercise of such rational capacities as she has. Putting to one side cases of impaired
rationality, this means that she must form her beliefs about the means to her
desired ends in a responsible manner, given the evidence available to her, and
she must subsequently act on such beliefs in an appropriate manner as well.
When she believes that there are alternative ways of realizing her desired ends,
she must prefer the more certain option, and when she is equally certain that
The Right, the Good and Uncertainty
143
acting in different ways will realize different desired ends, she must prefer the
option that realizes the desired end that she desires more. But since no mere
exercise of such rational capacities as an agent has will ensure that the non-evaluative facts manifest themselves to her, it follows that the mere fact that she
lacks knowledge does not, as such, render her liable to criticism.
Yet while this well explains why Jackson insists that right action is action that
maximizes expected value, it doesn’t explain why he thinks that right action
maximizes expected value. Indeed, anyone impressed by Jackson’s argument on
the non-evaluative facts side of things should surely suppose that an equally
impressive argument could be made for the conclusion that right action consists not in the maximization of expected value, but rather in the maximization
of expected value-as-the-agent-sees-things. For no mere exercise of such
capacities as an agent has looks like it will ensure that what is really valuable will
manifest itself to her either.
There are, after all, cultural circumstances in which it would be wildly
optimistic to suppose that agents could, merely through the exercise of their
own rational capacities, come to judge to be valuable what’s really valuable.
Cultures that are dominated by oppressive religions, and the like, would seem
to make it very difficult, perhaps even impossible, for ordinary people even to
contemplate alternatives to what is portrayed within their culture to be valuable. More generally, to the extent that we think that our own values are an
improvement on those of our parents, the wrong thing for us to think is that we
have been more responsible than our parents in the formation of our evaluative
beliefs. Indeed, this would be a remarkably arrogant thing to think. The right
thing to think is rather that the evidence available to us, as opposed to our parents, is different, and, as a result, we are simply better placed to form evaluative
beliefs.
If this is right, however, then it seems that the most that we could ever expect
of a normal agent—that is to say, again putting to one side cases of impaired
rationality—is that they form their evaluative commitments in a way that is
sensitive to such evidence as is available to them and that they form their desires
in a way that is sensitive to their evaluative commitments. Probabilities or
confidence levels come into the picture twice over. They come in once because
we have different levels of confidence that various means are means to our
desired ends. And then they come in again because we have different levels of
confidence about what we judge to be intrinsically valuable. When an agent is
equally confident about the intrinsic value of two things, but she judges one
more valuable than the other, then she must desire more strongly the one that
she judges more valuable, and when she judges two things to be equally intrinsically valuable, but is more confident of the value of the one than the other, then
144
Michael Smith
she must desire more that about which she is more confident. The upshot is
that whether we would criticize an agent for failing to maximize expected value
turns very much on that agent’s epistemic circumstances: whether or not they
had available to them evidence of what is really valuable.
The situation can be diagramed as follows (Smith 2004).
facts about → available
→ means-ends
means to
evidence
beliefs
desired ends
about means to
desired ends
⫹→
attempt → desired
end
desires for
ends
↑
facts about → available
→ judgements
what is
evidence about
about what
of value
what is of value
is of value
What we hold agents responsible for is, at most, the elements in bold italics, for
these are the elements whose relations are governed by the agent’s exercise of
such rational capacities as she has. Has the agent attended to all of the available
evidence as regards both what is of intrinsic value and means to her desired
ends in the formation of her evaluative judgements and means-end beliefs? Do
the agent’s desires for ends reflect her judgements about what is intrinsically
valuable? Is what the agent attempts to do an appropriate reflection of his
desires for ends and means-end beliefs? But if this is right then, contrary to Jackson,
we simply do not hold agents responsible for failing to maximize expected
value. At most we hold them responsible for failing to maximize expected valueas-they-see-things. Maximizing expected value-as-they-see-things is the right
thing for them to do, in the sense that is ‘most immediately relevant to
action.’ This is what we can legitimately expect them to do and criticize them
for failing to do.
Once this becomes clear it seems to me that we see much better the attractions of Moore’s definition of right action. Unlike Jackson, Moore adopts a
completely symmetrical approach to both evaluative and non-evaluative facts
as determinants of right action. Even though we can only hold agents responsible for failing to maximize expected value-as-they-see-things, that just goes to
show that our concept of a right action isn’t tied so closely as Jackson thinks to
what we can legitimately hold agents responsible for doing. If people live in an
oppressive culture in which they are epistemically cut off from what’s really
The Right, the Good and Uncertainty
145
valuable then the very best that they can do is to maximize expected value-asthey-see-things. But since, in so doing, they fail to maximize expected value,
even Jackson must conclude that they fail to do the right thing, albeit through
no fault of their own. The concept of a right action, in the sense relevant to
ethics, must therefore be the concept of a certain sort of ideal, one that abstracts
away from the possibility of this kind of error for which we cannot hold an
agent responsible.
But once we have distinguished the concept of right action from the concept
of what we can legitimately hold agents responsible for doing in this way, there
would seem to be no stable stopping point short of the Moorean definition.
People who maximize expected value, but fail to maximize value tout court, also
fail to act in accordance with an ideal. Moreover the explanation is much the
same as the explanation in the case where people maximize expected value-asthey-see-things, but fail to maximize expected value. In each case they are
epistemically cut from a relevant domain of facts, in this case, facts about what
the means to their desired ends really are. It is therefore this latter concept, the
concept of an action that maximizes value tout court, a concept that incorporates
two ideals—one on the evaluative side of things and the other on the nonevaluative side of things—that properly captures the concept of right action.
We should therefore accept Moore’s definition of right action, not Jackson’s.
4. The task of Practical Ethics
So far, so good; but what about the task of Practical Ethics? How do we get from
abstract knowledge of what it is about an action that makes it right—that it
maximizes value—to a decision about what to do in a concrete situation? Isn’t
Jackson right that the Moorean view will either force us to conceive of the task
of Practical Ethics in the completely wrong way, ensuring that we get the wrong
answer to which action we should choose—this is what Moore himself does—
or else make that transition altogether opaque? The answer is that it doesn’t,
and that it is instructive to see why not.
The fundamental problem with Moore’s conception of the task of Practical
Ethics is not his conception of right action, but rather his conception of moral
motivation. Suppose that right actions are those that maximize value. Does this
imply that, as a right-minded agent, I will be moved, at bottom, by an intrinsic
desire to maximize value? Though Moore is not explicit on the issue, this
does seem to be his view, for having this intrinsic desire is precisely what’s
required for agents to take an interest in which action is most likely to maximize
value when they decide what to do. Imagine again the two-drugs example.
146
Michael Smith
If what Jill cares about, fundamentally, is that she maximizes value, then of
course she will prescribe drug X rather than drug Y. For prescribing drug X has
a much better chance of achieving what really matters to her. In so far as we
have the reaction that we do to the two-drugs example, we implicitly assume
that this is not what we fundamentally care about. But in that case what do we
fundamentally care about? And how does what we fundamentally care about
lead us to make a more sensible decision, a decision that squares with our
account of right action?
Consider again the relations charted in the diagram above. If I believe that
experiencing pleasure is intrinsically valuable then, insofar as my desires are
formed correctly in response to my evaluative beliefs, I will have an intrinsic
desire that I experience pleasure. And if I believe that being autonomous is
intrinsically valuable, then, insofar as my desires are formed correctly in
response to my evaluative beliefs, I will have an intrinsic desire that I be
autonomous. Being right-minded thus requires not that I have an intrinsic
desire to maximize value but rather, since it requires that my desires are appropriately sensitive to my evaluative judgements, that I have intrinsic desires for
the things I judge to be valuable themselves. This suggests a completely
straightforward explanation of why, if I am right-minded, I will decide to
perform the action that maximizes expected value-as-I-see-things.
Suppose, to begin, that I believe that experiencing pleasure is more valuable
than being autonomous. If we abstract away from the levels of confidence associated with each of these beliefs—let’s assume I am equally confident—then,
insofar as my desires are formed correctly in response to my evaluative beliefs,
my intrinsic desire that I experience pleasure will be stronger than my intrinsic
desire that I be autonomous. And if we suppose that I am more confident of the
value of pleasure than the value of being autonomous, but we abstract away
from the degree of value I assign to each—let’s assume that I assign them equal
value—then, insofar as my desires are formed correctly in response to my
evaluative beliefs, my intrinsic desire that I experience pleasure will again be
stronger than my intrinsic desire that I be autonomous.
Putting these two conclusions together, suppose that I am very confident
that being autonomous is valuable, but not very valuable, and that I am fairly
confident—not as confident, but still quite confident—that experiencing
pleasure is valuable too, but more valuable than being autonomous. Insofar as
my desires are formed correctly in response to my evaluative beliefs the relative
strengths of my intrinsic desires to experience pleasure and be autonomous will
then depend entirely on the levels of confidence and associated degrees of
value. The strength of my intrinsic desires will track the product of the levels of
confidence and associated degrees of value: the greater that product, the greater
The Right, the Good and Uncertainty
147
the strength of my intrinsic desire. My intrinsic desire to be autonomous might
therefore even be stronger than my intrinsic desire that I experience pleasure if
the difference in the relative levels of confidence is greater than the difference
in the associated degrees of value.
Finally, let’s suppose that we plug these intrinsic desires into a standard
expected utility calculation, a calculation that takes account of the different
levels of confidence I have about the various means to my intrinsically desired
ends. It then emerges that, in so far as I am right-minded, I will decide and
choose to act not just so as to maximize expected utility, but so as to maximize
expected value-as-I-see-things.
This, it seems to me, is the right thing to say about the task of Practical
Ethics. We must decide to do what maximizes expected value-as-we-see-things,
and, when we offer advice to others, we must in effect tell them what we would
decide to do if we found ourselves in their circumstances. The ambiguity of
what it means to find ourselves in another’s circumstances—whether we imagine ourselves with our own values and the other person’s expectations, or both
our own values and our own expectations—supports and explains a profusion
of ‘oughts’ of the kind noticed by Jackson.
Importantly, however, note that even if I manage always to act so as to
maximize expected value-as-I-see-things, and so act responsibly and avoid all
criticism, and even if others manage always to do exactly what I would do if I
found myself in their circumstances, we can still quite happily admit that such
actions may yet fail to meet an ideal. For it is a simple fact of life that we act in
an environment in which we are epistemically cut off from all sorts of facts
about both what is really valuable and what the means to ends really are. We
can therefore quite happily admit that the right action for anyone to perform
in their circumstances, the one against which their own perfectly responsible
behaviour in those circumstances may well and quite appropriately be judged
to be a failure, is the act that maximizes value.
Conclusion
At one point Jackson says, in defence of his own definition of right action:
We need, if you like, a story from the inside of an agent to be part of any theory which
is properly a theory in ethics, and having the best consequences is a story from the outside. It is fine for a theory in physics to tell us about its central notions in a way which
leaves it obscure how to move from those notions to action, for that passage can be left
to something which is not physics; but the passage to action is the very business of
ethics. (Jackson 1991: 467)
148
Michael Smith
One way of putting the argument of this paper is that Jackson gets things
exactly the wrong way around. We already have a story from the inside of an
agent, a story that is not ethics. This is the story of rational decision-making as
outlined in the diagram above, the story that is formalized—or anyway
partially formalized—in decision theory. The passage to action is thus not the
business of ethics itself; or, at any rate, it is not the business of that part of ethics
itself whose concern is to provide a definition of right action. The business of
that part of ethics is to come up with a definition of right action which dovetails
in the right kind of way with the theory of rational decision-making. My
suggestion is that that definition will be an idealization of the story of rational
decision-making and that this is what Moore’s definition of right action provides.
References
Dreier, James (1993). ‘Structures of Normative Theories’, The Monist, 76: 22–40.
Jackson, Frank (1991). ‘Decision Theoretic Consequentialism and the Nearest and
Dearest Objection’, Ethics, 101: 461–82.
Louise, Jennie (2004). ‘Relativity of Value and the Consequentialist Umbrella’,
Philosophical Quarterly, 54: 518–36.
Moore, G. E. (1903). Principia Ethica (Cambridge: Cambridge University Press).
—— (1942). ‘A Reply to my Critics’, in P. A. Schilpp (ed.),The Philosophy of
G. E. Moore (Evanston, Ill.: Northwestern University): 543–667.
Rawls, John A. (1971). Theory of Justice (Cambridge, Mass.: Harvard University
Press).
Russell, B. (1910). ‘The Elements in Ethics’, repr. in Philosophical Essays (London:
Routledge, 1944): 13–59.
Scanlon, T. M. (1998). What We Owe to Each Other (Cambridge, Mass.: Harvard
University Press).
Smith, Michael (2003). ‘Neutral and Relative Value after Moore’, Ethics, 113:
576–98.
—— (2004). ‘The Structure of Orthonomy’, in J. Hyman and H. Steward (eds.),
Action and Agency, Royal Institute of Philosophy Supplement, 55 (Cambridge:
Cambridge Univerisity Press): 165–93.
7
Scanlon versus Moore on Goodness
Philip Stratton-Lake and Brad Hooker
G. E. Moore’s Principia Ethica started twentieth-century moral philosophy
with the open question argument. At the end of the century, T. M. Scanlon, in
his What We Owe to Each Other, says that he is led by Moore’s open question
argument, not to Moore’s own view of the relation of goodness to reasons,
but to the buck-passing account of goodness. The buck-passing account of
goodness is the view that goodness is not a property that itself provides practical
reasons (i.e. reasons to desire, to admire, to pursue, etc.) but rather is the purely
formal (higher-order) property of having some other properties that provide
reasons. According to the buck-passing account, the power to provide practical
reasons is passed from goodness itself to the properties on which goodness is
based. In this paper we will explore the question of whether there is a good
argument from the open question argument to the buck-passing account of
goodness. We will argue that the case for the buck-passing account of goodness
is stronger than some of Scanlon’s critics make out. We will not, however, offer
a full defence of the buck-passing account of goodness. Such a defence would
have to respond to the various objections to the buck-passing view itself, and
that is beyond the scope of this paper.
1. The open question argument
Moore presents the open question argument as an argument for the view that
naturalists commit a fallacy. Unfortunately, Moore was rather vague about
what this fallacy is supposed to be (see Frankena 1939). Moore most often portrays the naturalistic fallacy as the mistake of thinking that ‘good’ is definable.
But what is most important to Moore (1993: 19) is to insist not that ‘good’
cannot be defined, but that it cannot be defined in naturalistic terms, and we
For helpful comments on earlier drafts of this paper, we gratefully thank Roger Crisp, Jonathan Dancy,
Rahul Kumar, Rob Lawlor, Mike Ridge, Bart Streumer, Mark Timmons, and Andrew Williams.
150
Philip Stratton-Lake and Brad Hooker
will interpret his naturalistic fallacy in this way. Moore also thought that the
fallacy is committed if one identifies goodness with some metaphysical property, such as the property of being commanded by God. But Moore focuses on
naturalistic definitions, and we will follow him in this respect.
What is a natural property? Moore himself seemed unclear what the answer
is (Moore 1993: 12–13; T. Baldwin in Moore 1993: pp. xxi–xxii). One answer
he offers is that a natural property is one that can exist by itself in time (Moore
1993: 93). Another answer is that a natural property is one with which it is
the business of the natural sciences or psychology to deal (Moore 1993: 13).
There are problems with both these answers, and there are other possible
answers. How exactly to characterize natural properties remains controversial.
This is not the place to explore the natural/non-natural distinction. To give
some content to the discussion, however, we will adopt a version of the
epistemological account of this distinction. This epistemological account is
intimated by Moore’s definition in terms of the natural sciences and more
recently developed by Copp (2003: 181; cf. Sturgeon 2003: 534–8).
According to this account, a natural property is one that can be known by
empirical means, and so is an appropriate object of study by the empirical sciences. And, on this account, a non-natural property is one that cannot be
known by empirical means, and so is not an appropriate object of study by the
empirical sciences. If the naturalism/non-naturalism distinction is understood
in this way, the naturalistic fallacy is committed whenever goodness is defined
wholly in terms of empirical properties.
Moore’s open question provides the argument for the view that goodness
cannot be defined naturalistically, according to Moore. Moore works with two
versions of his question, which we will call the ‘property’ version and ‘object’
version.
According to the property version, the open question is ‘Is it good that A is
n?’, where n is the natural property with which good is to be identified. This is
the version of the question that Moore uses when he considers the attempt to
define ‘good’ as ‘that which we desire to desire’. He argues that ‘good’ cannot be
defined in this way, because when we
ask ourselves ‘Is it good to desire to desire A?’ it is apparent, on a little reflection, that
this question is itself as intelligible, as the original question ‘Is A good?’—that we are,
in fact, now asking for exactly the same information about the desire to desire A, for
which we formerly asked with regard to A itself. (Moore 1993: 67)
The object version of the question does not ask whether it is good that A is
n, but asks whether an A that is n is good. Here the question is not whether it
is good that something possesses the natural property in terms of which ‘good’
Scanlon versus Moore on Goodness
151
is to be defined, but whether the thing that has this property is good. This is the
more familiar version of the open question, and is the one with which we
will work.
Moore claims that all questions of the form ‘Is an A that is n good?’ have an
open feel that they would not have if goodness were the same thing as nness. It
is because such questions have this open feel that goodness cannot be a natural
property.
There are many explanations of the openness of questions of the form: ‘Is an
A which is n good?’ One of these explanations is Moore’s own. A second is
favored by non-cognitivists. A third comes from Scanlon.
Moore thought that such questions have an open feel because the answer is
unobvious and debatable. He took it that the answer would not be unobvious
and debatable if ‘good’ were definable as ‘n’.
In contrast, non-cognitivists hold that the openness of the open question
is explained by the possibility of motivational indifference to anything characterized in purely naturalistic terms. The non-cognitivist background to this
explanation starts with the claim that judging that something is good is per se
to be motivated to act in certain ways towards that thing. But judging that
something has some naturalistic property does not necessarily imply the
presence of such a motivational attitude towards that thing.
Scanlon offers a normative explanation of the openness of the open
question. Judgements about what is good entail practical conclusions about
what would be reasons for acting or responding in a certain way. Natural facts
may provide the ground for such practical conclusions, but judging that these
facts obtain need not involve explicitly drawing these conclusions. Scanlon
(1998: 96) writes,
Questions such as ‘This is C, but is it valuable?’ (where C is the term for some natural
or ‘metaphysical’ property) therefore, have an open feel, because they explicitly ask
whether a certain practical conclusion is to be drawn.
Judging that some natural fact obtains does not entail judging that we have
reason to respond to that thing is some positive way. We do not judge that we
have such reason until we judge that that thing is good.¹ So such judgements
are always normatively open.
Scanlon (1998: 96) says that he was led to the buck-passing account of value
by this account of the open feel of the open question. The normative account
¹ For Scanlon’s normative explanation of openness to work, he must assume not only that when we judge
that A is good we judge that we have reason to care about A, but also that we do not judge that we have such
reason until we judge that A is good. Otherwise it is not clear why the question whether we have reason to
care about A is normatively open until, and just until, we judge that A is good.
152
Philip Stratton-Lake and Brad Hooker
of the open question argument involves three things: the natural properties in
virtue of which A is good, the fact that A is good, and the fact that we have
reason to respond in certain positive ways towards the good, that is, towards A.
What led Scanlon to the buck-passing account of value is that he thinks that it
offers the best account of the relation of these three things.
2. Two preliminary points
Before we evaluate Scanlon’s arguments for the view that the buck-passing
account of value provides the best account of the relation between natural properties, reasons, and goodness, it will be useful to make two preliminary points.
First, the question Scanlon’s buck-passing account of goodness is supposed
to answer need not be restricted to the relation between the natural, goodness,
and normative reasons. The question is how the properties on which goodness
supervenes (the base properties) relate to goodness and normative reasons.
Whether the base properties must be natural properties will depend on
whether thick evaluative properties, such as the property of being kind, or
generous, can be divided into distinct natural and evaluative elements.
If thick evaluative properties can be divided in this way, then the normative
buck will be passed down to their naturalistic components. If, say, kindness can
be understood as a disposition to act in certain (naturalistically specifiable) ways
towards others, plus the fact that this disposition is good, then the base property
will not be kindness as such. Rather the base property will be the naturalistic
component of kindness, i.e. the disposition to act in certain ways towards others.
The evaluative component of kindness, the goodness of this disposition, will
not be a part of the base property, but will supervene on this naturalistic base.
If, on the other hand, thick evaluative properties like kindness do not
involve distinct naturalist and (thin) evaluative components, then someone’s
kindness itself could be located wholly within the base level. On this supposition, the normative buck could not be passed down to a distinct naturalistic
component of kindness, as there is no distinct naturalistic component to play
this reason-giving role. So if thick evaluative properties cannot be divided into
distinct naturalistic and (thin) evaluative components, the base properties need
not be naturalistic. They might be non-natural, thick evaluative properties.
Second, Scanlon distinguishes valuing something from claiming that the
thing has value:
To value something is to take oneself to have reasons for holding certain positive
attitudes towards it and for acting in certain ways in regard to it. Exactly what these
reasons are, and what actions and attitudes they support, will be different in different cases.
Scanlon versus Moore on Goodness
153
They generally include, as a common core, reasons for admiring the thing and for respecting it, although ‘respecting’ can involve quite different things in different cases. (95)
According to Scanlon, ‘[t]o claim that something is valuable (or that it is “of
value”) is to claim that others also have reason to value it, as you do’ (95).
The distinction between valuing something and thinking it valuable is a
useful one. As Scanlon points out, ‘it is natural to say, and would be odd to
deny, that I value my children; but it would be odd for me to put this by saying
that they are valuable (except in the sense that everyone is)’ (ibid.). This point
is brought out more strongly in comparative evaluative judgements. We
value our children more than any other people’s children, but we do not think
our children more valuable than any other children.
But given how Scanlon defines valuing something, judging that something
has value cannot be understood as judging that others also have reason to value
it, as you do. For then, judging that something is of value would turn out to be
judging that others have reason to judge that they have reason for holding
certain positive attitudes towards it and for acting in certain ways in regard to
it. This is, at best, awkward.
What Scanlon should have said is that for you to judge that something A is
good is for you to judge that you and others have reason for holding certain
positive attitudes (e.g. desiring or admiring A) and for acting in certain ways
(e.g. pursuing, promoting, or publicly applauding A). For you to judge that A
is good is not for you to judge that they have reason to judge that they have
reason for holding certain positive attitudes towards A and for acting in certain
ways towards it. Put more succinctly, the idea is that for A to be good is for more
or less everyone to have reason to care about A.
Bearing the above points in mind, we need to clarify the relation between A’s
base properties (be they natural or not), the property that A has of being good
(being of value), and the reasons we have to care about A.
Scanlon claims that these things can be related either according to the
Moorean view or according to the buck-passing account.
Moorean view:
The base properties of A make A good, and the
goodness of A gives us a reason to care about A.
Reasons supervene on goodness, and goodness
supervenes on the base properties.
Buck-passing account: The base properties of A make A good and provide
reasons to care about A. The goodness of A does
not provide an additional reason to care about A,
but rather is the purely formal (higher-order)
property of having some other properties that
provide reasons.
154
Philip Stratton-Lake and Brad Hooker
Moore’s own view, at least in its classic early formulation, combined three
theses: (1) the thesis that good is non-natural, (2) the thesis that ‘good’ is
unanalyzable, and (3) the buck-stopping thesis that we refer to here as the
‘Moorean view’.² This third thesis is the one that Scanlon primarily has in
mind when he contrasts his own, buck-passing account with Moore’s account.
But a related contrast between Moore’s account and Scanlon’s concerns
analyzability. The buck-passing account, unlike Moore’s own, does provide an
analysis of good, namely that ‘good A’ means ‘an A that has other properties
that provide reasons to care about it’.
3. Scanlon’s two arguments
Now let us consider the content of Scanlon’s two arguments for the
buck-passing account over the Moorean view. The first is that, at least in some
cases,
natural properties provide a complete explanation of the [practical] reasons we
have. . . . It is not clear what further work could be done by special reason-providing
properties of goodness and value, and even less clear how these properties could
provide reasons for action. (1998: 96)
The second is that
many different things can be said to be good or to be valuable, and the grounds for
these judgements vary widely. There does not seem to be a single, reason-giving property that is common to all these cases. (1998: 97)
Scanlon’s first argument is poorly stated. As we understand him, the point he
is making here is not that natural properties provide a complete explanation of
the reasons we have. His idea is that the fact that A has such-and-such natural
properties provides the reasons we have. So he thinks those natural properties
provide a complete list of the reasons we have when the base properties
(the properties on which goodness supervenes) are natural properties. We take
his point to be that thinking of the goodness of A as adding to the reasons
provided by the base properties of A commits a kind of category mistake. If the
reason why the holiday resort is good is because it is pleasant, then the reason
to choose the holiday resort is just that it is pleasant, not because it is pleasant
and because it is good.
² For example, Moore states, that ‘[t]he only reason I can have for aiming at “my own good”, is that it is
good absolutely that I should have something, which, if I have it, others cannot have’ (1993: 150).
Scanlon versus Moore on Goodness
155
But let us ask whether this is an argument not against the Moorean view that
goodness is the only reason, but against the view that goodness is an additional
reason? That is, contrast:
Moorean view:
The base properties of A make A good, and the goodness of
A gives us a reason to care about A. Reasons supervene on
goodness, and goodness supervenes on the base properties.
with
Alternative view: The base properties of A make A good and provide
reasons to care about A, and the goodness of A gives us
an additional reason to care about A. Goodness supervenes on the base properties, and reasons supervene
both on the base properties and on goodness.
Scanlon’s first argument (as we understand it) certainly seems to work against
what we’re here calling the alternative view. If the properties on which goodness
supervenes provide practical reasons, then unnecessary philosophical problems
are created if goodness is taken to supply an additional practical reason. The
alternative view seems to multiply reasons pointlessly. The buck-passing
account has what is needed from the alternative view but without the unnecessary proliferation.
But does Scanlon’s first argument work against the Moorean view? If
Scanlon’s argument is effective against the alternative view, it is even more
effective against the Moorean view. The strong intuition Scanlon points to is
that we seem to have exhausted the list of reasons to care about something once
we have listed the features that make it good. We do not need to add the
fact that it is good to complete the list. This intuition clearly assumes that the
good-making features are reason-giving, which the Moorean view simply
denies. So the intuition Scanlon mentions in his first argument is even more
incompatible with the Moorean view than it is with what we are calling the
alternative view.
Scanlon’s attack on the Moorean view may seem to put too much weight on
explanations in terms of natural properties. Scanlon offers the following
examples of ‘complete explanations’ of practical reasons: ‘the fact that a resort
is pleasant is a reason to recommend it to a friend, and the fact that a discovery
casts light on the causes of cancer is a reason to applaud it and to support
further research of that kind’ (1998: 97). Is Scanlon wrong to focus on natural
properties in these examples?
It may be objected that natural facts never provide reasons by themselves,
because, when fully stated, the reason-giving fact will always involve an
156
Philip Stratton-Lake and Brad Hooker
evaluative component. So, for example, it may not be the mere fact that
something is pleasant that gives us a reason to choose it, but the fact that it is
pleasant and the fact that this pleasure is innocent. Similarly, it may not be the
natural fact that some discovery has been made that gives us reason to applaud
it. The reason to applaud it is the fact that it is a significant discovery. But the
fact that it is a significant discovery may not be itself a natural fact.
We take no stand in this paper on the issue of whether reasons always, sometimes, or never are constituted by purely natural facts. But as we noted earlier,
it is not essential to Scanlon’s argument that the base properties be natural
properties. The issue is whether it is the base properties (whatever their nature
is) or the goodness of A that provides us with reason to care about A.
Let us now move to Scanlon’s second argument for the buck-passing theory
over the Moorean theory. The crucial part of Scanlon’s second argument is his
claim that ‘there does not seem to be a single, reason-giving property’ that is
common to all things that are good (1998: 97). As a first step in assessing that
claim, let us distinguish two different ways that there may conceivably be a
single, reason-giving property common to all goods.
One of these possibilities is that there may be a single, reason-giving property
at the supervening level. In fact, this possibility is just what Mooreans hold to be
the case—i.e. that the goodness shared by all good things is the reason-giving
property. This Moorean position was the target of Scanlon’s earlier argument
that a complete list of reasons mentions the base properties but not goodness.
The other way that there may conceivably be a single, reason-giving property common to all goods is that there is a reason at the level supervened upon,
the base level. To embrace this possibility would be to accept a monism about
grounds of goodness and of reasons. Such a view might hold, for example, that
all good things contribute to well-being.
Scanlon argues at length that well-being is not such a master value.
Admittedly, showing that well-being is not the master value does not show that
there is no master value. Still, well-being does seem to be the leading candidate
for being a master value. If Scanlon successfully argues against this candidate,
no other seems likely to do better. So let us grant to Scanlon that pluralism
about goodness is correct.
Scanlon’s buck-passing account of goodness in effect claims that, whatever
the practical reason-giving properties are, what they all have in common is only
that they generate practical reasons. This relation to practical reasons can be
expressed by saying that all things with those properties are good. The
Moorean account, in contrast, claims that they all have an extra property, goodness, and that this property is not merely a relation between other properties and
reasons.
Scanlon versus Moore on Goodness
157
Thus the Moorean view holds that there are three different kinds of thing to
explain the relations among: base properties, goodness, and reasons. And the
Moorean view holds that none of these are to be defined in terms of the others.
The buck-passing account holds that there are really only two things to explain
the relation between: base properties and reasons. And the buck-passing view
then defines goodness in terms of the base properties and reasons. This contrast
between the Moorean view and the buck-passing account reveals the buckpassing account’s greater parsimony. This greater parsimony counts in favour
of the buck-passing account.
But does Scanlon’s argument about the plurality of good do any real work
against the Moorean?
Suppose a pluralist view about the grounds of practical reasons—that is,
about what the reason-giving properties are. Suppose a, b, c, d, and e are the
reason-giving properties. On the buck-passing view, that things which are
either a, b, c, d, or e are good is nothing more than the fact that a, b, c, d, or
e provide reasons to care about those things.
For the sake of contrast, now suppose a monist view, e.g. that e is the only
reason-giving property. The buck-passing account, as such, is perfectly
compatible with this monistic view. A buck-passing form of this monistic view
would hold that the fact that only something that is e is good is nothing more
than the fact the only e is a reason-giving property.
Since the buck-passing account can take either pluralistic or monistic forms,
the buck-passing account is neutral as between pluralism and monism.
Similarly, the Moorean view that goodness ‘stops the buck’ is neutral between
monism and pluralism. Thus Scanlon’s argument from value pluralism cannot
establish that the buck-passing account is superior to the Moorean view.
4. Neglected alternatives
Consider the objection that, even if Scanlon succeeds in showing that the buckpassing account is better than the Moorean view, this hardly shows the
buck-passing account to be superior to other possible accounts. Scanlon
argues for the buck-passing account merely by trying to eliminate the
Moorean view. Some critics reply that there are other views Scanlon should
have considered.
According to Dancy (2000: 163), Scanlon’s argument works by eliminating
two alternatives—the teleological account of goodness and the Moorean buckstopping account. According to the teleological account of goodness, to be
good is to be to be promoted. In contrast to Dancy’s understanding, we think
158
Philip Stratton-Lake and Brad Hooker
that Scanlon’s argument does not work by eliminating the teleological account
of goodness.³
We take the teleological account to be just one specific form that the generic
buck-stopping theory might take. In other words, some versions of the buckstopping theory of goodness are not teleological. Indeed, we formulated
the Moorean view above so as to allow for such versions. A non-teleological
buck-stopping theory of goodness denies that the only appropriate response to
the good is promotion, but holds that the goodness of something provides a
reason to desire, admire, applaud, or (sometimes) promote that thing.
Furthermore, the teleological account of goodness does not dictate a buckstopping view of the relation between goodness and reasons. A teleological
buck-passing theory holds that the appropriate response to the good is always
promotion, but also holds that the reasons to promote do not come from the
goodness but from whatever base properties ground the goodness.
Since a buck-stopping theory of goodness can be non-teleological, rejecting
the teleological account does not establish the buck-passing view. Since accepting a
teleological account of goodness does not require rejecting the buck-passing view,
a defence of the buck-passing view need not eliminate the teleological account. So
Scanlon’s argument cannot work by eliminating the teleological account.
Scanlon has separate arguments against the teleological conception of
goodness. His main point is not that this conception implies that goodness is
reason-giving (as we have just explained, there is no such implication), but that
the teleological account artificially restricts the range of responses to the good
to just one—promotion (Scanlon 1998: 87–94). The argument for the buckpassing account of goodness does not proceed by denying that promotion is
the only appropriate response to goodness. It instead proceeds by denying that
goodness is a property that warrants a positive response from us in addition to
the positive responses warranted by the base properties. The buck-passing
account is not itself inconsistent with the view that the only response the base
properties warrant is promotion.
On the question, then, of whether goodness is reason-giving, Scanlon thinks
there is only one competing theory—the Moorean account, where the relevant
feature of this account is its being a buck-stopping account. If Scanlon is right
about this, then if he can show that his own account is better than Moore’s,
he will have shown that the buck-passing account is the best account of the
³ It should be noted here that Scanlon occasionally seems to think of the buck-passing view as an alternative to the teleological view (1998: 93). But this is confused. The teleological view is, primarily, about what
the appropriate response to the good is, whereas the buck-passing view is about what gives us reason to
respond in the appropriate way, whatever this is.
Scanlon versus Moore on Goodness
159
relation of goodness, the ground of goodness, and reasons. But Dancy
maintains that there are at least two theories that Scanlon ignores, and thus that
Scanlon’s argument from elimination does not work.
The first theory that Dancy (2000: 164) accuses Scanlon of ignoring is a
Rossian view according to which the sort of things that can ground reasons
need not be the same as the sort of things that can ground goodness. Ross maintained that actions can be right but not good, and that motives can be good but
not right (1930: 4 ff.). But if an act is right (which for Ross meant ‘obligatory’),
then I have reason to do it. The reason I have to do this act is provided by
whatever grounds the obligation to do it. This ground cannot be the same as
the ground of its goodness, for, on Ross’s view, obligatory acts are never good.
In this respect, then, Ross’s view differs from Scanlon’s view that reasons and
goodness always share the same ground.
The difference between Ross’s view and Scanlon’s should not be exaggerated.
When Ross says that acts are never good, he means that they are never intrinsically good. But, in chapter 3 of The Right and the Good, Ross distinguishes four
senses of ‘good’: intrinsic, instrumental, attributive and contributive (1930:
72–3). Although Ross denies that acts are ever intrinsically good, he does not
deny that they can be good in some other sense. So although he holds that the
reason we have to do some act cannot be provided by the ground of the act’s
intrinsic goodness since an act cannot have intrinsic goodness, he can allow
that a reason to do the act may be provided by the ground of whatever nonintrinsic goodness the act has. The buck-passing view is an account not merely
of intrinsic goodness, but of all forms of goodness. So the buck-passing view is
compatible with the view that we could have reason to do acts that are not
intrinsically good.
Nonetheless, there still seems to be a divergence between Ross’s view and the
buck-passing view. Ross’s view allows that we might have reason to care about
some act that is in no way good, i.e. is neither intrinsically, instrumentally,
attributively, or contributively good. Since the act is in no way good, the reason
we have to care about it cannot be provided by whatever makes it good (contra
Scanlon), or by its goodness (contra Moore). Since Ross’s view is different from
Moore’s view, the argument Scanlon provides against Moore’s view does not
dispose of Ross’s view.
The second view that Scanlon fails to consider is Dancy’s own. This view
shares elements with both Moore’s and Scanlon’s accounts. It accords with
Moore’s account insofar as it denies that the property of being good can be
identified with the second-order property of having other reason-giving
properties. It accords with Scanlon’s view in that it denies that goodness is a
property that adds to the reasons provided by its ground (Dancy 2000: 164).
160
Philip Stratton-Lake and Brad Hooker
Dancy is right that the view we have just attributed to him and the view he
attributed to Ross need to be considered. Scanlon’s argument by elimination
needs to eliminate these views, if the buck-passing view is to be the only
alternative not eliminated. Nevertheless, these views are indeed inferior to the
buck-passing view, as we shall now explain.
Consider first Ross’s view, as sketched by Dancy. The buck-passing account
of goodness can allow that an agent can have reason to do some act which is in
no way good. It would allow this if the features that gave the agent a reason to
care about his doing act A did not give anyone else reason to care about his
doing act A. The buck-passing view does not entail that if the agent has a reason to care about A, then A is good (in some way). Rather, the buck-passing
account of goodness entails the following bi-conditional:
A is good if and only if A has properties that both give an agent reason to
care about A and give others reason to care about A.
Consider cases where you are the only person who has reason to care about
whether you do A. In such cases the fact that your doing A is in no way good
does not conflict with the above bi-conditional, and so is compatible with a
buck-passing account of goodness.
Ross’s view would conflict with the buck-passing view only if Ross held that
there could be cases in which both (1) your doing A has features that give you and
others reason to care about your doing A, and yet (2) your doing A is in no way
good. Ross might allow for the possibility of (1) and (2) because the reason you
and others have to care about doing A stems from what makes A obligatory, and
obligatory acts need not be good in any way. We do not think this is a plausible
view. If you are obligated to do A, then the features that give you reason to do A
will also give others reason to approve of your doing A. And if others have reason
to approve of your doing A, then your doing A must be good in some way.⁴
This is because the attitude of approval is only ever appropriate when it is
directed at things that are good. Ross thought that this is because it is the fact
that A is good that makes approval an appropriate response to A (1939: 261).
This makes Ross look like a buck-stopper. However, one need not agree with
Ross about that. One might just as plausibly maintain that approval is appropriate only towards things that are good because what makes the attitude
of approval appropriate is what makes its object good.⁵ In any case, if others
⁴ Ross would certainly not deny that if others have reason to approve of something, then that thing is
good (1939: 261).
⁵ Our view is that we do not approve of certain things because they are good. Rather, we approve of
different things for different reasons. We approve of good things because of their good-making features. We
admire good acts because they are heroic, or brave, or selfless, etc., not because they are good (though we
cannot approve of them unless we think they are good).
Scanlon versus Moore on Goodness
161
have reason to approve of your doing A then your doing A must be good in
some way.
Turn now to the view we attributed to Dancy himself. This view agrees with
Scanlon that, say, the pleasantness of a resort could be a reason to choose it, and
agrees that the fact that it is a good resort does not give us an additional reason to
choose it, but denies that the fact that it is good is the fact that it has other features
that provide reasons to choose it. Dancy thinks that the second-order property of
having other properties that provide reasons to choose something is distinct from
the property that thing has of being good, though if something is good it will have
this distinct second-order property. On this view, then, we always have reasons to
choose things that are good (though not because they are good).
We think there is reason to prefer the buck-passing account of goodness to
Dancy’s view. Dancy maintains both that we always have reason to care about
the good and that goodness never adds to the reasons provided by its ground.
But unlike the buck-passing view, his account leaves unexplained why goodness cannot provide us with an additional reason. In this respect, then, the
buck-passing account is a better option for those who agree with Scanlon that
goodness is never reason-giving. For the buck-passing account of goodness
explains why the fact that something is good never gives us a reason to care
about it. On the buck-passing account, the fact that something is good is the
fact that it has other properties that provide reasons to care about it, and the
fact that it has such properties cannot provide an extra reason to care about it.
Roger Crisp picks out a different neglected position in Scanlon’s argument.
Crisp starts from Moore’s famous example about choosing between two
uninhabited worlds, one beautiful, the other an ugly pile of filth. Moore held
that the beautiful world is better than the ugly one, and that this fact gives us
reason to bring it into existence. But, Crisp argues, we need not embrace the
buck-passing account of the relation between value and reasons if we reject
Moore’s view. We might opt for what Crisp calls a ‘welfarist buck-keeping nonwelfarist buck-denying’ account. Crisp refers to this account as W. He frames
W as follows:
[V]alues are properties that make their bearers good, or better, whilst reasons are facts
that speak in favour of acting in a certain way: According to view W, the beautiful
universe is indeed valuable; it is better that it exist than not. But since it matters to no
one whether it exists or not, there is no reason to bring it about. (Crisp 2000: 240)
Since W works with a welfarist view about reasons, W holds we have no reason
to bring about such a world if no one benefits from it. But since we are here
working with a non-welfarist account of value, such a world could, nonetheless, be good.
162
Philip Stratton-Lake and Brad Hooker
W does not seem as plausible as the buck-passing view. If W is true, then the
beautiful world can be good and the ugly not, without there being any reason
to prefer the good world to the bad one. But it seems that, if the beautiful world
is good and the ugly one not good, then there is a reason to prefer the good
world. So W seems false.
Crisp might reply by saying that no one would judge a beautiful world good
unless they cared about such things or were too depressed to care about what
they judged valuable. No doubt this is right, but it is not at all clear why it
should be right if W is true. If, as W allows, one’s value judgements are independent from one’s concerns and the reasons that go with these, it is not at all
clear why we have to care about something to think that it is good.
Crisp might now issue a different and more general objection. On Scanlon’s
account, the goodness of A is nothing more or other than there being practical
reasons to care about A. But there is a difficulty with taking reasons for action
as being built into (or analytically entailed by) the very concept of goodness.
Crisp’s W, for example, holds precisely that the grounds of and criteria for
goodness simply do part company with the grounds of and criteria for practical reasons. And, in making that claim, W is hardly alone!
Consider philosophers who hold both (i) that some form of consequentialism,
or the Categorical Imperative, or some form of contractualism, or Rossian pro
tanto duties determine what is morally right, and (ii) that reasons for action are
always determined exclusively by the agent’s desires (or alternatively by the
agent’s long-term self-interest). All such philosophers presumably agree with
Crisp’s W in finding it intelligible that your doing A would be good yet you have
no reason to do A, nor anyone else reason to care about your doing A. Since the
buck-passing account claims that A is good if and only if there are reasons to pursue, desire, admire, or approve of A, the buck-passing account rules out many
(seemingly intelligible) combinations of theories about goodness and reasons.
We admit that such combinations of views are intelligible (as well as popular). But are they as plausible, on balance, as the buck-passing view? In the spirit
of the normative explanation of the open question argument, we want to note
that any combination of theories that introduces a gap between goodness and
reasons invites a ‘so what?’ objection to judgements that something is good. In
other words, if you can judge that A is good without judging that you have any
reason to care about A, your response to instances of good might rightly be ‘so
what?’ On the contrary, if there can’t be a ‘so what?’ objection to a judgement
that something is good, this seems to imply that judgements that A is good do
indeed entail the normative judgement that there is reason to care about A.
As an illustration, consider Bernard Williams’s example of Owen Wingrave.
Wingrave comes from a family with a strong military tradition, but the
Scanlon versus Moore on Goodness
163
thought of a life in the army leaves him cold. Now Wingrave could acknowledge
all of the facts that his parents might cite in trying to persuade him to pursue a
life in the army. He might accept that in pursuing such a life he would be
carrying on the family tradition, that he would acquire various skills he would
otherwise not acquire, that he would be more self-disciplined and organized,
and so on. He could accept all this and still perfectly coherently respond ‘so
what?’ What is not, in our view, coherent is for him to accept that these considerations make an army life a good life for him while responding ‘so what?’
This is incoherent because to accept that these considerations make an army
life a good life is, at the same time, to accept that they count in favour of living
this sort of life. To accept that they count in favour of living this sort of life is,
of course, to accept that they provide reasons to live such a life.
That the buck-passing account of goodness rules out the possibility of
simultaneously accepting that A is good and responding ‘so what?’ provides
strong prima facie evidence in favour of the buck-passing view. Any view that
denies that judgements that A is good entail judgements that we have reason to
care about A invites a ‘so-what?’ objection to the judgement that A is good.
That is the fate awaiting any combination of a theory of goodness and a theory
of practical reasons that with one hand specifies the ground of and criteria for
goodness and with the other hand specifies a quite different ground of and
criteria for practical reasons.
5. Multiplying gaps
Let us now move on to other arguments. Moorean realism takes goodness to be
a real, simple, non-natural property that supervenes on other properties. A
difficulty with Moorean realism, according to Blackburn (1971; 1984: ch. 6;
1985; 1998: 315–17), is that it really has no explanation of why goodness
supervenes on other properties. If goodness is a simple property, then it is not
clear why it must supervene on other properties.
But on Scanlon’s view, goodness is not a simple indefinable property, but is
the property of having other properties that give us reason to care. As StrattonLake (2002: 15–16) points out, given this account of goodness, there is no
mystery why it must supervene on other properties. It must supervene on other
properties, because it is the property those properties have of providing reasons.
For Scanlon, reasons just are such properties of things as that they produce
pleasure or might lead to an improvement in our understanding of cancer.
The reasons to care about things are properties that can be described without
use of the concept of ‘practical reasons’. Goodness, in turn, is the non-natural
164
Philip Stratton-Lake and Brad Hooker
property those properties have of providing reasons (Parfit 1997: 124; Scanlon
1998: 57, 60, 61, 97; McNaughton and Rawling 2003: 30–3).
It may be argued that the buck-passing account of goodness does not solve
the problem, but merely moves it back. Goodness is defined in terms of
reasons, but the property of being a reason supervenes on other properties.
Blackburn could, therefore, insist that the puzzle remains: why must the
property of being a reason supervene on other properties?
We do not see the difficulty here. It is a conceptual truth that if you have a
reason to care about A, then there must be something that provides that reason.
This something is what the reason supervenes on. If the reason you have to go
for a walk is that it will be pleasant, then this reason supervenes on the pleasantness of the walk. If the reason you have to cycle to work is that it will keep
you fit, then this reason supervenes on the fact that cycling will keep you fit.
Reasons supervene on other properties because there must be something that
provides the reason. There is nothing mysterious about this, nothing here that
stands in need of further explanation.
One might ask why reasons supervene on one property rather than another.
The question might be why, for example, the fact that you have promised to
feed the neighbours’ cat while they are away gives you a reason to do so, whereas
the fact that grass is green does not. This is not, however, a metaethical question, but a normative one. The answer will depend on normative principles
about what people should do and why.
Here is a related argument in favour of Scanlon’s account over the Moorean
one. Scanlon holds that the openness of the open question argument does not
stem from the supposed fact that if goodness were the same as some natural
property this would be obvious to us. He thinks the openness of the open
question argument stems from the fact that judgements about whether A is
good express practical conclusions about reasons we have to care about A. The
open feel is dispelled once we judge that A is good because in judging that A is
good we judge that we have reason to care about A.
In contrast, the Moorean account does not hold that we judge we have
reason to care about A insofar as we judge that A is good. According to the
Moorean account, the goodness of something and the practical reasons that
the goodness supplies are different things. So Mooreans must hold that to
judge that we have reason to care about A is to make a further judgement over
and above the judgement that it is good. They must hold that, in this respect,
judging that A is good is just like judging that it has some natural, reason-giving
property. We may think that certain natural properties as well as goodness are
reason-giving; but, just as we do not explicitly judge that we have reason to
care about A simply insofar as we judge that A has some natural property, so, on
Scanlon versus Moore on Goodness
165
the Moorean account, to judge that something is good is not yet to conclude
that we have reason to care about it.
In short, according to Scanlon, (a) there is openness between judging that A
has some natural property and judging that A is good, (b) there is openness
between judging that A has some natural property and judging that there is
reason to care about A, but (c) there is no open feel between judging that A is
good and judging that there is reason to care about A. According to the
Moorean view, again (a) there is openness between judging that A has some
natural property and judging that A is good, and again (b) there is openness
between judging that A has some natural property and judging that there is
reason to care about A, and (c*) there is an open feel between judging that A is
good and judging that there is reason to care about A. If we think (c) rather than
(c*) is correct, then we have to line up with Scanlon.
6. The open question again
If the buck-passing account of goodness is right, then to be good is to have
properties that provide reasons to care. So, if the buck-passing account is
right, then not only does goodness entail reasons to care shared by the agent
and others, but also, conversely, reasons to care shared by the agent and others
entail goodness. The buck-passing account’s commitment to the entailment
of goodness from reasons to care makes the buck-passing account vulnerable
to a series of objections. These objections insist that we can coherently judge
that we and others have reason to care about something yet deny that this
thing is good.
There are various forms this sort of objection can take, and we cannot address
them all here.⁶ But one way of capturing the general idea behind such objections can be expressed as an open question objection to Scanlon’s buck-passing
account of goodness. It may be insisted that the buck-passing analysis is false
because the question ‘Is something that has properties that provide reasons for
the agent and others to care about it good?’ has an open feel. The Moorean
argument that led Scanlon to the buck-passing account of goodness may,
therefore, provide a strong reason to reject it.
⁶ A particularly forceful form of this objection is put forward by Rabinowicz and Rønnow-Rasmussen
(2004). They consider the possibility of an evil demon who demands of us, under threat of severe punishment, that we admire him precisely on account of his determination to punish us if we don’t. Here we clearly
have reason to admire the demon, and to admire him on account of his maliciousness. Nonetheless, his maliciousness is in no way good. Whether the buck-passing account of goodness can rebut this objection is not
something we have space to consider here—but see Stratton-Lake (2005).
166
Philip Stratton-Lake and Brad Hooker
Is it an open question whether something that has properties that provide
reasons for the agent and others to care about it is good? If there is openness to
the question, it cannot be normative openness. For once we accept that something has properties that provide reasons to care about it, the question of
whether there is reason to care about it is clearly closed. So if Scanlon’s normative explanation of the open feel of the open question is the best account of
that question, then his account of goodness cannot be vulnerable to the open
question argument. Hence, if his account of goodness is vulnerable to an open
question, the question must be open in the sense that his definition either
is non-obvious and debatable or leaves open the possibility of motivational
indifference.
If it is open whether something that has properties that provide reasons to
care about it is good, and if this openness is merely a consequence of Scanlon’s
definition of goodness being non-obvious and debatable, then it is not clear
that the openness raises difficulties for Scanlon. For, insofar as this definition
of goodness is interesting and informative, one would hope that it is, prima
facie, non-obvious and debatable. We accept that, if Scanlon’s account of
‘good’ is an accurate account of what we mean when we judge that something
is good, the question ‘Is something that has properties that provide reasons to
care about it good?’ is ultimately closed. But it strikes us as unsurprising
that this question should initially feel open in the sense that the answer is not
obvious.
Is this question motivationally open? The motivational explanation of
openness is premised on the view that to judge that something is good is per se
to have a directed (motivational) attitude towards that thing. This is because
judgements about goodness are essentially practical judgements, and on this
view, practicality is understood motivationally. But judgements about what
we have reason to do and feel are also essentially practical judgements. So if the
essentially practical nature of evaluative judgements is to be understood
motivationally, then the practical nature of judgements about what we have
reason to do and feel is presumably to be understood motivationally. This
would mean that to judge that we have reason to respond in certain positive
ways towards something would per se involve having a positive attitude towards
that thing. But then the question ‘Is something that has properties that provide
reasons to care about it good?’ would not be motivationally open. Our view is,
then, that someone who thinks that Moore’s open question argument is best
understood in motivational terms would not expect there to be an open feel to
the question whether something that has properties providing reasons to care
about it is good.
Scanlon versus Moore on Goodness
167
We do not, therefore, think that Scanlon should be troubled by an open
question objection to the buck-passing account of goodness. The only sense in
which it is likely to be open is in the sense that the definition is non-obvious
and thus debatable. But this feeling of openness is to be expected from an
interesting and informative account of goodness. If this feeling persists after
reflection, this would cast doubt on the analysis of goodness in terms of
reasons. But if the analysis survives such reflection—which it will have to do if
it is a plausible analysis—this feeling of openness would dissolve.
References
Blackburn, S. (1971). ‘Moral Realism’, in J. Casey (ed.), Morality and Moral
Reasoning. London: Methuen.
—— (1984). Spreading the Word. Oxford: Oxford University Press.
—— (1985). ‘Supervenience Revisited’, in I. Hacking (ed.), Exercises in Analysis.
Cambridge: Cambridge University Press. Reprinted in Blackburn’s Essays in QuasiRealism. New York: Oxford University Press, 1993: 130–48.
—— (1998). Ruling Passions. Oxford: Oxford University Press.
Copp, D. (2003). ‘Why Naturalism?’ Ethical Theory and Moral Practice, 6: 179–200.
Crisp, R. (2000). ‘Contractualism and the Good’ (review article of T. M. Scanlon,
What We Owe to Each Other), Philosophical Books, 41: 235–47.
Dancy, J. (2000). ‘Should We Pass the Buck?’, in A. O’Hear (ed.), The Good, the True
and the Beautiful. Cambridge: Cambridge University Press: 159–73.
Frankena, W. (1939). ‘The Naturalistic Fallacy’, Mind, 48: 464–77.
McNaughton, D., and Rawling, P. (2003). ‘Naturalism and Normativity’,
Proceedings of the Aristotelian Society Supplementary Volume, 77: 23–45.
Moore, G. E. (1993). Principia Ethica, 2nd edn by Thomas Baldwin. Cambridge:
Cambridge University Press. Original edition published in 1903.
Parfit, D. (1997). ‘Reason and Motivation’, Proceedings of the Aristotelian Society
Supplementary Volume, 71: 99–130.
Rabinowicz, W., and Rønnow-Rasmussen, T. (2004). ‘The Strike of the Demon:
On Fitting Pro-attitudes and Value’, Ethics, 114: 391–423.
Ross, W. D. (1930). The Right and the Good. Oxford: Clarendon Press. Reprinted with
an introduction and bibliography by Philip Stratton-Lake. Oxford University Press,
2002.
—— (1939). Foundations of Ethics. Oxford: Clarendon Press.
Scanlon, T. (1998). What We Owe to Each Other. Cambridge, Mass.: Harvard
University Press.
168
Philip Stratton-Lake and Brad Hooker
Stratton-Lake, P. (2002). ‘Introduction’, in P. Stratton-Lake (ed.), Ethical
Intuitionism. Oxford: Clarendon Press: 1–28.
—— (2005). ‘How to Deal with Evil Demons: Comment on Rabinowicz and
Rønnow-Rasmussen’, Ethics, 115: 788–98.
Sturgeon, N. (2003). ‘Moore on Ethical Naturalism’, Ethics, 113: 528–56.
8
Opening Questions, Following Rules
Paul Bloomfield
There are two types of assessment of G. E. Moore’s intellectual character by
Ludwig Wittgenstein recorded in Norman Malcolm’s Memoir of Wittgenstein.
First is that Moore was ‘deep’, honest and serious; at one point Wittgenstein
‘observed that if one were trying to find exactly the right words to express a fine
distinction of thought, Moore was absolutely the best person to consult’ (1958:
67). The second is that
Wittgenstein once remarked that what Moore primarily did, as a philosopher, was to
‘destroy premature solutions’ of philosophical problems. . . . But he added that he did
not believe Moore would recognize a correct solution if he were presented with one. (66)
Malcolm writes that Wittgenstein thought Bertrand Russell ‘was extremely
“bright” is how he put it: Moore, in comparison, was less so’ (68). If we put all
this together we get the idea that Moore was ‘deep but not too bright’ and from
anyone but Wittgenstein, this would certainly be ‘Damning With Faint
Praise’. Regardless of any possible insult to Moore, however, Wittgenstein may
have been right.
At least in regard to the open question argument, it seems clear that Moore
himself was not at all clear about what he was on to. Perhaps everyone agrees
that he was on to something and that it was something both deep and important. Exactly what it is, however, is still a bit of a mystery. The problem arises
when we note that we have defined ‘good’ in such a way that we call all tokens
of type X we’ve encountered in the past ‘good’ and here in the present we are
encountering another token X. Why is it not a foregone conclusion that we
think this new X is good? Why does there always seem to be an open question
with regard to the goodness of each instance of X we encounter? Why do all
putative definitions of ‘good’ leave this lingering doubt? If we believed some
particular definition of ‘good’ was the true one, then it would be ‘true by
I’d like to thank Donald Baxter, Terry Horgan, Len Krimmerman, Michael P. Lynch, Mark Timmons, and
David Schmidtz for their helpful comments and discussion.
170
Paul Bloomfield
definition’ that an item which satisfied the criteria of the definition is ‘good’.
But instead of an ‘open and shut case’, at each new possible predication of the
word ‘good’, we are presented with an open question about whether the item
in question is, in fact, ‘good’.
Such considerations very well might lead us to conclude that ‘good’ is
indefinable, as they obviously led Moore. Even if we agree with Moore about
this, one might nevertheless have a lingering doubt about whether indefinability can really explain what is going on in the open question argument.
Have we ever really gotten a satisfactory account of what keeps the open
question open?
The canonical interpretation of the open question argument is that there is
a peculiar form of normativity accompanying the use of the word ‘good’ and
this normativity keeps us from definitively closing our open questions; there is
something peculiar about the way the predication of ‘goodness’ guides action.
Goodness comes with some special ‘normative force’ which necessarily incites
(defeasible) motivation. Because such force cannot be captured by a definition,
all putative definitions of ‘goodness’ strike us as incomplete and we get the
open question. This interpretation has had a large influence on the course of
metaethics in the twentieth century. Darwall, Gibbard, and Railton wrote in
their 1992 Philosophical Review ‘Fin de Siecle’ piece that non-cognitivism is the
‘real historical beneficiary of the open question argument’ (1992: 119).
Indeed, it seems fair to say that the non-realist family of theory that includes
non-cognitivism, emotivism, prescriptivism, quasi-realism, norm-expressivism,
etc. was the dominant metaethical family of the last century. Given the traditional interpretation presented briefly above, it is a short step to inferring that
it is, in fact, the pro-attitude attending predications of ‘good’ which keeps the
definition of ‘good’ open: we can always imagine ourselves not being motivated
by some item, or a token of a type of item, which we called ‘good’ in the past.
The normative force of ‘goodness’ must be all in the head, for if it were ‘out
there’ in the world, it would be ontologically ‘queer’. If goodness has to be
accompanied by a motivationally charged, psychological state of mind, then
‘goodness’ cannot be defined in terms of some mind-independent, external
entities, as moral realists, like Moore, thought. In this way, moral non-realism
found perhaps its strongest support in Moore’s open question argument.
If, however, there were good reason for thinking that the sort of normativity
which keeps the open question open is not something special (queer) about
‘good’ (or ‘right’ or their cognates), if what we have is actually not a sui generis
form of moral normativity, but rather just a sort of normativity which we find
in another area of philosophy, then the consequences for metaethics are apparent: the dominant metaethical theory loses perhaps its strongest motivation for
Opening Questions, Following Rules
171
acceptance.¹ This is the idea to be explored below. In particular, the conclusion
is that the normativity which keeps Moore’s ethical open question open is the
same sort of normativity which is driving Wittgenstein’s rule following considerations.² The main thesis of this essay is that there is no peculiar or queer sui
generis form of normativity which infects the meaning of the word ‘good’;
whatever it is that keeps the definition of ‘good’ open also keeps the definitions
of other words, like ‘plus’, ‘mass’, and ‘triangle’ open. In general, we simply cannot be sure that our words are being used now in a way that is consistent with
the way we used them up until now.
Regardless of whether or not this is a genuine problem or a pseudo-problem,
the point is that there is only one problem here, not one for ethics and a different one for everything else. The questions ‘how ought I to use this word?’, ‘what
ought I to believe in this situation?’, and ‘what ought I to do now?’ stem from
semantics, epistemology, and morality respectively but the underlying problem of normativity is the same in all: how can features of the world establish
conditions under which it makes sense for us to say that there are ways we
ought to conduct ourselves (with regard to our actions, our speech, or our
beliefs) and other ways which ought not to be followed? Even if normativity is
a mystery, even if it is somehow in the end rightly called ‘queer’, still, there is no
need to multiply mysteries beyond what is required and, a fortiori, there is no
reason to think that the normativity attaching to how we ought to behave
morally is different than that attaching to how we ought to speak semantically
or what we ought to believe epistemologically. If this is true, then the metaethical situation is far different than what is required for metaethical expressivism
(et al.) to be supported in the way that it is typically taken to be supported.
If one were to reason from the open question argument to expressivism
when engaging metaethics, then, assuming the argument below is sound,
¹ A comment late in production has shown me that just a few sentences here can ward off what amounts
to a very serious objection. Some may read this paper and think I am defending the idea that there is no
distinctive moral normativity but that I am trying to reduce moral normativity to semantic normativity, or
perhaps eliminate the former, supplanting it with the latter. This is mistaken. I do think there are general features of semantic, epistemic, and moral normativity such that each of these are a form of normativity, a satisfactory general account of which is still in the offing. (For my own attempt to begin such a theory, see
Bloomfield 2001: 143–52.) The limited conclusion of this paper, which concerns Moore’s open question
argument and the meaning of the word ‘good’, is only that the meanings of moral terms do not carry with
them any queer sort of normativity beyond the semantic normativity carried by the meanings of other words.
² This is a possibility which Darwall, Gibbard, and Railton (1992) acknowledge but do not pursue in
footnote 5 of their paper.
One who supports the traditional interpretation of the open question argument might reason as follows:
if the open question argument supports non-realism in other areas of philosophy beyond metaethics, then
so be it and non-realism should thereby spread into those areas. This may be the case, but then it must be
acknowledged that moral realists have no special burden to carry in defending their metaethical theory.
What we have then is a general metaphysical debate about realism and we have left metaethics per se behind.
172
Paul Bloomfield
consistency would demand the same reasoning applies when engaging in
semantics and epistemology. Since expressivism is not a going general theory of
semantics or epistemology, one should seriously doubt the reasoning that leads
to it in metaethics.
Obviously, these issues are very complicated. The crux of the matter can be
driven in upon first by clearing away surrounding red herrings that might
otherwise create a variety of possible conflations; after this the workings of
Moore’s open question argument and Wittgenstein’s rule following considerations may be more directly inspected. So, first, this means distinguishing the
open question argument from the misbegotten naturalistic fallacy as well as
from the ‘paradox of analysis’; it also means distinguishing the rule following
considerations from the argument against private language as well as from
concerns about family resemblance. Finally, and prior to all this, a large caveat
is required. The aim of this essay is admittedly polemical and as such Moore
and Wittgenstein are treated as historical stalking horses and not philosophical
saints: the goal is not to present new scholarship that sheds light on a solution
to the open question argument or the rule following considerations as traditionally construed, nor is it to shed light on the philosophical methods of either
Moore or Wittgenstein. Undoubtedly, below certain points of philosophical
development are emphasized and others are completely ignored. The choices
are driven by the interests of philosophy (particularly of metaethics) and not of
scholarship or hagiography. The suggestion is that, given the improved
perspective of hindsight on these particular issues, the best way to understand
what both Moore and Wittgenstein were up to is to see them as pursuing the
same set of issues, clustered around the idea of semantic normativity, though
both pursued them through different questions and with obviously different
methodologies and philosophical temperaments.
We can begin with ‘naturalistic fallacy’ and dispatch with it quickly enough.
All that is truly required for this is to cite Frankena’s ‘The Naturalistic Fallacy’
(1939/1986), but perhaps just a few more words here can sufficiently sum up
the situation. In his Principia, Moore, of course, recognized that the ‘naturalistic fallacy’ never really had anything special to do with naturalism, as he applied
it to metaphysical, as well as to natural, definitions of ‘good’. We learn from
C. Lewy (1964) that as early as 1921, Moore saw that there was no real fallacy
involved, insofar as fallacies require inferences and Moore thought that those
who committed the ‘naturalistic fallacy’ were merely confused. About what
were they confused? It seems that each of them thought that one of a variety of
possible definitions of ‘good’ were complete and correct. How do we know
they were all confused? Because the open question argument shows us that any
such definition is not actually complete. The ‘naturalistic fallacy’ is best
Opening Questions, Following Rules
173
thought of as a substantial philosophical thesis, the content of which is that the
open question argument shows us that all definitions of ‘good’ are at best
incomplete. Since all the work here is being done by the open question argument, our attention can be focused upon it and we may leave the naturalistic
fallacy (almost) behind us.
The ‘almost’ is because it will help us to keep it in view while attending to the
paradox of analysis. In Principia’s discussion of the naturalistic fallacy, Moore
uses language which is very similar to the language found in discussions of the
paradox of analysis. For example ‘ “That we should desire to desire A is good”
is not merely equivalent to “that A should be good is good” ’ (§ 13). In fact,
Lewy (1964) gives one interpretation of the naturalistic fallacy that takes it to
be ‘a particular instance’ of the paradox of analysis. As Lewy notes, this is a
mistake. The paradox of analysis asks us how any definition or analysis can be
both complete and informative; quoting Alonzo Church, ‘This would seem to
reduce [a successful] analysis to something trivial and uninformative’ (1946:
132). The important point for our work here is that the paradox of analysis
obviously assumes that the definition under consideration is complete and this
is explicitly what the open question argument is supposed to show is impossible for any definition of ‘good’. Insofar as ‘good’ is concerned, the paradox of
analysis assumes and the open question argument denies the possibility of a
complete definition of ‘good’ and as such we can leave the paradox of analysis
behind.
The relationship between Wittgenstein’s rule following considerations and
his argument against private language is more complicated. The traditional
view is that the former are a step toward establishing the latter, since the former
immediately precedes the latter, which begins at § 243 within Philosophical
Investigations. Others (e.g. Kripke 1982) think the rule following considerations are what constitute the argument against private language, the conclusion
of which is thus given in § 202;³ on this view, what the tradition takes as the
private language argument is actually an application of that argument to
sensations which give the prima facie appearance of being a counterexample to
the private language argument. Our concern here, however, is not with private
³ In § 202, Wittgenstein says, ‘Hence it is not possible to obey a rule “privately”: otherwise thinking one
was obeying a rule would be the same thing as obeying it.’ Note however, contra Kripke, that in § 199 (on
the same page as § 202) he says, ‘To obey a rule, to make a report, to give an order, to play a game of chess are
customs (uses, institutions)’ where institutions are surely public; at § 206 he says, ‘Following a rule is analogous to obeying an order. We are trained to do so; we react to an order in a particular way. But what if one
person reacts in one way and another in another to the order and the training? Which one is right?’; here the
concern is interpersonal. And in § 204, the concern is with rules which no one ever follows. In general,
Wittgenstein does not seem to be particularly, much less primarily, concerned with private rules at this point
in his dialectic.
174
Paul Bloomfield
language at all. This can be seen because the rule following considerations
present a set of problems for public languages just as much as for private
language: the question ‘what is it for us to all be now following the same rule?’
arises through the rule following considerations with as much force, and
perhaps more, than the question ‘what is it for me to be following now the same
rule I followed yesterday?’ The intrapersonal diachronic difference between
me today and me yesterday is identical from a formal point of view to a synchronic interpersonal difference. Problems regarding the meanings of words
are not purely private problems, but are public language problems as well. So,
even if we agree with Kripke in thinking that the conclusion to the private
language argument comes at § 202, we can still see this as a special application
of a more general problem about following rules. For our present purposes, we
are trying to isolate whatever it is that is generating the problems we find in
understanding how we follow rules. Since there is no a priori reason to think
that issues concerning privacy are at the bottom of the rule following considerations, we can set aside problems concerning private language in our consideration
of the rule following considerations.
It might appear that Wittgenstein’s notion of family resemblance is driving
at least part of the rule following considerations.⁴ If one thought that rules
needed necessary and sufficient conditions to be fulfilled in order for the rule
to apply, then pointing out the fact that we apply some rules which do not
admit of necessary and sufficient conditions might show us a way into problems concerning our following rules. Admittedly, the application of family
resemblance concepts is a bit mysterious; we find ourselves saying odd things
like ‘I can’t quite put my finger on it, but something about you reminds me of
your parents.’ But again, we do not want to supplant a general problem with a
particular instance of it (indeed this is contrary to the spirit of the essay) and
that is the mistake we would make were we to think that family resemblance
drives the rule following considerations. The reason for this is that the rule
following considerations may apply to rules driven by family resemblance
⁴ Wittgenstein did, at least at one point, think that some of the problems in understanding ‘good’ were
due to something like family resemblance. Quoting from lecture notes written by Moore in 1931 of lectures
given by Wittgenstein: ‘And of the word “good” he [Wittgenstein] said similarly [as he said of “beautiful”]
that each different way in which one person, A, can convince another, B, that so-and-so is ‘good’ fixes the
meaning in which ‘good’ is used in that discussion— ‘fixes the grammar of that discussion’; but that there
will be ‘gradual transitions’, from one of these meanings to another, “which take the place of something in
common” ’ (Wittgenstein 1993: 104). ‘Good’ may be a vague predicate, but I think it is a mistake to think
that the mysteries surrounding ‘good’ are to be located by attending to its vagueness per se; we will not find a
solution to those mysteries by adopting any particular theory of vagueness or by solving the Sorities paradox.
Maybe I am wrong and the troubles surrounding ‘good’ are due to a family resemblance problem; even so,
remember that the thesis of the paper is to argue that ‘good’ presents no sui generis problems. If I am wrong
and the mystery of ‘good’ is to be located in its character as a family resemblance notion, this still does not
demonstrate that ‘good’ presents any sui generis problems.
Opening Questions, Following Rules
175
concepts, like ‘game’, but they also apply to rules which can be made as precise
as anyone could like: Kripke’s contribution to the dialectic is that he brought
the generality of the problem to the forefront. It arises for ‘plus’ just as much as
it does for ‘game’. The problem is located in how it is that we succeed in
following rules, and whether or not the rules are precise or driven by family
resemblance, the same problem arises. Granted, there may be extra complications with family resemblance concepts but these are not needed to generate
the rule following considerations per se, and as such we can leave behind these
complications and focus on the rule following considerations directly.
Having cleared the board of some possible red herrings, let us now turn forthrightly to the open question argument and the rule following considerations
respectively. The open question argument comes in § 13 of Principia Ethica as
the argument which, all by itself, is supposed to show that the meaning of ‘good’
is simple and indefinable. Moore gives us an argument by cases beginning with
the idea that the only options to the simplicity and indefinability of ‘good’ are
either that it means something complex or that it has no meaning at all. The
open question argument is used to show that both of these alternatives fail.
In the case in which ‘good’ denotes something complex, the argument goes
as follows. We are explicitly asked to apply such a complex definition of ‘good’
to a particular case, A, asking ‘Is A good?’, and we may say that ‘A is good’ if and
only if it is one of the kind of thing which falls under the extension of the complex definition of ‘good’ under consideration. So, working with Moore’s own
example, let’s use the complex notion that good things are the things which we
‘desire to desire’ and then we can answer the question ‘Is A good?’ by noting
that it is if and only if A is one of the things which we desire to desire. So far, so
good. We can pursue the matter further, however, by asking a second question
‘Is it good to desire to desire A?’ and when we do so, we find out that the fact
that A is one of the things which we desire to desire does not settle this second
question. In fact, the information we are asking about in the second question
is the same as the information we were asking about in the first question ‘Is A
good?’: having the definition and applying it to a particular case does not automatically close the question about whether or not that particular case falls
under the extension of the original predicate. As Moore says,
It may indeed be true that what we desire to desire is always also good; perhaps even the
converse may be true: but it is very doubtful whether this is the case, and the mere fact
that we understand very well what is meant by doubting it, shews clearly that we have
two different notions before our mind. (§ 13, 1902/1988:16)
The empirical fact that every individual thing which in the past we have desired
to desire has been good is insufficient to show us that ‘good’ denotes only those
176
Paul Bloomfield
things which we desire to desire. At the most, what can be shown is that the
extension of ‘good’ is the same as the extension of what we ‘desire to desire’, but
even this would not allow us to define ‘good’ in terms of being something
which we desire to desire. It does not show us that it is impossible for there to
be something which we desire to desire but which in fact is not good at all, nor
that it is impossible for there to be something which deserves to be called ‘good’
even though it is not something which we desire to desire. Either of these two
possibilities represent an ‘open question’ by showing us that it is not ‘true by
definition’ nor by ‘the meanings of the words’ that what it is to be good is
identical to what it is to be what we desire to desire.
This is supposed to show us that any complex definition of ‘good’ will not
work, though it is not clear what role complexity plays in the argument.
Leaving this issue aside for the moment, the other alternative is that ‘good’ fails
to denote anything at all, that it is meaningless, and Moore tries to use the open
question argument, once again, to show us that this alternative fails. The argument now goes like this: if we think that it is true by definition that ‘pleasure is
good’, then we have only one notion before our mind and if this is the case then
it cannot make sense to go on and ask ‘Is pleasure after all good?’ (1902/1988:
16). Moore takes it as obvious, and perhaps it is, that we all understand the
question ‘Is this good?’ and that this is a distinct question from ‘Is this pleasurable or desirable or approved?’ So, once again, given an identification of what
is ‘good’ with what is, say, ‘pleasurable’, we are still left with the open question
embodied by meaningfully asking a question of the following form: ‘We know
it is pleasurable, but is it good?’ So because the question ‘Is this thing, A, good?’
is meaningful and distinct from asking about A if it is anything else other than
‘good’, we can conclude that ‘good’ is not meaningless.
Does this actually show that ‘good’ is not meaningless or that it fails to
denote anything? This seems dubious at best because the argument as it stands
seems to rest on the assumption that if what is ‘good’ is in fact identical to (‘⫽’),
e.g., what is ‘pleasurable’, then either ‘good’ is meaningless or it fails to denote
anything, and this seems like a false assumption. From the premise identifying
what is ‘good’ with what is ‘pleasurable’, one cannot conclude that ‘good’ is
meaningless or fails to denote; and yet this is the inference which Moore seems
to make in this second application of the open question argument. Contra
Moore, it seems that if the premise is true then ‘good’ does not fail to denote
anything, rather it denotes all the pleasurable things in the world.⁵
⁵ To say ‘good’ is to mean ‘pleasure’ but this may not be the same as intending to use ‘good’ to mean ‘pleasure’.
As Church (1946) notes in regard to the paradox of analysis, Frege’s distinction between meaning (sense) and
denotation (reference) is surely lurking in the background here. ‘Good’ may be the sense while ‘pleasure’ is
denoted.
Opening Questions, Following Rules
177
Settling this particular question is not necessary either, since the idea that
‘good’ is meaningless is not plausible in the first place. All parties to present
debate in metaethics seem to agree that ‘good’ is not meaningless. (Even if
expressivism is right and ‘good’ can no more be defined than ‘hooray’, we
would not thereby conclude that ‘good’ is meaningless.) If we can assume that
‘good’ is meaningful, we can nevertheless learn something from the second
application of the open question argument: no possible definition of ‘good’
could be the true or correct one if the only reason for accepting it is through an
appeal to the meaning of the word. This may sound trivial, but in fact it is not.
It may make more sense to say it this way: we cannot understand the meaning
of ‘good’ through any possible definition. What we learn, finally, from the
open question argument is that any definition of ‘good’ is open to question
because it can never be a foregone conclusion that the definition, as it stands,
works for every present or future application of the word ‘good’: for any new
case, we can always question whether or not the item is good even after we note
that it fulfills the criteria given in the definition we have supposedly accepted.
Our past experience with using the word in a particular way does not tell us that
we ought to continue to apply the word to things in the world in that same way
in the present or in the future.
The open question argument may therefore be summed up as follows. Any
putative definition of the word ‘good’ shows itself to be incomplete through its
failure in the practical application of ‘good’ to a particular case, P; such that,
when we encounter P we cannot automatically conclude ‘P is good’ even
though it may be apparent that P fulfills the criteria laid out in the putative
definition. The doubt always remains that ‘P is good’ is false even though it
satisfies the semantic rule constituted by that definition. This doubt, this open
question, cannot be closed by appeal to past things which have satisfied the definition nor by the meaning of ‘good’ as given by the definition. The semantic
rule for ‘good’ must be seen as, in this sense, open-ended. According to Moore,
the way we tell whether or not ‘P is good’ is through ‘intuition’; no definition
of ‘good’ can be proved true.⁶ Therefore, regardless of the particulars of the
definition, we can conclude that it is incomplete, that it has left something out,
that the meaning of ‘good’ is not fully captured by the terms specified in the
definition.
⁶ Moore writes, ‘Again, I would wish it observed that, when I call such propositions “Intuitions,” I mean
merely to assert that they are incapable of proof; I imply nothing whatever as to the manner or origins of our
cognition of them. Still less do I imply (as most Intuitionists have done) that any proposition whatever is
true, because we cognise it in a particular way or by the exercise of any particular faculty: I hold, on the contrary, that in every way in which it is possible to cognise a true proposition, it is also possible to cognise a false
proposition’ (1902/1988: preface, p. x). This is a form of the argument for moral realism from error which
merits separate attention.
178
Paul Bloomfield
What is it that is left out? What is it that keeps the open question open?
Moore wants to answer the question by pointing to the simplicity of goodness.
But while simplicity might rule out the possibility of analysis, it is far from
obvious why we cannot trust ourselves to some degree to be able to sometimes
recognize goodness when we see it, especially since we all seem to be able to use
the word ‘good’ in a variety of ways that indicate that we know what it means.
The problem is not in affirming or denying the possibility of giving a full analysis of goodness or a definition of ‘good’. The problem is, as just mentioned, a
practical one regarding how it is that we can properly use a word without being
able to ever be sure that we are using it correctly. The problem is not over
whether or not the rule governing the use of the word ‘good’ ought to engage
in the face of something either simple or complex. The problem is in understanding how the meaning of a word can be governed by a rule. Simplicity and
complexity are beside the point. While Moore is to be credited for pointing out
the open question and for explicating some of its implications, he does not
satisfactorily explain why the open question stays open.
Turning now to Wittgenstein, the rule following considerations are not so
neatly compacted in a page or two as the open question argument is in
Principia; arguably they run all the way through Philosophical Investigations
insofar as the idea of a ‘language game’ (first mentioned in § 7) is perhaps the
central theme of the book and rules constitute the formal aspects of all games,
including logic. If people don’t follow the rules then there will come a point at
which they have stopped playing the game. The giving of definitions and the
meanings of words can rightly be considered the central examples of rules in
the book. In § 81 and § 82, we find the following:
All this [concerning the idea of Ramsey’s that logic is a ‘normative science’], however,
can only appear in the right light when one has attained greater clarity about the concepts of understanding, meaning, and thinking. For it will then also become clear that
what can lead us to think that if anyone utters a sentence and means or understands it
he is operating a calculus according to definite rules.
82. What do I call ‘the rule by which he proceeds’?—The hypothesis that satisfactorily describes his use of words, which we observe;—But what if observation does not
enable us to see any clear rule, and the question brings none to light?—For he did
indeed give me a definition when I asked him what he understood by ‘N’, but he was
prepared to withdraw and alter it.—So how am I to determine the rule according to
which he is playing? He does not know it himself . . .
In § 84, Wittgenstein asks what a game looks like whose rules ‘never let a doubt
creep in’, and in § 85 he says, ‘A rule stands there like a sign-post’; but it does not
tell us to follow in the direction of the finger rather than in the direction of the
wrist. Even chalk marks on the ground can be interpreted in different ways
Opening Questions, Following Rules
179
(§ 85). Doubt always remains about whether or not we are following the rule
that we are intending to follow.
In terms of structure, a plausible reading may take the ‘rule following
considerations’ to begin at § 143 where we are asked to consider writing down
a ‘series of signs according to a certain formation rule’. At § 147, Wittgenstein
notes that one can ‘know the application of the rule of the [algebraic] series
quite apart from remembering actual applications to particular numbers’. At
§ 151, we are introduced to the idea that any finite amount of data underdetermines a rule. From § 156 to § 171, we get a digression on reading, and at
§ 179 we return to our case at § 151. This is followed by the example of trying
to teach someone ‘⫹2’ in such a way that the pupil will not ‘change’ to ‘⫹4’ (a
different rule) after reaching an arbitrary point, like 1,000. The notion of
‘queerness’ makes its appearance at § 194 and the problem is summed up in full
force for the first time at § 201:
This was our paradox: no course of action could be determined by a rule, because every
course of action can be made out to accord with the rule. The answer was: if everything
can be made out to accord with the rule, then it can also be made out to conflict with
it. And so there would be neither accord nor conflict there.
It can be seen that there is a misunderstanding here from the fact that in the course
of our argument we give one interpretation after another; as if each one contented us at
least for a moment, until we thought of yet another standing behind it. What this
shews is that there is a way of grasping a rule which is not an interpretation, but which
is exhibited in what we call ‘obeying the rule’ and ‘going against it’ in actual cases.
Immediately after this, Wittgenstein becomes occupied with the idea of
‘obeying a rule’ and keeps almost exclusively to this thought until § 243 when
he famously turns his attention to the particular problem of reidentifying ‘private’ sensations (again, according to a rule). After this, in the Investigations,
what we have learned about rules is somewhat in the background but must be
seen as constituting the bedrock of subsequent discussions of ‘meaning as use’,
the point or purpose of a game, the notion of ‘intention’, etc. One cannot hope
to capture the breadth or depth of Wittgenstein’s dialectic, but the present
discussion can carry on by attending to three points: overcoming the underdetermination of a rule by data, the supposed ‘queerness’ of thinking that
‘future development must in some way already be present in the act of grasping
the use and yet isn’t present’, and, finally, the idea of ‘obeying a rule’.
At § 151, we find a series of numbers: 1, 5, 11, 19, 29; and it is pointed out
to us that different mathematical functions can be used to describe the series: it
can be an ⫽ n2 ⫹ n ⫺ 1 or one could spot the series 4, 6, 8, 10 being added
to each successor. There is no fact of the matter about which function is the
function defining the series; the options are logically equivalent. If we are
180
Paul Bloomfield
looking for a determinate answer about what is really going on in that series of
numbers, the situation only worsens when we note that the options may
capture equally well all past data but nevertheless diverge in the very next case
or at any point in the future. When our pupil works in a way that seems to be
in accord with the rule ‘⫹2’ until 1,000 is reached and then carries on 1,004,
1,008, 1,012, we think that something has gone wrong and cannot help but to
wonder at what rule our pupil was following all along. The problem is not
that we cannot know the rule which our pupil was following; rather the
problem is that there may not have been any fact at all about what rule
our pupil was following. The reason for this is that we do not know how to
capture in a single application the entire meaning of a rule, and multiplying
applications does not help because the same incompleteness is present in each
new application.
‘The steps are really already taken even before I take them in writing or orally or in
thought.’ And it seemed as if they were in some unique way predetermined, anticipated—
as only the act of meaning can anticipate reality. (§ 188)
We do overcome this underdetermination of data: the fact that we are not
(yet) extinct shows that we are doing something right. We can learn rules which
help us to survive on our own and communicate with each other. We do succeed in playing games, we do seem to be able to follow rules regarding, say,
which side of the street to drive on, and the exceptions are aberrations which do
not last long, on pain of death. (We are perhaps touching on the essence of natural selection.) It would be wrong to say that all of human life or all of life in
general is a game; this would be demeaning. Life is the ‘real thing’; games are
not. Nevertheless, the formal relationship which human life bears to rules is the
same as the one which human games bear to rules.
We do overcome this underdetermination of data, but Wittgenstein is not
as clear as one could hope about how this is done. We ‘grasp the whole use of
the word in a flash’ (§ 191), in a way which determines our future usage while
not being determined by past usage.⁷ This may sound mysterious and
Wittgenstein is sensitive to this. He writes, ‘When does one have the
thought: the possible movements of a machine are already there in it in some
mysterious way?—Well, when one is doing philosophy’ (§ 194). Is this sort
of grasping rightly considered a ‘mystery’ something to be thought of as
‘queer’? Wittgenstein says it is only philosophers who will see it as queer.
Philosophers will say, ‘I don’t mean that what I do now (in grasping a sense)
determines the future use causally and as a matter of experience, but in that
⁷ § 138 is the earliest use of the idea that we ‘grasp’ the meaning of a word ‘in a flash’.
Opening Questions, Following Rules
181
queer way, the use itself is in some sense present’ (§ 195).⁸ And Wittgenstein
thinks that the philosophers are right except for their invocation of queerness. We think on one hand that ‘there is no doubt that we understand the
word, and on the other hand its meaning lies in its use’ (§ 197). And here
there is no special queerness, as long as we remember that to ‘understand a
language means to be master of a technique’ (§ 199). We can grasp meanings
and understand rules, we do it all the time, it is how we survive, and though
we may not as yet have a full explanation for the phenomenon, we need not
think that grasping the meaning of a word ‘in a flash’ or following a rule is ‘a
queer process’: amoeba behave in a manner closely akin to the way in which
we obey rules each time an amoeba encounters acidic water and swims in the
opposite direction. At some point, justifications for behavior end, our ‘spade
is turned’, and at that point we are left with nothing to say but, ‘This is simply what we do’ (§ 217).⁹
We are still left, however, with the question of how rules are followed.
Wittgenstein’s answer to this is that what we grasp when we grasp a rule or
understand the meaning of a word is not an interpretation of the rule but is
rather the grasp of a practice or a technique. We make a hypothesis about how
to proceed and then employ the hypothesis, all the while keeping an open mind
about the possible need to amend the hypothesis in the future.¹⁰ Wittgenstein
thinks that we can understand how we follow rules by considering the ‘analogy’
of it to how we ‘obey a rule’. We hold on to the idea that some actions ‘obey’ the
rule and others ‘go against it’, and so we may conclude that ‘to think one is
obeying a rule is not to obey a rule’ (§ 202). Thus, it is the normativity behind
the determinations of what obeys the rule and what goes against it that is the
crux of the project of figuring out the meanings of words. Importantly, this is a
problem of practical rationality, much like a chess problem is a problem of
practical rationality. The distinction between theory and technique dissolves.
Once we grasp the rule, we do not ‘choose’ our move; rather, we obey ‘blindly’
(§ 219). We carry on with our hypotheses until our experience trips us up.
⁸ ‘. . . When we do philosophy, we are like savages, primitive people, who hear the expressions of civilized
men, put a false interpretation on them, and then draw the queerest conclusions from it’ (§ 194).
⁹ I have changed here from Wittgenstein’s first person singular to first person plural. The original § 217
is ‘This is simply what I do.’ The quotation marks do not here indicate an quotation from the text, but are to
indicate what we might say in the described circumstance.
Those with more realist tendencies than Wittgenstein might think that the bedrock is found in some very
unmysterious ‘fact of the matter’ (‘if we eat those berries we die’) upon which our practices are founded or
without which our practices cannot continue. The bedrock may be the world when the game is to dig a ditch;
the bedrock may be our conventions when the game is proper table etiquette.
¹⁰ ‘82. What do I call “the rule by which he proceeds”?—The hypothesis that satisfactorily describes his
use of words . . .’
182
Paul Bloomfield
When we do get tripped up we give explanations to ‘remove or avert’ the
misunderstanding (§ 87).
Must I know whether I understand a word? Don’t I also sometimes imagine myself to
understand a word (as I may imagine I understand a calculation) and then realize that
I did not understand it? (‘I thought I knew what “relative” and “absolute” motion
meant, but I see that I don’t know.’). (footnote, following § 138)
The oblique reference to Einstein is helpful. The word ‘mass’ used to denote
a quantity which is invariant with regard to speed. After Einstein, the word
denoted a quantity which does vary with speed. We do not need to take sides in
the debate about Kuhnian incommensurability in order to consider the rule
constituting the meaning of ‘mass’ and wonder whether it always meant what
Einstein taught us it meant or whether he changed the rule so that the word
started to mean something different than it used to before him. The important
point for us is that we can conceive of Einstein as questioning the rule constituting the meaning of the word ‘mass’; Einstein opened up the question about
the meaning of the word ‘mass’ when everyone else had long thought that this
particular question was permanently closed. He forced us to question whether
or not we were going to continue to apply the word ‘mass’ as we thought we
were up until that point. Our pre-Einsteinian hypothesis regarding the meaning of ‘mass’ was shown to require adjustment in order to comport with new
possibilities. We learned that if we continued to apply the word under exactly
the same conditions in which we applied it in the past, we would sometimes ‘go
against’ the rule governing the meaning of the word. We need not bother here
with the distinction between retrospectively identifying unnoticed mistakes
and the deliberate changing of our ways to better ‘obey’ the rule. We need not
ask whether it was the same rule all along and Einstein helped us to get clear
about it or if he changed the rule.
What is crucial is that we see the question which Einstein opened with
regard to the definition of ‘mass’ as being the same as the open question which
Moore pointed out with regard to ‘good’.¹¹ The meaning of ‘mass’ may have
seemed more closed than the meaning of ‘good’ ever could be, but if we
remember the stricture to ‘question our assumptions’, then we will learn that
the meanings of perhaps all our words, and most certainly our more theoretically laden words, like ‘mass’ and ‘good’, always have an open question attached
to them. Their meanings can never be assumed to be certain in such a way that
their future applications are necessarily or logically determined by present
¹¹ Interestingly, Moore and Einstein were working on Principia and the Special Theory of Relativity,
respectively, at approximately the same time.
Opening Questions, Following Rules
183
usage. In Riemannian space, ‘triangles’ can be composed of three right angles.
Why should the meanings of our words be held sacred? We will continue to
carry on as we have, humanely plodding along, one step at a time, adjusting or
adapting as we need to, and, crucially, there is nothing ‘queer’ at all about it.
The proper question to be asking is not ‘how can the meaning of “good”
guide (or goad) behavior?’, nor even ‘how can the meaning of “justified” guide
my behavior?’ but rather ‘how can the meaning of any word, like “plus” or
“mass” or “triangle”, guide behavior?’ given that speaking and writing are quite
literally forms of behavior. To move to full generality, the question we really
ought to be asking is ‘how can anything in the world guide behavior?’ In what
sense is it possible for behavior to be guided? Guided by what? With what
authority? To what end?
Luckily, once again, these questions do not need determinate answers to
pursue the matter at hand. The goal here is not to figure out the ‘mystery’ of
normativity, but to show that the open question argument does not support
moral non-realism as most moral philosophers think, because what keeps the
open question open is no special sui generis type of normativity that attaches to
the words ‘good’ or ‘right’ or their moral cognates. The way to do this is to show
how the rule following considerations provide a broader explanation of the
open question argument allowing us to see the latter as a problem which does
not uniquely concern the word ‘good’ or morality but rather as an instance of
the general phenomenon or the problem of normativity. As Moore says in the
preface of Principia, the first question he is concerned with is ‘What kind of
things ought to exist for their own sakes?’; the second is ‘What kind of actions
ought we to perform?’ What we have formulated here is the problem of normativity in its fullest human generality. Moore pursues these problems through
investigating the nature of intrinsic value or the properties of the goodness of
objects and the rightness of actions. This would have to include not only ‘good
people’ but ‘good knives’ and ‘good chess players’, as well as proper or fitting or
appropriate speech as a form of ‘right’ action. With each fresh circumstance in
our lives, day by day, often moment by moment, we are confronted with the
question ‘what ought to be done here?’ and we cannot assume that we can
immediately gain our answer by looking to the past or unthinkingly applying
old rules. In almost all cases, we must inspect the circumstance anew to make
sure that we are responding to it appropriately; in all other cases we are, one
might say, on ‘autopilot’. How to sort out all these issues is the grand problem
of normativity and incorporates the programs of both Principia Ethica and
Philosophical Investigations.
Proving this in any systematic way may in fact be impossible. What can be
done is to show how, at a number of points, Wittgenstein’s concerns mirror
184
Paul Bloomfield
those found that naturally arise when considering the open question argument:
similarities between Moore’s ‘intuitions’ and Wittgenstein’s ‘grasping in a
flash’; that the ‘queerness’ of ‘good’ stands or falls with the ‘queerness’ of ‘mass’;
and finally that the only normativity needed to keep the open question open is
of the same sort which keeps all linguistic and epistemic behavior infused with
an ‘open question’. There is nothing which marks off behavior with regards to
morality as ‘special’ which does not also mark off behavior with regards to meaning and belief formation generally: all this is equally normative. Wittgenstein’s
concerns are Moore’s, only writ larger, deeper, and broader.
Both Moore and Wittgenstein focus on both the unprovable aspects of our
intuitions about ‘good’ and our grasping of rules, respectively. Moore says we
have intuitions about what counts as ‘good’ and, as noted above, what he
means by this is that our intuitions about what is ‘good’ are unprovable. For
Moore, the unprovability of the contents of intuitions does not prevent us from
having moral knowledge; such knowledge is simply not open to discursive
proof. For Wittgenstein, whatever it is that takes place in the grasping of a
meaning ‘in a flash’, it may be seen as a way to sum up the shift from the
Tractatus to the Investigations: life and language cannot be pictured by truth
tables; in the end there are no proofs to be had. If the rule following considerations have shown us anything, it must be that we cannot ever prove that we
understand the meaning of a word: no amount of using it properly proves
anything, given the fact that our next application of it may seem to others as
our pupil saying, ‘. . . 996, 998, 1,000, 1,004, 1,008 . . .’ seems to us.
We cannot know with any certainty that we understand; again, at bottom
our justifications run out and we say, ‘Well, that’s just how we’ve been doing
it . . .’ We may seem to be on firmer ground with respect to knowing the
meaning of other words than we are with ‘good’ but what Einstein shows with
regard to ‘mass’ or Riemann shows with regard to ‘triangle’, this terra firma may
be quite illusory. This lack of firm ground is a problem running all through
philosophy. In philosophy of math, there is still no real consensus on what a
number is. In philosophy of mind, we cannot rule out the possibility that
eliminative materialists about the mind are wrong and that, in fact, we have no
‘beliefs’ or ‘attitudes’. In epistemology, we cannot simply stipulate that epistemic skeptics are wrong for we must keep an open mind about what we say
we ‘know’, or what we are ‘justified’ in believing. In metaphysics, we need only
to gesture towards Berkeley on the meaning of ‘substance’ or Hume on the
meaning of ‘causality’. In more modern times, we learn that these problems go
beyond the meanings of terms which we typically think of as being ‘theory
laden’: Peter van Inwagan (1990), not to mention many quantum physicists,
tells us that the words ‘river’ and ‘mountain’ apply to things which are radically
Opening Questions, Following Rules
185
different than we typically think, and Derek Parfit’s (1984) thought experiments
about ‘identity’ tell us that it may not be predicated of particular cases in
obvious ways.
Are these words ‘action-guiding’ in any important sense? Well, of course
they are. The answers to questions of identity may determine our actions in any
number of circumstances, e.g. in cases of proposed euthanasia. Whether or not
‘allowing’ is as much of a ‘cause’ as ‘making’ has been at the center of many
important life and death debates. We need not come down on one side of any
of these debates to see that the meanings of the words under contention, the
rules according to which we may think, speak, write, and act, either already
have open questions attached to them or the philosophers who initiate the
debates are the ones who are doing the opening. The meaning of ‘good’ presents us with undoubtedly difficult problems, and problems whose solutions
do not admit of proof; but so far there are no reasons to think of these problems
as being particular to ethics or moral philosophy.
Under these circumstances, it is hard to see as mere coincidence Wittgenstein’s
(or his translator, Anscombe’s) use of the word ‘queer’ in the context of the rule
following considerations when this very word has subsequently played a large
role in the development of perhaps the most significant attack on the property
of ‘moral goodness’, as found, most famously, in John Mackie’s Ethics:
Inventing Right and Wrong (1977). Things which we call ‘good’ are supposed to
have some special power over us; such items are said to necessarily affect our
motivations when we recognize them. The implausibility of thinking that
ontologically real, ‘mind-independent’ properties could have such power over
us has become one of the leading reasons for rejecting moral realism, and, as
noted above, the ability to explain why we are motivated by the items we call
‘good’ has been perhaps the strongest argument in favor of non-realistic
metaethical theories like norm-expressivism and quasi-realism. But what is so
queer? Thinking or calling something ‘good’ means that we will react to it in a
way different than if we thought it ‘bad’ or thought it was neither ‘good’ nor
‘bad’. So we can conclude that how we think about things affects how we
behave toward them. There are rules according to which we act when we
recognize something as ‘repulsive’ and other rules when we find something
‘attractive’. Well, of course. And perhaps there is a mystery here, but there is no
special mystery about ‘good’ or ‘bad’ or their moral cognates: from this point
of view, all meaning has behavioral implications coming along with it. The
behavior of a physicist may be guided differently by thinking that ‘mass’ does
or does not vary with speed. Where is the special queerness about moral terms?
Well, thinking something is ‘good’ is often said to require us to be motivated
toward it in a particular way; we must feel positively about it. This is the
186
Paul Bloomfield
position often termed ‘internalism’. It would be decisive perhaps to secure a sui
generis queerness for the meaning of ‘good’ if there were no open question
about the possibility of being repulsed or at least feeling negatively about
something which is recognized as ‘good’. Internalism is not itself a ‘foregone
conclusion’ but a substantive thesis for which a philosopher must argue. There
is an open question whether internalism is true for at least the following reason.
Moore pointed out that an item’s fulfilling the criteria of a definition of
‘good’ does not all by itself allow us to conclude that this item is ‘good’; we must
have an intuition of its goodness. But there is the further open question being
pressed by someone not moved by internalism: why is it impossible to feel
anything but positive towards the things which we recognize as ‘good’? This
question remains open because we can imagine people receiving training
which makes them respond negatively to what they recognize as ‘good’. This
could be the opposite of the sort of training which occurs at the end of Burgess’s
A Clockwork Orange: there the subject is made to feel ill by images of violence
and other things which are recognizably ‘bad’ but which the subject felt positively towards prior to the training. Surely, it would be possible, though inhumanly malevolent and evil, to train people to be repulsed by things which they
recognize as ‘good’. In any case, it seems that we could train people to have the
same positively ‘queer’ motivations toward any old thing which we typically
think we have towards those items we call ‘good’. We can be just as motivated
and ‘queerly’ committed to solving an outstanding mathematical problem,
writing a string quartet, or making the longest chain of paper clips in the world,
as we are to being a friend, promoting world peace, or working towards the
elimination of hunger. If it is my goal to make the longest chain of paper clips
in the world, then I will necessarily, automatically be motivated by recognizing
any paper clips in the room, just as I may necessarily, automatically be motivated by helping my friends when I recognize that they are in need. There is
nothing special about the motivations which most frequently come when people recognize things as being ‘good’. We can be motivated positively or negatively by these things, just as we can be towards anything in the world. In any
case, our motivations are affected by how we think of things, how we think of
things affects how we behave toward those things, and if this is queer here then
there is equal queerness everywhere. But ‘ubiquitous queerness’ is either an
oxymoron or a contradiction in terms. As such, whatever it is, it ceases to be
queer. Wittgenstein is right: queerness is a philosopher’s invention.
It is a philosophical platitude nowadays that there are normative aspects to
both semantics and meaning as well as to epistemology and justification. Why
should we think that the normativity of morality is something special or even
primary? One can create a modus ponens or a modus tollens from how one
Opening Questions, Following Rules
187
answers this question. If one thinks that moral normativity is primary or
special, then one may be led to conclude that metaethics should ‘turn imperialistic’ and that we should see semantics or epistemology become subfields of
metaethics. In his Hempel Lecture (1992), Allan Gibbard suggests that semantic
normativity is a special case of moral normativity and in her book on virtue
epistemology, Virtues of Mind (1996), Linda Zagzebski suggests that epistemology should be enveloped as a subfield of metaethics. This does not put the cart
before the horse, because there is no a priori reason to think that epistemology
or language precedes morality any more than there is to think that morality
precedes epistemology or language.
If one answers the above question about the ‘specialness’ or ‘primacy’ of moral
normativity in the negative, then there is no reason to think that the normativity
which attaches to morality is sui generis, in a way that requires moral
non-realism to explain. There are skeptics about epistemology, meaning, and
morality; but there are very few expressivists in the field of epistemology, and
there are also few, if any, who think that all language deserves an expressivist’s
interpretation. Being a global non-cognitivist or expressivist would be a difficult
trick to pull off, and one would have to wonder at the motivation behind it.
We are human beings, members of Homo sapiens. We are agents who form
beliefs according to rules, communicate with our conspecifics according to
rules, and use rules generally to ‘guide’ almost all our behavior. We are concerned with how we ought to act, what we ought to believe, and what we ought
to say to each other. These are aspects of our existence which might be thought
to be central to, if not essential to, the ‘human condition’ and there is no way
around these practical problems. Still, obviously, we carry on. Most of us, most
of the time, are guided by what we ought to do, or at least we try to be guided
by what we ought to do, or at least we ought to be guided by what we ought to
do. The regulation of conduct falling under the provenance of morality is no
more nor less normative than the regulation of conduct falling under the
provenance of communication or epistemology. At the very least, the burden
of proof should be on those who claim a special normativity for morality which
is somehow missing from communication or epistemology.
Of course, it would be wonderful to have a full explanation of normativity
in general. The question has received some attention, but few truly ingenious
ideas have been found and there is little if any consensus about the nature of
normativity. At least nowadays, philosophers all explicitly recognize the problem. In large part, we are capable of seeing the problem now, in its full scope,
due to the work of Moore and Wittgenstein; we might indeed think that they
established the philosophical problematic for the twentieth century by helping
us to formulate our questions and we are just beginning now to wrestle our
188
Paul Bloomfield
way towards answers. Moore’s open question argument started us off by forcing us to confront the problems involved in following the semantic rules of
moral terms, but it was Wittgenstein’s rule following considerations which
showed us that the problem lies not in moral terms per se, or even in semantic
rules per se, but in the very idea of ‘rule-guided behavior’. Perhaps we will be
realists about normativity, or instead, in the end, we may adopt one form of
non-realism or another. But when we do so, our stance will not be one which is
a metaethical stance; rather it will be a stance toward normativity in general. It
will only be subsequent to this that we will be forced to deal with the repercussions of our general theory of normativity as they ramify through epistemology,
the philosophy of language, and moral philosophy.
References
Bloomfield, Paul (2001). Moral Reality. New York: Oxford University Press.
Church, Alonzo (1946). ‘A Review of 4 Articles on “The Paradox of Analysis” ’,
Journal of Symbolic Logic, 11: 132–3.
Darwall, S., Gibbard, A., and Railton, P. (1992). ‘Toward a Fin de Siecle Ethics’,
Philosophical Review, 101/1: 115–89.
Frankena, William (1939/1986). ‘The Naturalistic Fallacy’. Reprinted in P. Foot
(ed.), Theories of Ethics. Oxford: Oxford University Press: 50–63.
Gibbard, Allan (1992). ‘Normative and Explanatory Meaning’, Hempel Lecture I
(draft).
Kripke, Saul (1982). Wittgenstein: On Rules and Private Language. Cambridge, Mass.:
Harvard University Press.
Lewy, Casimir (1964). ‘G. E. Moore on the Naturalistic Fallacy’, Proceedings of the
British Academy, 50: 251–62.
Mackie, John (1977). Ethics: Inventing Right and Wrong. London: Penguin Books.
Malcolm, Norman (1958). Ludwig Wittgenstein. London: Oxford University Press.
Moore, G. E. (1902/1988). Principia Ethicia. Buffalo, NY: Prometheus Books.
—— (1942). ‘A Reply to my Critics’, in P. A. Schilpp (ed.), The Philosophy of G. E.
Moore. LaSalle, Ill.: Open Court.
Parfit, Derek (1984). Reasons and Persons. Oxford: Clarendon Press.
van Inwagan, Peter (1990). Material Beings. Ithaca, NY: Cornell University Press.
Wittgenstein, Ludwig (1929). ‘Lecture on Ethics’. Delivered in November 1929 to
the Heretics Society, Cambridge University.
—— (1958). Philosophical Investigations, trans G. E. M. Anscombe. New York:
Macmillan Publishing.
Opening Questions, Following Rules
189
—— (1993). Philosophical Occasions, 1912–1951, ed. J. C. Klagge and A. Nordmann.
Indianapolis: Hackett Publishing.
Wright, Crispin (1984). ‘Kripke’s Account of the Argument against Private
Language’, Journal of Philosophy: 759–78.
Zagzebski, Linda (1996). Virtues of the Mind. Cambridge: Cambridge University
Press.
This page intentionally left blank
9
Was Moore a Moorean?
Jamie Dreier
Mooreanism in metaethics consists of two theses: that Good¹ is unanalyzable
and that it is a non-natural property. It is an important position in twentiethcentury metaethics, not, primarily, because it was thought to be correct or even
especially plausible, but as a kind of backdrop against which some of the main
debates took place. Some metaethicists today are self-avowed Mooreans, but
not many. More commonly, philosophers interested in the nature of Good will
either deny the first thesis or else accept the first thesis but maintain that the
denial of the second thesis is compatible with the first. The latter seems to be
the most popular approach these days. So, for instance, Expressivists agree that
Good is unanalyzable, but deny that it is a non-natural property because, they
think, ‘Good’ is not the name of a property at all. And Cornell naturalists also
agree that Good is unanalyzable, but argue that this unanalyzability is a feature
of a word or concept and that an unanalyzable word may still be the name of
a complex natural property. By and large, Principia Ethica (Moore 1903, hereafter PE) set the stage for twentieth-century metaethics by challenging
philosophers to show how the second, more metaphysical thesis could be
avoided in the face of the very compelling first (and more conceptual) thesis.
Mooreanism itself threatens naturalistically inclined metaethicists: it is the
position they will be forced to adopt if they can neither see how to reject
the first thesis nor show how the first thesis can be made compatible with the
rejection of the second.
Thanks to Jordan Bleicher for research assistance and helpful discussion at the inception of this paper; to
audiences at Macquarie University, Charles Sturt University, and the Australian National University for
comments.
¹ I will follow Moore in using the word ‘Good’ as a (purported) name for a property, instead of calling it
‘goodness’ as most philosophers now would. Although Moore says that he is talking about the property, and
not about the word ‘good’ or the concept of good, it is not always clear that that is what he is in fact talking
about. However, I will take him at his word. Moore very often does not capitalize the word, though sometimes he does. I will capitalize it for the sake of clarity.
192
Jamie Dreier
There aren’t many Mooreans around these days. G. E. Moore himself was
presumably one. But was he, really? I want to raise a doubt about Moore’s own
position with respect to the second thesis of Mooreanism. Moore does say that
Good is a ‘non-natural’ property. But we properly attribute this view to him
only if in so saying he meant by ‘non-natural’ something that we mean when
we use the term. Philippa Foot once pointed out that we cannot attribute to
someone the belief that he is sitting on a pile of hay just because he says, ‘I am
sitting on a pile of hay.’ For suppose he goes on to explain: ‘yes, it has four legs
and a stiff back and was produced for the purposes of providing people with a
place to sit.’ Then we should conclude that by ‘pile of hay’ he means what we
call a chair. Similarly, we should be careful in attributing to Moore the thesis
that Good is a non-natural property, at least until we are confident that we have
some clear idea of what he meant by a ‘non-natural property’. Attention to
Moore’s writings should shake our confidence. For nowhere in his work is there
a clear and coherent explanation of which properties are non-natural that is
also consonant with his claim that Good is non-natural.
I will proceed as follows. In the first section, I investigate the locus classicus of
Mooreanism, namely, PE chapter I. Moore’s official explanation of what makes
a property non-natural is, to say the least, unhelpful. I then turn to another
chapter of PE, the one on ‘metaphysical ethics’, because ‘metaphysical’ is supposed to contrast with ‘natural’. Some of Moore’s remarks in that chapter are
very suggestive, but they are ultimately unsatisfying. In the third section, I look
at some of Moore’s later work, especially ‘The Conception of Intrinsic Value’
(1965), where he has another try at what appears to be the natural
property/non-natural property distinction put in other terms. A passage
identified by Broad as the key passage is indeed a crucial one, although not
exactly for the reason that Broad thought it was. In the last section I explain
why I think Moore was almost an Expressivist, and more important, what
distinguishes his view from Expressivism. Roughly, my suggestion is that
Moore differs from Expressivists in the way he takes certain phenomena to be
explained. It’s that special kind of explanation that distinguishes Moore as a
Moorean after all.
1. Natural properties in Principia Ethica IB
I’ll begin by looking at what Moore said in PE about non-natural properties.
First I’ll look at chapter I (‘The Subject Matter of Ethics’) where the main argument that Good is a simple, unanalyzable, non-natural property takes place.
The good news is that in that chapter Moore offers an explicit criterion for the
Was Moore a Moorean?
193
naturalness of a property. The bad news is that the criterion makes no sense.
Then I’ll turn to chapter IV (‘Metaphysical Ethics’), where Moore discusses
theories according to which Good is not a natural property. We might expect
to find Moore’s own view counted among the ‘metaphysical’ theories, but we
don’t. After a detour through the (alleged) distinction between being and
existence, I’ll conclude that there is no coherent conception of ‘non-natural’
properties in PE.
Simple vs. non-natural
I am no historian, but I do know that to figure out what a philosopher meant
by something, it’s a good idea to look at the arguments he gave for what he said,
and also, of course, to see what use he made of his idea. Moore does not seem to
have made much use of the idea that Good is non-natural. However, he is widely
supposed to have given some argument for it, namely, the open question argument. But the open question argument does not appear to be an argument to
the conclusion that Good is non-natural. It appears to be an argument to the
conclusion that Good is simple, unanalyzable—the first Moorean thesis.
Now, it might be thought that there really is no more to Moore’s claim that
Good is non-natural than that it is unanalyzable. The ‘non-natural’ in the
description ‘simple, non-natural, unanalyzable property’ is a redundancy,
according to this interpretation. But this cannot be right. Recall this passage
from PE:
There is, therefore, no intrinsic difficulty in the contention that ‘good’ denotes a simple
and indefinable quality. There are many other instances of such qualities. Consider
yellow, for example. We may try to define it, by describing its physical equivalent; we
may state what kind of light-vibrations must stimulate the normal eye, in order that we
may perceive it. But a moment’s reflection is sufficient to shew that those light-vibrations
are not themselves what we mean by yellow. They are not what we perceive. Indeed we
should never have been able to discover their existence, unless we had first been struck
by the patent difference of quality between the different colours. The most we can be
entitled to say of those vibrations is that they are what corresponds in space to the
yellow which we actually perceive. (1903: 10 § 10)
If it were not obvious (as Moore thinks) that yellow is a simple property, we
could determine that it was simple by an open question argument. I can
wonder whether something is yellow even when I know that it reflects ‘lightvibrations’ of such and such a kind; but I cannot wonder whether it reflects
those light-vibrations when I know that it does; so yellow cannot be the same
thing as reflecting those light-vibrations. The case of yellow is supposed to
soften us up, in case we were inclined to be skeptical about the possibility of
194
Jamie Dreier
simple properties. However, Moore did not think that yellow is a non-natural
property. So he is not thinking of ‘non-natural’ as just being, or following from,
simple. And this makes sense, after all. Since there are natural properties, there
must be simple ones. So showing that a property is simple could not be taken
to show that it is non-natural.
The official PE Account of ‘natural property’
In section 26 of PE, Moore gives a general explanation of what he counts as
‘nature’, and what he means by a ‘natural object’. Natural objects are those
which exist in time, he says, meaning to include experiences (which he thinks
do not exist in space but clearly do exist in time) and exclude numbers (on
which more below). Nature is the totality of what exists in time. However,
There is, indeed, no difficulty about the ‘objects’ themselves, in the sense in which I
have just used the term. It is easy to say which of them are natural, and which (if any)
are not natural. But when we begin to consider the properties of objects, then I fear the
problem is more difficult. Which among the properties of natural objects are natural
properties and which are not? For I do not deny that good is a property of certain
natural objects: certain of them, I think, are good; and yet I have said that ‘good’ itself
is not a natural property. Well, my test for these too also concerns their existence in
time. Can we imagine ‘good’ as existing by itself in time, and not merely as a property
of some natural object? For myself, I cannot so imagine it, whereas with the greater
number of properties of objects—those which I call the natural properties—their
existence does seem to me to be independent of the existence of those objects. They are,
in fact, rather parts of which the object is made up than mere predicates which attach
to it. If they were all taken away, no object would be left, not even a bare substance: for they
are in themselves substantial and give to the object all the substance that it has. But this
is not so with good. If indeed good were a feeling, as some would have us believe, then
it would exist in time. But that is why to call it so is to commit the naturalistic fallacy.
It will always remain pertinent to ask, whether the feeling itself is good; and if so, then
good cannot itself be identical with any feeling. (1903: 41 § 26)
Moore’s test is a thought experiment that we can all try. Let’s try it out on Good.
If you find that you cannot conceive of Good existing by itself in time, and not
merely as a property of some object, then you will agree that Good is not a
natural property. And, I’ll wager, you will all agree that Good does not pass this
test; you cannot imagine it existing by itself in time. The only problem is that
you may have quite a lot of difficulty imagining any property existing by itself
in time and not merely as a property of some object. I have this problem myself.
I can’t help thinking that maybe I am not understanding what it is that Moore
wants me to try to imagine.
Was Moore a Moorean?
195
There is some faint plausibility to the idea that we could imagine yellow, say,
existing by itself in time. I am supposing that Moore wants us to focus on a trope,
as we would now say; for example, the yellowness of a certain yellow flower.
Maybe we are to imagine this yellowness just floating there—the idea that the
yellowness is a part of the yellow object suggests this, as if the yellowness were the
two-dimensional part, the surface. Or maybe we are to imagine a shaped yellow
patch—though in that case, as I put it into words it sounds as if there is a yellow
object: the patch. Maybe we are to imagine an entirely yellow visual field? In any
case, whatever faint plausibility there is to the idea that yellow might pass the test,
I can’t see any at all in the thought that, say, having a one-kilogram mass might
pass it. I don’t know how to begin to try to imagine the property of being one kilogram in mass, existing by itself in time, and not as the property of any object.²
Broad, in his contribution to the Library of Living Philosophers volume on
Moore, wrote:
Now it seems to me that every characteristic of a natural object answers to Moore’s
criterion of non-naturalness, and that no characteristic could possibly be natural in his
sense. I do not believe for a moment that a penny is a whole of which brownness and
roundness are parts, nor do I believe that the brownness or roundness of a penny could
exist in time all by itself. Hence, if I accepted Moore’s account, I should have to reckon
brownness, roundness, pleasantness, etc., as non-natural characteristics. Yet he
certainly counts them as natural characteristics.³
It is hard to disagree. It turns out that it was impossible for Moore himself to disagree with Broad. In his ‘Reply’, Moore confessed: ‘This suggestion which I made
in Principia seems to me now to be utterly silly and preposterous’ (Schilpp 1968:
581–2). And then, ‘I agree, then, that in Principia I did not give any tenable
explanation of what I meant by saying that “good” was not a natural property.’
This is surprising, I think. There is no viable explanation of what a natural
property is in PE. So we don’t know what Moore meant by the hard-to-believe
feature of his view, because he gave no tenable explanation of what he meant by
saying that ‘good’ was a non-natural property. Now, an obvious strategy would
be to look to Moore’s other work to see if he had something illuminating to say
about which properties are natural. I will do that in a few minutes. For a little
² In some other contexts, when Moore wants an example of a natural property, he chooses experiential
properties: having an experience of a scarlet patch is an important example in his 1968: 587. Possibly, then, he
was thinking of an example like that when he described the test for natural properties in PE. I don’t see how
it would work, if we really do try to imagine the property having an experience of a scarlet patch without anybody doing the having, so to speak. So maybe Moore was thinking of the experience itself as existing in time
all by itself, and not as the experience of some person. It is not clear to me that this makes any sense, but in
any case it is not what Moore needs. The experience is an object, not a property.
³ Broad 1968: 59. In that essay, Broad was grappling with more or less the same problems of interpretation of Moore as I am. It will be obvious in what follows that I found Broad’s paper very helpful.
196
Jamie Dreier
while longer, though, I want to stay in the confines of PE. Let’s see what Moore
says about non-natural properties there.
2. Metaphysical ethics
Chapter IV of PE is about what Moore calls ‘metaphysical’ conceptions of
Good. What does he mean by ‘metaphysical’?
What, then, is to be understood by ‘metaphysical’? I use the term, as I explained in
Chapter II., in opposition to ‘natural’. I call those philosophers preeminently ‘metaphysical’ who have recognised most clearly that not everything which is is a ‘natural
object.’ ‘Metaphysicians’ have, therefore, the great merit of insisting that our knowledge is not confined to the things which we can touch and see and feel. They have
always been much occupied, not only with that other class of natural objects which
consists in mental facts, but also with the class of objects or properties of objects, which
certainly do not exist in time. . . . To this class, as I have said, belongs what we mean by
the adjective ‘good’. It is not goodness, but only the things or qualities which are good,
which can exist in time—can have duration, and begin and cease to exist—can be
objects of perception. (1903: 110–11 § 66)
So it looks like a metaphysical conception of Good is one according to which
Good is a non-natural property. That is, of course, what Moore believes; he
was, I dare say, one of those philosophers ‘who have recognised most clearly
that not everything which is is a “natural object”.’ So one might expect that the
chapter on metaphysical theories would include Moore’s own view. But instead
that chapter is purely critical. The chapter is obviously aimed at refuting metaphysical conceptions. How can this be? What did Moore believe that the
Metaphysicians did not?
It turns out that the answer is evident in a short bit I left out in the passage
quoted above. Here it is (in boldface):
‘Metaphysicians’. . . have always been much occupied . . . with the class of objects or
properties of objects, which certainly do not exist in time, are not therefore parts of
Nature, and which, in fact, do not exist at all. (Moore 1903: 110 § 66)
Good is a non-natural property, according to Moore; the Metaphysicians were
correct about that. Where they went wrong was in thinking that Good exists.
What is this supposed to mean? Well, Moore holds the same ontological views
about numbers.
It is quite certain that two natural objects may exist; but it is equally certain that two
itself does not exist and never can. Two and two are four. But that does not mean that
Was Moore a Moorean?
197
either two or four exists. Yet it certainly means something. Two is somehow, although it
does not exist . . . (1903: 111 § 66)
Likewise, Good does not exist. Everyone knows that Moore held Good to be a
simple, non-natural property. Nobody mentions that he also held that Good
doesn’t exist. How is it that this last feature of his metaethics, every bit as radical
as the familiar ones, has received so little subsequent attention? My provocative
question has a mundane answer. Moore’s claim that Good does not exist has no
significance except as a historical curiosity.
The distinction between being and existence is very mysterious to me. No
doubt it was commonplace in Cambridge in the early twentieth century, so
much so that Moore didn’t think he needed to explain it. But there is not much
point in looking to this distinction to try to work out what Moore really
thought about Good, since he later gave up on the distinction. In the middle of
a series of lectures given in 1911, Moore announced:
I used to hold very strongly, what many other people are also inclined to hold, that the
words ‘being’ and ‘existence’ do stand for two entirely different properties; and that
though everything which exists must also ‘be’, yet many things which ‘are’ nevertheless
do emphatically not exist. I did, in fact, actually hold this view when I began these
lectures; and I have based the whole scheme of the lectures upon the distinction. . . .
But nevertheless I am inclined to think that I was wrong, and that there is no such
distinction between ‘being’ and ‘existence’ as I thought there was. There is, of course, a
distinction of usage, but . . . (Moore 1953: 300)
but, he goes on to say, it has no real importance. We do not feel right saying that
numbers exist, though we are happy to say that there are such things, but
Moore has decided by early 1911 that this is a superficial fact of language.
I might say a few words about a kind of move I have now made twice. The
official PE criterion for naturalness (of properties), and now the being/
existence distinction are both bits of metaphysics that Moore came to reject, and
I have felt warranted in rejecting them as elements of an explanation of what he
meant (by distinguishing natural from non-natural properties). This move
looks shaky. It is entirely possible that Moore meant something at a certain time
and then later came to reject it. Could that be what’s happening here? I don’t
think it is. In neither case did Moore have a definite substantial view that he later
decided there were reasons for discarding. Rather, in each case, when he reflects
on what he once said or thought, he cannot see any sense in it at all. The
being/existence distinction looks to him like no distinction at all, and the independence in time criterion for natural properties, he says, strikes him as silly and
preposterous. What seems to have happened is something like this: Moore did
198
Jamie Dreier
have some distinction in mind between natural and non-natural (properties),
but he had only an intuitive grasp of it. When he tried to make it explicit, he was
proposing suggestions for capturing what seemed to him an entirely intuitive
idea. These suggestions he later decided were wrong, and indeed empty.
This suggestion, that the distinctions were misguided and not what Moore
really meant, seems more clearly true about ‘natural’ properties than about
being/existence. In the latter case, it is more plausible that he did think there was
an ontological chasm between the existent and the non-existent, among things
that are, and that this was what he really did have in mind when he wanted to
distinguish himself from ‘metaphysicians’. But there is a separate reason for
eliminating this hypothesis. That reason becomes clear when we look at what
Moore was actually doing in the chapter on Metaphysical theories.
What’s wrong with Metaphysical theories?
Moore has an objection to Metaphysical ethics that doesn’t depend on the being/
existence distinction, namely that these theories commit the naturalistic fallacy.
It was for this reason that I described ‘Metaphysical Ethics’ in Chapter II. as based upon
the naturalistic fallacy. To hold that from any proposition asserting ‘Reality is of this
nature’ we can infer, or obtain confirmation for, any proposition asserting ‘This is good in
itself ’ is to commit the naturalistic fallacy. . . . Such an assertion involves the naturalistic
fallacy. It rests upon the failure to perceive that any truth which asserts ‘This is good in
itself ’ is quite unique in kind—that it cannot be reduced to any assertion about reality,
and therefore must remain unaffected by any conclusions we may reach about the
nature of reality. This confusion as to the unique nature of ethical truths is, I have said,
involved in all those ethical theories which I have called metaphysical. (1903: 114 § 67)
Metaphysicians identify Good with some metaphysical non-natural property,
and yet they are guilty of the naturalistic fallacy, because they reduce statements
of what is Good to assertions ‘about reality’. At this point, one might conclude
that Moore is being willfully opaque. The one fallacy a Metaphysician could
not be guilty of is the naturalistic fallacy. And what does Moore mean by an
assertion ‘about reality’? Maybe this: reality is the totality of what exists.
Statements of value are not about reality, but about, well, whatever is left over
when reality is removed. But this can’t be right; it can’t be what Moore meant.
If it were, then Moore would have had a very different criticism of
Metaphysical theories of Good from the one he actually had. He would have
said that Metaphysicians were right, except that they say that Good exists.
There would be a good chance of converting a Metaphysician’s theory to a
correct theory by changing all occurrences of the verb ‘to exist’ to occurrences
of the verb ‘to be’. But Moore says nothing like this. His actual criticism of
Was Moore a Moorean?
199
Metaphysicians is that they, like Naturalists, want to reduce statements of value
to other statements (statements ‘about reality’). That is, in a way, the same
fallacy as the Naturalists commit, only the type of the reduction statement is
different. ‘Naturalistic fallacy’ is a bad name for this fallacy, but it is still the
same fallacy in each case.⁴
That Moore’s point about the ‘naturalistic fallacy’ is really a point about
reducibility rather than having anything to do with the natural suggests that the
distinction he had in mind when he insisted that Good was in some special class
of its own was not a distinction that had especially to do with any division between
the natural and the non-natural as most philosophers would think of that division.
Again, what he meant to be saying about Good by calling it non-natural cannot be
merely that it is unanalyzable, since yellow is also unanalyzable. Still, what he had
in mind clearly did have something to do with Good’s unanalyzability. In the next
section we turn to a later paper of Moore’s, ‘The Conception of Intrinsic Value’. In
that paper together with his reflections on it Moore confirms that what he meant
in PE by calling Good non-natural does indeed have to do with its peculiar logical
relation to the other, natural properties that good things have. Before turning to
‘The Conception of Intrinsic Value’, let me make one quick observation about this
claim of Moore’s, which I just quoted above in its context.
[A]ny truth which asserts ‘This is good in itself ’ is quite unique in kind—that it cannot be reduced to any assertion about reality, and therefore must remain unaffected by
any conclusions we may reach about the nature of reality. (Moore 1903: 114 § 67)
When Moore says that statements of value are not about reality, he sounds
an awful lot like an Expressivist. Of course, he wasn’t an Expressivist. But he
was awfully close. Toward the end of this paper, I’ll be arguing that Moore was
awfully close to being an Expressivist, and I will try to say what distinguished
Moore’s actual view from Expressivism.
3. Natural properties in ‘The Conception of Intrinsic Value’
In ‘The Conception of Intrinsic Value’, Moore explains what he means by
saying that some value is intrinsic.
To say that a kind of value is ‘intrinsic’ means merely that the question of whether a thing
possesses it, and in what degree it possesses it, depends solely on the intrinsic nature of the
thing in question. (1965: 260; italics in original)
⁴ Support for this interpretation comes from some contemporary sympathizers of Moore’s. Frankena
(1939) (not really a sympathizer) suggested ‘definist fallacy’; Wood (1937) suggested ‘valuational fallacy’;
J. Laird says: ‘The essential point is the irreducibility of values to non-values’ (1926: 94 n.).
200
Jamie Dreier
But although the intrinsic value of something does depend solely on the intrinsic
nature of the thing, Moore insists, rather perversely, that value is not an
intrinsic property. It is a property. And it is intrinsic. But it is not an intrinsic
property. This is a puzzle. It is not, as Moore appears to have thought, a metaphysical puzzle (how could this strange thing be, that a property which is
intrinsic turns out not to be an intrinsic property?), but for us rather a terminological puzzle (what on earth was Moore talking about?). Broad suggested that
what Moore meant by a ‘natural property’ in PE is what he here, in ‘The
Conception of Intrinsic Value’, means by an ‘intrinsic property’. I think this is
a very good suggestion. Moore had identified, inarticulately, something he
found odd about Good, something that led him to say that it is non-natural. In
‘The Conception of Intrinsic Value’, he again finds himself confronted with
the same oddness, and this time it leads him to say that even intrinsic goodness
is not an intrinsic property.
In his ‘Reply’, Moore agrees with Broad that the natural/non-natural distinction of PE is the very same distinction as the intrinsic property/non-intrinsic
property distinction of ‘Conception’. He apologizes for the outlandish terminology that forced him to say that some properties that are intrinsic are not
intrinsic properties. Moore writes:
[I] will now drop the awkward terminology of Philosophical Studies, and speak instead
of a distinction between intrinsic properties which are ‘natural’ and intrinsic properties
which are not ‘natural;’ since I think that the feature, additional to that of being ‘intrinsic,’ which a property must possess in order to be what I called in Philosophical Studies
an ‘intrinsic property,’ is the very one which would have led me, in Principia, to call it
a natural intrinsic property. The question is: What is this feature? What is the difference
between a ‘natural’ intrinsic property and one which is not natural? Mr. Broad says truly
enough that in Philosophical Studies I give no clear account of this distinction.⁵
Moore did give some account of the distinction in ‘The Conception’, only not
a clear one. In fairness to Moore, he confessed right away that he could not give
a clear account. Here is what he wrote in ‘The Conception of Intrinsic Value’:
[T]here must be some characteristic belonging to intrinsic properties which predicates
of value never possess. And it seems to me quite obvious that there is; only I can’t see
what it is. . . . I can only vaguely express the kind of difference I feel there to be by saying
that intrinsic properties seem to describe the intrinsic nature of what possesses them in
a sense in which predicates of value never do. If you could enumerate all the intrinsic
properties a given thing possessed, you would have given a complete description of it,
and would not need to mention any predicates of value it possessed; whereas no
⁵ Moore 1968: 584–5. Here and hereafter, Moore’s references to Philosophical Studies are always to ‘The
Conception of Intrinsic Value’, which appears in the collection of his papers, Philosophical Studies.
Was Moore a Moorean?
201
description of a given thing could be complete which omitted any intrinsic property.
(1965: 272 ff.)
Like Broad, I think this passage contains the crux of the matter. First let’s see
what Broad made of the distinction. Next, we’ll take a look at what Moore said
about it upon reflection. Unfortunately, what he said renders the distinction
more confusing. I’ll try to untangle things a little, and then return to the crucial
passage with my own suggestion about what was going on, and how in light of
everything I understand the non-naturalism of Moore’s metaethics.
What Broad said
Broad says,
I am inclined to think that the fact which Moore has in mind here is that goodness, in
the primary sense, is always dependent on the presence of certain non-ethical characteristics which I should call ‘good-making’. . . . We might, therefore, distinguish the
characteristics of a thing into the following two classes, viz., ultimate and derivative.
Goodness would certainly fall into the class of derivative characteristics. (1968: 60)
This is the distinction we would now recognize as the distinction between
supervening (‘derivative’) properties and the base of properties on which
they supervene, the ‘ultimate’ properties, or as Hare puts it, the subjacent
properties.⁶ Broad’s suggestion is that when Moore says that a ‘complete
description’ of something might still leave out certain properties, he must be
thinking of the supervening properties. The point would then be that goodness
is a supervening property. Broad gives an example of supervening properties:
Now there is a sense in which one might say that a thing could not be completely
described if any of its ultimate characteristics were omitted, but that it could be completely described without mentioning all its derivative characteristics. In describing a
circle, e.g., it is not necessary to mention explicitly any of the innumerable properties
of circles which follow of necessity from their definition together with the axioms of
Euclidean geometry. (1968: 60–1)
But Broad then points out that in the first place, the way in which goodness
supervenes on the nature of an object seems to be importantly different from
the way in which the derived properties of circles supervene on the ultimate
⁶ It is not quite the same distinction. First, subjacent properties need not be ultimate, since they may
themselves supervene on further properties. Ultimate properties, presumably, supervene on no other properties. Second, supervenience may obtain in a trivial way, as for example the property of being self-identical
supervenes on shape merely because it supervenes on anything and nothing. ‘Derivative’ suggests that the
properties from which the derivative ones derive figure in the derivation. Broad’s idea of being derivative is,
quite reasonably, a vague one; supervenience is a sharper tool with which to cut the same material.
202
Jamie Dreier
ones; and in the second place, properties that supervene on natural properties
certainly do not seem generally to be non-natural in any recognizable intuitive
sense.
In the first place, the way in which ethical properties of a thing depend on its non-ethical
properties seems to be quite unlike the way in which the remaining properties of a
circle depend on its defining properties. In the latter case the dependence is equivalent
to the fact that the possession of the remaining properties can be inferred deductively
from the axioms of Euclid and the presence of the defining properties. But the
connexion between the non-ethical bad-making characteristic of being an emotion of
delight at another’s pain and the ethical characteristic of being morally evil is certainly
not of this nature.
Moreover, it is surely quite as evident that pleasantness and unpleasantness are derivative characteristics of an experience as that goodness and badness are. (Broad 1968: 61)
The first point is somewhat suggestive. The supervenient properties of a circle
follow deductively from the subjacent properties, whereas, at least according to
Broad, and surely according to Moore, goodness does not follow deductively
from any of the ‘good-making’ properties, the natural properties on which
goodness supervenes. This failure to follow deductively leaves ethical properties
in a peculiar realm, for the explanation for supervenience present in the relatively straightforward case of geometric properties is not available for ethical
ones. Though Broad doesn’t say so, it’s somewhat plausible that it is precisely
this peculiar feature of ethical properties that Moore had in mind—maybe
only in the back of his mind—when he called them non-natural, and again
when he said that they are not intrinsic.
What Moore said in reply to Broad
Now, a couple of things that Moore says in reply to Broad tend to strengthen this
interpretation. First, he says that his original criterion for being an intrinsic
property was not quite correct, since some intrinsic properties—the ones that
‘follow from’ other natural intrinsic properties—could in fact be left out of a
‘complete description’ of an object. In the ‘Reply’, he changes the criterion:
No description of a thing could be complete which omitted any of those among its natural intrinsic properties, which are such that no other natural intrinsic properties which
the thing possesses entail them. (1968; italics in original)
So by this criterion, the supervening properties of a circle would count as
natural intrinsic properties, even though a complete description of a circle could
omit them. Here, apparently, Moore intends to be allowing the derivative
properties of circles into the class of natural intrinsic properties while still
Was Moore a Moorean?
203
keeping ethical properties out of the class. And this does tend to strengthen the
hypothesis that what’s ‘non-natural’ about ethical properties is precisely that
they supervene on the natural intrinsic base without ‘following from’ it.
Second, here is Moore’s confirmation that he always did believe that goodness supervenes on natural properties:
It is true, indeed, that I should never have thought of suggesting that goodness was
‘non-natural,’ unless I had supposed that it was ‘derivative’ in the sense that, whenever
a thing is good (in the sense in question) its goodness (in Mr. Broad’s words) ‘depends
on the presence of certain non-ethical characteristics’ possessed by the thing in
question: I have always supposed that it did so ‘depend’ . . . (Moore 1968: 588)
So far so good. But then Moore throws a wrench into the works:
I have always supposed that it did so ‘depend,’ in the sense that, if a thing is good (in my
sense), then that it is so follows from the fact that it possesses certain natural intrinsic
properties, which are such that from the fact that it is good it does not follow conversely
that it has those properties. (1968: 588)
This passage is a hermeneutic disaster. I was strongly tempted to pretend it
didn’t exist. For here Moore declares that the fact that a thing is good follows
from the fact that it possesses certain natural properties. And this, I had always
thought, was precisely the claim that he had so famously refuted with the open
question argument.
I still think so. And I think that what he says in the disastrous passage is
something he simply did not mean. One way of avoiding the awful implication
is to suppose that when Moore says that goodness ‘follows’ from the natural
properties, he is talking about something other than logical implication.
Unfortunately, in the very same essay, Moore advises us that he has always been
using ‘follows from’ in the logical sense:
This partially explains how I have been using ‘follows’ as applied to propositions: I have
been so using it that it can only be said with truth that a proposition q follows from
another p, when it can also be said with truth that ‘p, but not q’ is self-contradictory.
(1968: 607)
My hypothesis is that he has not, in fact, been using ‘follows’ in the way that he
claims he has. Moore identifies looser senses of ‘follows’, and as an example he
says that
Before aeroplanes were invented you could have said with perfect truth, on hearing that
a man was in New York the day before yesterday, ‘Then it follows that he wasn’t in
London yesterday.’ And this is a way in which ‘follows’ is used in ordinary speech.
(1968: 607)
204
Jamie Dreier
The way in which the fact that something is good follows from the fact that it
has such-and-such natural properties is not just like the way it follows from the
fact that a man was in New York the day before yesterday that he wasn’t in
London yesterday, since the latter is a kind of natural necessity and the former
is metaphysical. Still, we may, and I think we must if we are to avoid attributing to Moore a completely incoherent view, suppose that in the infamous
Admission of Supervenience passage Moore was using ‘follows from’ in one of
the non-logical, more ordinary senses. Back in ‘The Conception of Intrinsic
Value’ Moore mulls over the kinds of necessary implication, and in connection
with just the same point about supervenience:
You cannot say that an intrinsic property is a property such that, if one thing possesses
it and another does not, the intrinsic nature of the two things must be different. For this
is the very thing which we are maintaining to be true of predicates of intrinsic value,
while at the same time we say that they are not intrinsic properties. Such a definition of
‘intrinsic property’ would therefore only be possible if, we could say that the necessity
there is that, if x and y possess different intrinsic properties, their nature must be
different, is a necessity of a different kind from the necessity there is that, if x and y are
of different intrinsic values, their nature must be different, although both necessities
are unconditional. And it seems to me possible that this is the true explanation. But, if
so, it obviously adds to the difficulty of explaining the meaning of the unconditional
‘must,’ since, in this case, there would be two different meanings of ‘must,’ both unconditional, and yet neither, apparently, identical with the logical ‘must.’ (1965: 272)
Broad called ‘The Conception of Intrinsic Value’ a ‘very difficult paper’, no
doubt on the basis of passages like this one. I do think, however, that Moore’s
struggle with the perplexing idea that the goodness of a thing follows with some
kind of necessity from its natural intrinsic qualities even though it never does
follow deductively that the thing is good, just from propositions about its natural intrinsic qualities, justifies discarding his self-interpretive dictum about
what he always means by ‘follows from’.
4. What disinguishes a Moorean?
To summarize, then: there is something very odd about ethical properties,
namely, that although they supervene on the straightforwardly natural properties they do not follow deductively from them. This odd feature somehow led
Moore to say that ethical properties are not natural, and also, later, that they are
not intrinsic properties.
Just why might it have led him to say these things? Well, we might think of
the supervenient properties of something that do not strictly reduce to the
Was Moore a Moorean?
205
intrinsic properties as somehow ‘floating’ on top of the proper intrinsic
properties, ‘emerging’ from them, so to speak, and so not quite ‘inhering’ in them.
All very metaphorical, but perhaps enough to lead Moore to think of them as
lacking a certain sense of ‘intrinsicness’. But why non-natural? Because they,
the ethical properties, do not reduce to the paradigm of natural properties? But
then we immediately run into the old problem: yellow is just as Good an
example of a supervening property as Good is—it supervenes, though perhaps
not quite in the same way as Good, on the scientific properties of its objects—
and yet yellow is used by Moore as an example of a natural irreducible property.
So, suggestive as all of this talk of supervenience may be, I am afraid it is not
after all the source of Moore’s idea that Good is non-natural. Still, I do think we
had the right passage in our sights just a moment ago.
I can only vaguely express the kind of difference I feel there to be by saying that intrinsic
properties seem to describe the intrinsic nature of what possesses them in a sense in
which predicates of value never do. If you could enumerate all the intrinsic properties
a given thing possessed, you would have given a complete description of it, and would
not need to mention any predicates of value it possessed; whereas no description of a
given thing could be complete which omitted any intrinsic property.
I think this is the crux of the matter, for Moore. We don’t describe anything by
calling it good.
What do you call a metaethicist who maintains the following theses?
Ethical terms do not reduce to non-ethical ones.
‘Good’ is not the name of any natural property.
Ethical statements are not about reality.
Ethical statements do not describe anything.
Without further information, we would probably conclude that he is an
Expressivist. So, was Moore an Expressivist? I would hate to think so. What
Moore says about predicates of value failing to describe their object sounds like
something Allan Gibbard (an Expressivist) or Mark Timmons (whose view is a
close relative of Expressivism) might say (Gibbard 1990, Timmons 1999). My
hypothesis is that Moore was a proto-Expressivist. I am not saying this merely to
be perverse. I think it explains the data. And not just these that I have been presenting; I think there are other surprising things Moore says that are also very
close to what Expressivists say. For instance, his views about the connection
between ethical judgment and the will are much more like R. M. Hare’s or
Gibbard’s than has generally been recognized.⁷ But it would be alarming, and
⁷ I hope to explore this parallel in another paper.
206
Jamie Dreier
to my mind disappointing, to discover that Moore was an Expressivist.
Disappointing because it has been important historically, and it still is
dialectically important today, to have Mooreanism as a contrastive position to
Expressivism.
So, why was Moore a Moorean after all? What did he believe that
Expressivists do not believe? Moore actually says that some properties describe
an object, and that Good does not. That’s a funny thing to say. We would say (I
take it) that some words, or a person, describe an object. But Moore says that
it’s properties. Did he mean this literally? Moore is not always careful about
distinguishing use and mention, so it’s possible that when he said that properties describe an object he meant that the predicates that denote those properties
describe the object. Still, I think it is significant that he insists that properties
describe objects. The point is that description inheres in properties, not in
predicates.
This much seems clear: the metaethics of PE takes as its centerpiece
the divide between the ethical and the non-ethical. I’d like to suggest that
Moore misdescribed this distinction as the distinction between the natural and
the non-natural. Even supposing that he had something coherent in mind when
he wrote of natural and non-natural properties (and I am not at all sure that he
had), his assertion that Metaphysical ethics is also guilty of the naturalistic
fallacy shows that it’s unlikely that natural is what he really thought Good
lacked. Instead, I am suggesting, the distinction he was after was the one
between the descriptive and the evaluative. Combine what he says in his
Library of Living Philosophers essay with what he said in ‘The Conception of
Intrinsic Value’, and it looks like Moore came close to the same conclusion
about what he must have meant in PE.
But if the controversial thesis of Principia Ethica is that the evaluative is not
reducible to the descriptive, what made him a Moorean rather than an
Expressivist? It must be that according to Mooreans, the gap between the
descriptive and the ethical is a gap between properties. All of us (almost) think
that there is a logical gap between description and evaluation. Hare famously
argued that the gap is accounted for by proper attention to forms of expression:
moral judgments do not describe because they prescribe, they don’t fail to
describe because they are about special non-descriptive properties. Allan
Gibbard puts the responsibility on normative concepts. But Mooreans think the
logical gap between description and evaluation (or prescription, or judgments
of what to do) is explained by a matching gap between kinds of properties.
In some of his recent work, Allan Gibbard has suggested that one way of
individuating properties is by what they explain. I am suggesting that a way of
distinguishing what Moore believed from what Gibbard believes is according
Was Moore a Moorean?
207
to what explains the gap between description and evaluation. Expressivists say
the explanation is to be found in the different roles that concepts play in
planning; Prescriptivists say it is to be found in the different speech acts that
declarative sentences can perform. Mooreans say it is to be found in the
different sorts of properties that there are.
I don’t see how Mooreans could be right about this. But at least, if I am right,
I now know what makes them Mooreans. I know what it is they believe that I
cannot believe.
References
Broad, C. D. (1968). ‘Moore’s Ethical Doctrines’. In Schilpp 1968.
Frankena, William K. (1939). ‘The Naturalistic Fallacy’. Mind, 48: 464–77.
Gibbard, A. (1990). Wise Choices, Apt Feelings. Cambridge, Mass: Harvard University
Press.
Laird, J. (1926). A Study in Moral Theory. London: G. Allen & Unwin.
Moore, G. E. (1903). Principia Ethica. Cambridge: Cambridge University Press.
—— (1953). Some Main Problems of Philosophy. London: G. Allen & Unwin.
—— (1965). ‘The Conception of Intrinsic Value’. In Philosophical Studies. Totowa,
NJ: Littlefield, Adams & Co.
—— (1968). ‘A Reply to my Critics’. In Schilpp 1968.
Schilpp, P. A. (ed.) (1968). The Philosophy of G. E. Moore. 2 vols. La Salle, Ill.: Open
Court.
Timmons, M. (1999). Morality without Foundations. Oxford: Oxford University Press.
Wood, L. (1937). ‘Cognition and Moral Value’. Journal of Philosophy, 34/9: 234–9.
This page intentionally left blank
10
Ethics as Philosophy
A Defense of Ethical Nonnaturalism
Russ Shafer-Landau
It is early days in the Principia Ethica when Moore presents us with his famous
argument from elimination on behalf of ethical nonnaturalism. Moore
thought that there were three options when it came to a diagnosis of ethical
concepts. Ethical notions are either meaningless, susceptible of a naturalistic
analysis, or nonnatural. Neither of the first two options appealed. Ethical
nonnaturalism won by default.
It took only a long generation before nonnaturalism’s star began to fade. The
ascendancy of noncognitivist views, and the resurgence of naturalistic ones,
were prompted by the perception that nonnaturalism suffered from fatal flaws.
By century’s end, this perception had become so widespread as to be rightly
considered a bit of conventional philosophical wisdom. I think that this
critical, and often dismissive, attitude is mistaken—as mistaken as Moore’s
contention that he had identified a line of argument that proved, once and for
all, the falsity of ethical naturalism.
It is a commonplace that philosophical preferences are cyclical, and that the
runt of the litter in one era may elbow out its competition in later times. Ethical
nonnaturalism, for decades consigned to second-class status, is due a reappraisal.
The nonnaturalism I favor is a brand of moral realism. As I understand it,
moral realism is the view that says that most moral judgments are beliefs,
some of which are true, and, when true, are so by virtue of correctly representing the existence of truth-makers for their respective contents. Further,
and crucially, true moral judgments are made true in some way other than by
virtue of the attitudes taken towards their content by any actual or idealized
human agent.
My thanks to Paul Bloomfield, whose sense and gentle suasion I too frequently ignored in preparing the final
version of this paper. Thanks also to Terence Cuneo, who gave me a number of incisive criticisms and suggestions for improvement.
210
Russ Shafer-Landau
Not all nonnaturalisms are realistic—Kantian views, for instance, reject the
assimilation of moral to natural properties, and yet also reject realism. I will
proceed on the assumption (unargued here) that realism is the best path for
nonnaturalists.¹ After describing nonnaturalism, and identifying the most
serious worries that face it, I will undertake a partial defense against a number
of those problems. My preferred strategy for doing so invokes a parallel
between philosophy in general, and ethics in particular. My contention is that
once we pay special attention to this relationship, a number of the traditional
concerns about nonnaturalism begin to seem less pressing than they have for a
long while.
1. The nature of nonnaturalism
Ethical nonnaturalism is, first and foremost, a metaphysical doctrine. It claims
that there are instantiated moral properties that are not natural properties.
There are instantiated moral properties: this element distinguishes nonnaturalism from all forms of moral nihilism.² Moral nihilism is the view that either
there are no moral properties, or there are, but none that are ever instantiated.
On this view, there is no moral reality at all. Nonnaturalism asserts the
existence of moral properties—there really is such a thing as being (e.g.)
morally forbidden, or morally good—and asserts, too, that these properties are
sometimes exemplified.
Nonnaturalism per se is, in two respects, neutral as to the exact nature of moral
properties. It is first of all ecumenical regarding the conditions under which these
properties are instantiated. Many substantive normative ethical theories are
compatible with a nonnaturalistic metaphysics. And, second, nonnaturalism
is compatible with any number of specific views about what a property is.
Nonnaturalists can await a verdict from the metaphysicians on this question,
and incorporate their best answer into a comprehensive metaethic.
Ethical nonnaturalism is true only if moral properties are not natural
properties. There is no agreement on what makes a property a natural one. The
criteria that immediately come to mind—a property is natural in virtue of its
¹ I argue against alternatives to realism in Part I of Moral Realism: A Defence (Oxford: Clarendon Press,
2003). Many of the ideas that appear in this essay receive much fuller development in that book. I have also
appropriated a couple of sentences from my Whatever Happened to Good and Evil? (New York: Oxford
University Press, 2004). My thanks to the Press for permission to rely on this material here.
² The following, in my book, all qualify as versions of nihilism: emotivism (e.g. Ayer 1936; Stevenson
1937, 1948), prescriptivism (Hare 1952, 1963), expressivism (Blackburn 1993, 1998; Gibbard 1990),
assertoric nondescriptivism (Timmons 1999, Skorupski 1999), and error theories (Mackie 1977; Garner
1994; Joyce 2001).
Ethics as Philosophy
211
being instantiated nonconventionally, materially, or tangibly, or being describable
without reliance on evaluative vocabulary—will not do. If nonnaturalists are
right, then morality is not conventional. That doesn’t make them naturalists.
Some paradigmatically natural properties are realized nonmaterially (e.g. being
a vacuum). Not every realization of a natural property is tangible—one can’t
literally touch a muon or a gluon. And, as Nicholas Sturgeon rightly points out
in a recent essay (2003), one can’t always tell a natural property by the vocabulary in which it is described. If naturalists are correct, then moral worth, virtue,
and rightness are natural properties, though all are surely also evaluative
notions. The debate between naturalists and nonnaturalists cannot be solved
just by noting salient features of the terms we use to describe it.
I suggest that we take a different tack, that of offering a disciplinary conception. The natural is whatever is the object of study by the natural sciences. The
obvious problem of having just pushed the definitional problem back a step is
usually solved by defining the natural sciences ostensively: something is natural
just in case it figures ineliminably in true propositions that emerge from (on
some accounts, the perfected versions of ) physics, chemistry, molecular
biology, astronomy, etc. Leave astrology off that list. Film studies, too.
It would be nice to have an account of what makes a natural science natural.
This for two reasons. First, we’ll want to know how to fill in that ‘etc.’, especially
for contentious cases. And ethics is likely to be among the most contentious.
But, second, we presumably don’t want to limit the natural domain to the
things studied by (even perfected versions of ) the sciences we now recognize—
we want to allow for the emergence of natural sciences that we haven’t yet
dreamed of.
I don’t think that ethics is a natural science. Its fundamental principles are
not inductive generalizations. It is not primarily concerned with causal efficacy.
Its central principles are not descriptive of historical contingencies. The
phenomena it does describe are supervenient as a matter of conceptual requirement. It allows for a much greater degree of indeterminacy and vagueness
than is found in typical natural sciences. It has only a very little concern for
mathematical quantification and precision. Unlike any of the recognized
sciences, its truths are normative truths that direct and guide, rather than (in
the first instance) predict the course of future events or explain what has already
occurred. Moral truths provide justifying reasons that are often ignored.
Physics and geology and hydrology don’t do that.
Of course, ethical naturalism may be correct even if ethics is not a science.
Ethical naturalism is true so long as moral properties are the proper object of
natural scientific study. They might be, even if ethics isn’t the science that does
the studying.
212
Russ Shafer-Landau
But if ethics isn’t the relevant science, then what is? Sociology and psychology,
if we admit them into the pantheon, might do yeoman’s work—they will be
required to aid us, for instance, in determining how many people are made
happy by various actions, or what it was that actually motivated agents to do
what they did. All quite helpful information. But neither sociology nor psychology, nor any other science that I know of, will do much to fix the content of
our fundamental moral principles. That is the job of ethics proper.
Suppose my truncated case against the scientific status of ethics is way off
base. So let us grant, provisionally, that ethics is a science. Still, is it a natural
science? We might, for instance, also allow mathematics to qualify as a science.
But not many think of it as a natural science. What explains this, it seems to me,
is the nature of mathematical investigation. Mathematics is done in a largely a
priori fashion. And this leads us to the following thought: a science is a natural
science just in case its fundamental principles are discoverable a posteriori,
through reliance primarily on empirical evidence.
Any definition of naturalism is bound to be to some extent stipulative. Yet
this characterization is true to all instances of avowed ethical naturalism that
I know of, and does seem to capture what many have thought to be essential to
the classification. But notice, now, that the difference between naturalists and
nonnaturalists, usually taken to be a metaphysical one, turns out on this
accounting to be epistemological. Ethical nonnaturalists are those who claim
that moral properties are not natural ones. This means, on the present understanding, that fundamental moral truths are discoverable a priori. If I am right,
prospects for nonnaturalism depend crucially on the possibility of gaining such
knowledge.
It isn’t my intention here to offer a direct defense of a priori moral knowledge.³
Instead, I will present some ancillary arguments that constitute an invitation to
see ethical enquiry as importantly dissimilar from empirical, natural scientific
investigation. The rationale and the structure of these supporting arguments is
best appreciated by noting the classical objections to nonnaturalism.
2. Objections to nonnaturalism
Many have opposed ethical nonnaturalism just because they have thought a
priori moral knowledge an impossibility. Scientific confirmation is our model
of epistemic success; if morality fails to conform to the model, the only alternative
³ I offer a somewhat programmatic defense in Moral Realism, ch. 11. See also Audi 1997, 1999, 2004, and
the fine collection edited by Stratton-Lake (2002).
Ethics as Philosophy
213
(it has seemed) is moral skepticism. Before I sketch my preferred path to
resisting this conclusion, we must consider other potent objections to
nonnaturalism.
The first of these is motivational: moral judgments are thought to bear a
necessary connection with motivation. But if nonnaturalists are right, then
moral judgments represent efforts to report on the instantiations of nonnatural
properties. How could such (putative) reports necessarily engage the will?
The second worry focuses on practical reasons. Many believe that our moral
obligations constitute or entail categorical reasons, i.e. reasons that apply independently of our inclinations. But how could the instantiations of nonnatural
properties supply such reasons? Our reasons stem, ultimately, from what each
of us most cares about. But the link between our existing concerns, and nonnatural properties, will be tenuous, at best.
A third concern is metaphysical: if moral properties are not natural properties, then what can they be? Everything else we know of in the universe is
(potentially) scientifically explicable. Nonnaturalism insists on a class of sui
generis properties, and this (it is said) amounts to ontological profligacy.
Next comes the epistemological concern. If scientific methods don’t answer
moral questions, then how are they to be answered? For nonnaturalists, this
problem has always taken a quite specific form. Nonnaturalists, by definition,
deny the identity of moral and natural properties. They are also agreed that, for
any moral and natural property, there are no relations of necessary coextension
that link them.⁴ There is no way to fill in the following blank: necessarily,
something is good [right/virtuous/praiseworthy, etc.] if and only if it is _____
[fill in a term designating a natural property]. In the absence of the relevant
biconditionals, it is easy to see how we might be epistemically confounded. If
we don’t know the natural conditions under which moral properties are invariably instantiated, then it is hard to know when they are instantiated at all. That
is precisely the epistemic problem facing nonnaturalism.
The epistemological problem is very closely related to the last significant
problem for nonnaturalism—that of explaining persistent moral disagreement.
When we compare the plausibility of two hypotheses introduced to account
for such disagreement—(i) some parties are misapprehending the relations
linking nonnatural and natural properties, or (ii) the interlocutors are giving
⁴ I don’t believe that this is a defining feature of nonnaturalism, but it is a claim that all nonnaturalists I
am familiar with have endorsed. Of course, if necessary coextension entailed identity, then such aversion is
readily explicable. But that entailment fails, or so I believe. Being self-identical and being either red or not
red are necessarily coextensive properties. They aren’t identical. Being a plane figure with a side and being one
with an angle are also necessarily coextensive. But being an angle is not the same thing as being a side. (For
the contrary view, see Jackson 1998: 126–7.)
214
Russ Shafer-Landau
voice to parochial commitments they’ve been raised and socialized to accept—the
simpler explanation has always seemed to be the latter. Rather than attributing
some cognitive failing(s) to people involved in moral disagreements, we do
better to see such cases as exemplifying what Mackie has called ‘patterns of
objectification’ (1977: 42 ff.). The reason, of course, is that citing cognitive
failings in this context is plausible only in the presence of a successful moral
epistemology that provides epistemic standards for measuring such failings.
And that is just what we don’t have. Why invoke sui generis properties when
familiar kinds of projection can do the trick?
These longstanding worries are legitimate and serious. There is obviously no
practicable way to reply to all of them here. Since that is so, I will simply summarize my preferred direction of argument for some cases, and then devote the
rest of the paper to the general strategy that I think, with development, can lead
to satisfying solutions for the remainder of the worries.
I’ll take a pass here on the first two problems. My explanation for this in the
case of the motivational problem is that it is, as traditionally conceived, a
difficulty equally for any cognitivist. Cognitivists see moral judgments as
truth-evaluable beliefs, as intended efforts to report on the nature of a moral
reality. If, as Hume claims, beliefs of any stripe are unable to motivate all by
themselves, then every cognitivist view (ideal-observer theories, varieties of
relativism and subjectivism, sensibility theories, Kantian views, and the miscellany of moral realisms) will fall to the ax. I don’t think any argument can be that
powerful. In any event, nothing about nonnaturalism per se makes it any more
vulnerable to this worry than any of its competitors in the cognitivist camp.
There are two ways to defend against the motivational argument. The first is
to argue that beliefs—especially evaluative ones—are capable of motivating all
by themselves, or, by themselves, are able to generate derived desires that
together yield motivation. The second is to deny that moral judgments necessarily motivate, and to provide an explanation for why such judgments, when
sincerely held, nevertheless do manage to motivate as often as they do. I think
that both strategies can be successfully brought off, though obviously not here.
I have a different basis for avoiding any discussion of the worries regarding
practical reason. Here, some of realism’s cognitivist competitors are at an
advantage. Specifically, all of those who make our moral duties some function
of our attitudes can readily explain why we have reason to do what morality
requires of us. It takes a lot more work to show that realists—naturalists or nonnaturalists—can do as good a job of explaining morality’s normativity.
The typical realist response to these charges is to deny that moral obligations do
in fact entail reasons for action. Sometimes we may lack any reason to do as morality demands. The standards that specify the content of our moral obligations do
Ethics as Philosophy
215
not have their source in human attitudes; reasons do; therefore there can be the
relevant disconnect. My way out of this problem begins by insisting that moral
duties do entail excellent practical reasons, but proceeds to deny that such reasons must be reachable from sound deliberation that starts from our existing
commitments. I don’t see any short way of substantiating these points, however, and I don’t expect anyone just to take my word that successful solutions to
these problems are possible.⁵ Instead, I suggest that we bracket these concerns
at this point. Even if my promissory notes can be adequately cashed, that still
leaves plenty of other problems for the nonnaturalist realist to deal with.
3. Ethics as philosophy
Ethics is a branch of philosophy. Few would dispute that. Yet this fact has
significant, wide-ranging implications, many of which have gone little noticed
in debates about the status of ethical judgments. My central claim is that there
are very close parallels between ethical investigation and that pursued in philosophy quite generally. These parallels provide excellent reason for rejecting
some of the main lines of criticism just rehearsed. They also generate positive
reasons to favor nonnaturalism over its competitors.
I locate the central claim within a central argument. Here it is:
1. Ethics is a species of enquiry; philosophy is its genus.
2. A species inherits the essential traits of its genus.
3. There are (among others) two essential traits of philosophy: the realistic
status of its truths, and its status as something other than a natural
science.
4. Therefore nonnaturalistic ethical realism is true.
In what follows, I will try to make the first premise plausible. I’ll just assume
the truth of premise (2). As for (3), I will say something (but not nearly enough)
to defend the claim that philosophy is something other than a natural science.
I will say a good deal more about why we should construe philosophical truths
realistically.
To see ethics as philosophy is to appreciate a certain kind and degree of
methodological similarity. Let me start with an admission: I don’t have a nice,
precise characterization to offer of the philosophical method. In large part,
⁵ I address the motivational problems in Part III, and the worries to do with practical reason in Part IV,
of Moral Realism.
216
Russ Shafer-Landau
I think philosophy is distinguished from other disciplines as much by the
questions it asks as by the methods it uses to pursue their answers. We know
philosophy when we see it, and can distinguish it, in practice, fairly easily from
any of the disciplines whose lines sometimes blur into its territory (religious
studies, anthropology, economics, psychology, sociology, etc.).
To get an approximation of what it is to pursue a philosophical question,
consider what would have to happen in order to answer the question of
whether there are universals. First, we’d have to remove ambiguities and fix our
subject, so as to avoid talking past one another. Then we would have to gather
and bring forward any empirical evidence that was thought to be relevant to
determining an answer to our question. We would have a careful look at
arguments already on offer, from both those whose views are largely in sync
with our own, and the best of our critics. We would test our claims for logical
consistency, and decide, in the face of contradiction, which of our competing
views is least well supported. And we would have to investigate to see whether
ours, among competitors, best exemplifies a host of theoretical virtues—economy,
stability across cases, avoidance of ad hoc assumptions, preservation of existing
beliefs, explanatory breadth, etc.
Every single stage of this process is fraught with controversy, since all but the
second (that of obtaining relevant facts) primarily involves philosophical
exploration. ‘Universal’ is not univocal. The appraisal of a sympathetic argument,
as well as a critic’s challenges, is no less a philosophical task. The consistency of
one’s views isn’t always transparent. The comparative assessment of whether one
view best exemplifies a theoretical virtue is often a very tangled affair, and in any
event is not to be resolved by appealing exclusively to the evidence of the senses.
So far as I can tell, things are just the same when it comes to ethical
investigation. When there is a disagreement, for instance, about the morality of
abortion, we must first ensure that we are not talking past one another. We have
to settle on the meaning of the relevant terms (e.g. what it is to be human, or a
person), so that we are on the same page. We must gain the relevant nonmoral
facts. We have to advance a view that is internally consistent, or amend it so as
to make it so. We try out the best of our arguments against the competition,
and vice versa, and make changes accordingly. We see whether our view best
comports with a battery of theoretical virtues. And, finally, the ultimate
resolution of the issue is not to be decided solely or even primarily by direct
appeal to empirical evidence. We rely on such evidence to tell us what is the
case. To know what ought to be the case, we need to push further. That is not,
in the first instance, an empirical matter.
Philosophy is not primarily an empirical discipline, but an a priori one. Its
truths are ordinarily discoverable, when they are, not exclusively by appeal to
Ethics as Philosophy
217
what our senses can tell us. We don’t bump into such things as universals, free will,
or modalities; we can’t see them, or hear or touch them. We may have reason to deny
the existence of such things, but not because we aren’t sure what they taste like.
Dismissing such things from our ontology, or ratifying their inclusion in it, is something that no scientist is able to do. Such things are dealt with in an a priori way.
At this point I’d like to invoke another of my promissory notes.
Substantiating the claim that fundamental philosophical truths are a priori is
work for a paper unto itself (at the least). This isn’t that paper. Yet this claim
about philosophy, while contentious, isn’t on the face of it that implausible. Of
course there are those who deny the very possibility or existence of a priori
knowledge. But for all others, basic philosophical principles should be quite
attractive candidates. Philosophy must run a close second to mathematics as an
exemplar of an a priori discipline (if indeed there are any such exemplars). Part
of this is explicable by reference to the metaphysically or conceptually necessary status of the principles that are the object of philosophical investigation.
And part of this is explicable by reflection on cases. Consider for a moment
Leibniz’s law of the indiscernibility of identicals, or the modal principle that
anything that is necessary is possibly necessary. These certainly don’t seem to be
inductive generalizations, or conclusions of inferences to the best explanation.
The role of sensory evidence in establishing such claims is peripheral, at best.
I might be mistaken about this, and nothing to come will absolutely protect
against this possibility. But the view that makes the justification of such principles a matter of empirical confirmation is (much) more contentious than the
one I am prepared to rely on.
As ethics is a branch of philosophy, we have excellent reason to think that
fundamental ethical principles share the same status as fundamental philosophical principles. When we want to know whether something is right or
wrong, admirable or vicious, we will certainly want to know what’s going on in
the world. The evidence of our senses may tell us that happiness has been
maximized, or that the words of a promise have been uttered, but that’s only the
beginning, not the end, of our ethical investigations. When trying to verify the
basic standards that govern the application of moral predicates, we will
only secondarily (if at all) advert to what the physicists and botanists and
hydrologists say. The conditions under which actions are right, and motives
and characters good, aren’t confirmed by the folks with lab coats. They are confirmed, if at all, by those who think philosophically. And much of that thinking, especially when focused on non-derivative, core principles, is undertaken
without clear reliance on what we can see, or hear, or touch.
Since doing ethics is doing a kind of philosophy, we shouldn’t be surprised at
the similarities just mentioned. But these similarities can insulate nonnaturalism
218
Russ Shafer-Landau
from some of the standard critical charges, enumerated above in section 2. In
what follows, I will rely on the parallels between the species (ethics) and its
genus (philosophy) in a way that aids nonnaturalist realists in answering three
of the most pressing objections against their views.
The first objection says that the intractability of ethical disagreement sustains an antirealist diagnosis of ethical thought and talk. The second criticism
claims that this disagreement in any event undermines any justified belief we
may have for our moral views, provided that they are meant to tell us about how
the world really is. The third asserts that the causal inefficacy of moral facts
provides excellent reason to deny their existence. Note that success in refuting
such criticisms is still short of a complete defense against the metaphysical and
epistemological objections I outlined above. But the criticisms I discuss are
some of the most important elements of these families of objections, and I think
that they can be met. That is work enough for a day, if it can be accomplished.
4. Moral disagreement as a metaphysical objection
Nonnaturalism stands for the idea that there are nonscientific moral properties. But if there are such things, why is there so much disagreement about
them? Many believe that objective properties of any kind must be such as to
garner consensus about (the conditions of ) their instantiation, at least among
people who are well situated to appreciate such things. But it doesn’t take an
expert to realize that such consensus is extremely elusive in ethics. So persistent
moral disagreement presents us with a choice. Perhaps there are no moral properties at all. Or there are, but ones that are not objective—rather, they are constituted by partial or parochial human attitudes. As such, the conditions of
their instantiation, and the instantiations themselves, would presumably be
empirically verifiable, at least in principle. Either way, the nonnaturalist loses.
There are really two ways to run this skeptical argument, though they
usually remain entangled in the literature. One is as an argument that seeks to
draw an inference to the best explanation, the explananda being the scope of
actual ethical disagreement we see in our world. The second is as an a priori
argument that has us anticipating persistent disagreement amongst even hypothetical, idealized moral deliberators. In both cases, the presence of intractable
disagreement is said to be sufficient to draw an antirealist conclusion: there are
no real, objective moral standards that could serve as guideposts to our moral
investigations. In ethics, we make it all up.
The first version of the argument, as an inference to the best explanation, is
inconclusive at best. Certainly there is intractable moral disagreement—plenty
Ethics as Philosophy
219
of it. But just as surely, such disagreement might be well explained as a product
of insufficient nonmoral information, or adequate information insufficiently
‘processed.’ Such processing failures cover a wide range of cases, from errors of
instrumental reasoning, to a failure of nerve, sympathy, empathy, or imagination. One explanation (not the only one) of these errors is that there’s typically
much more personally at stake in ethical matters than in scientific ones, and
these stakes tend to introduce biasing factors that skew correct perception. It
may be that for any given real-world ethical disagreement, we could cite at least
one of these failings as an explanation for its continued existence.
I think that one’s expectations of (lack of) consensus are largely an expression of
one’s antecedent metaethical commitments, rather than anything that could
serve as an independent argument in this context. Imagine away all of the failings
mentioned in the previous paragraph: will there or won’t there be any disagreement left to threaten moral realism? I’m not sure. If not, then the realist can rest
easy. But suppose disagreement persists, even in the counterfactual situation in
which we rid our agents of the flaws that impede correct moral reasoning. Even
here, however, realists can sustain their view with a minimum of damage. They
will have to say that impeccable reasoning may nevertheless fail to land on the
truth. There can be a gap between epistemic accessibility and truth. If we are to
posit an absence of consensus amongst even perfected enquirers, then the idealized picture of moral enquirers will fail to guard against their fallibility.
At this point we can introduce the ethics–philosophy parallel and use it to
defend nonnaturalism from the argument from disagreement. The breadth
and depth of philosophical disagreement is just as great as that found within
ethics (perhaps greater). There’s still no consensus on the merits of compatibilism,
the analysis of knowledge, or the relation of the mental and the physical. Nor
is there broad agreement about which methods are needed to unequivocally
confirm the right answers for us.
If the intractability of disagreement in an area is best explained by antirealist
assumptions about its status, then we must be global philosophical antirealists.
The judgments we render, and the arguments we offer on their behalf, must all
be seen either as incapable of truth, as expressions of conative commitments
only, or as claims whose truth is contingent on personal or interpersonal
endorsement. But that’s not a very plausible take on the status of our philosophical views. There is a truth—a real, objective truth—about whether the
mental is identical to the physical, or about whether certain kinds of freedom
are compatible with determinism. Once we are sure of our terms and concepts,
the judgments that affirm or deny the existence of such things are literally
either true or false, in as robust a sense as we can imagine. We don’t have the
final say about the truth of such judgments, and the content of these judgments
220
Russ Shafer-Landau
is indeed something other than whatever conative or practical commitments
contingently accompany them.
I invite you to reflect on the status of the philosophical judgments you hold
most dear, and have worked most carefully to defend. Do you imagine that
your views, and their supporting arguments, are either untrue, or possessed of
only the sort of minimal truth that is attainable by having been sincerely
endorsed from within a parochial perspective? No matter how skeptical you
might be about some alleged philosophical entities (universals, free will, or
moral facts), you presumably take your confident opinions about such matters
as having registered a real truth, one that is a function neither of your attitudes
towards it, nor of the language you have used to comprehend it. That truth, you
believe, is independent of the circle you inhabit, the agreements you’ve entered,
the conventions you are part of, and the era in which you find yourself.
And yet one’s philosophical views are bound to be as controversial as one’s
ethical views. Disagreements in core (and peripheral) philosophical areas are
apparently intractable. Empirical evidence hasn’t yet been able to solve any
major philosophical problem, and any prediction that it someday might is as
likely to divide philosophers as any other philosophical question. If intractable
disagreement about verdicts and methods is enough to warrant an antirealist
diagnosis of an area, then the whole of philosophy must be demoted. That simply is implausible: there really is (or isn’t) such a thing as probabilistic causation, numbers without spatio-temporal location, actions that are both free and
determined, etc. My say-so doesn’t make it so. Neither does anyone else’s.
The philosophical stance that denies the existence of nonnatural moral
properties is itself the subject of intractable disagreement. If such disagreement
is sufficient to undermine the realistic status of the controversial judgments,
then the views of the ethical naturalist or moral antirealist cannot be objectively
correct. They are either untrue, or are true reports of the attitudes they themselves take toward nonnaturalism, or are noncognitive expressions that reflect
their own practical commitments. If they are any of those things, then they
cannot rationally command the allegiance of their detractors. Nonnaturalists
needn’t be making any error when rejecting such views.
The alternative is to see our beliefs about such matters as aspiring to, and
possibly succeeding in, representing a philosophical reality not of our own
making. This reality is constituted by a set of truths whose alethic status is independent of our endorsement of their content. And this despite the presence of
intractable philosophical disagreement.
Of course, one might say that were we free of the shortcomings that beset all
of us actual enquirers, we would converge on a set of philosophical claims
about free will, causation, etc. The disputes that seem to us so intractable
Ethics as Philosophy
221
would vanish with more information, more efficient and comprehensive
application of that information, etc. That may be so. But then we have every
reason to render the same verdict in the ethical case. Since ethics is a branch of
philosophy, it would be very surprising to come to any other conclusion.
In other words, even if, at the end of the day, there is (much) more disagreement within ethics than there is in science, this is hardly probative evidence for
the falsity of nonnaturalism. For there is just as likely to be (much) more
disagreement in philosophy generally than there is in science. That doesn’t
license a verdict of global philosophical antirealism. It certainly doesn’t license
us in the claim that philosophy is a natural science. So long as it is true (as
almost every working philosopher presupposes) that there is an objectively
correct view about central philosophical puzzles, then we have just as much
reason to accord the same status to ethical matters. We will have our opinions
about such matters, but they are ultimately answerable to a truth not of our
own making. Things are no different in ethics.
5. Moral disagreement as an epistemic defeater
For any nontrivial moral view one holds, there are bound to be others who disagree with it. This very fact is probably not enough to undermine any epistemic
justification one may have for the belief. One might, after all, be unaware of the
disagreement, and this ignorance might be nonculpable. Yet what of the
ordinary situation, where we realize that our own moral views fail to command
universal allegiance? Suppose not only that you know of such disagreement,
but that you also rightly believe that your opponents, reasoning correctly from
their own incompatible but justified beliefs, will never come over to your side.
What does that do to the status of your own beliefs?
As I see it, such awareness does not, by itself, constitute a defeater of one’s
views. It does not entail that one ought to suspend judgment about what one
believes. For one may well think—and this is the usual case—that one has
justifying reasons that the other is failing to appreciate. That she is reasoning
impeccably from her own starting points does not mean that her beliefs must
be true, for her starting points may be way off base. And, as you will see things,
they almost certainly are.
Surely it is possible that any defense you offer of your contested views will
invoke other beliefs that are as controversial as the ones you are intending to
support. In fact, this happens all the time in moral discussions. Perhaps, for
many such cases, there is nothing one can do but beg the question. And
question-begging arguments never confer justification.
222
Russ Shafer-Landau
There are two things to say here. First, one’s belief might continue to be
justified, even if defending it to others has one begging questions. A belief ’s justification is distinct from an agent’s ability to justify it to others. So long as the
belief was initially justified, it is possible that its justification survives, despite
an agent’s inability to advance considerations that an audience finds compelling. (Someone rightly convinced that tomatoes are fruits might be justified
in her belief, even if she’s unable to bring others around to the idea.) Second,
there is excellent reason to believe that the presence of another’s incompatible,
justified belief doesn’t always undermine justification; indeed, there might
even be a case for thinking that question-begging arguments can supply
positive justification for one’s contested beliefs.
We can see this with the help of a series of examples. Suppose that you
are engaged in conversation with a principled fanatic. He thinks that the
fundamental ethical imperative is to gain power over others; everything else is
subsidiary to this primary goal. Any argument you offer for beneficence is
bound to be treated as the product of an effective brainwashing. Nothing
you can say will convince him. Moreover, suppose that he’s not contradicting
himself, and isn’t making any false empirical claims to support his ultimate principle. In the context of your conversation, you are bound to beg the question.
But you might be justified in your beliefs anyway. For the presence of
an intelligent, consistent, and indefatigable opponent does not necessarily
undermine a belief that one is otherwise justified in holding. This is a general
point. It holds for one’s ethical views, but also for perceptual, memorial, and
philosophical ones, as well.
To simplify, consider a case in which one’s perceptual beliefs later form the
basis of a memorial belief. I saw and remember talking to my hated nemesis
Smith the moment before he made that fatal misstep that no one else witnessed. I try to convince others of what I have seen, and am met with disbelief.
Let’s add that my relations with Smith are well known. That others have
excellent reason to doubt my word is compatible with my original belief, and
its memorial descendant, both being highly justified. In this case, not only do
the incompatible, well-justified beliefs of others fail to undermine my justification, but my own question-begging attitudes (e.g. regarding my own innocence
in this case) do appear to be enough to constitute positive justification for the
beliefs I hold.
We can broaden the picture in an obvious way. Informed, rational, and
attentive skeptics, possessed of internally consistent and coherent attitudes,
might remain unconvinced by any of our empirical claims. According to
this version of the argument from disagreement, that resistance defeats any justification we might have for our empirical beliefs. Though we can’t absolutely
Ethics as Philosophy
223
discount that possibility, the conclusion is so drastic as to call into question the
soundness of the argument that generated it. If we assume, as everyone reading
this will, that we do have some positively justified empirical beliefs, then, so far
as I can tell, it follows that question-begging grounds can confer positive justification. For anything one might cite as evidence on behalf of one’s empirical
beliefs will surely be regarded as question-begging by the skeptic.
A similar story can be told regarding all of our philosophical beliefs. The
most brilliant philosophers, rational, open-minded, and well informed, have
failed to agree amongst themselves on just about every key philosophical issue.
If pervasive and intractable disagreement signaled an absence of justification,
this would mean that none of those philosophers (much less the rest of us)
would be at all justified in holding the philosophical views that they (we) do.
But this seems false; it’s certainly belied by anyone who actually undertakes
sincerely to argue philosophically. One who has developed a theoretically
sophisticated take on some philosophical issue, coming to grips with deep
criticisms and developing novel and integrated positive proposals, is surely justified to some extent in thinking her views correct. Of course such a person will
see that some others will fail to be convinced—even some others who are as
smart, ingenious, and imaginative as she is. She will recognize her fallibility and
appreciate a salient feature of philosophical history—namely, the failure of
greater minds to attract even near-unanimity on most of the major points that
they had advanced. Still, awareness of this history, and the skepticism of some
of her contemporaries, is not enough to force her to suspend judgment on the
views that she has so skillfully defended.
I see no reason to register a different verdict for ethics. Deep disagreement
there, as elsewhere, should give one pause. It can sap one’s confidence, and if it
does, then that (but not the disagreement per se) may be sufficient to undercut
one’s justification. But this is no different from the general case. Provided that
one brings to a dispute a moral belief that is justified, then exposure to conflicting belief needn’t defeat one’s justification, even if one is unable to convince
an intelligent other of the error of his ways.
The present argument against the epistemic justification of moral belief
relies on the following principle (or something very like it):
(E) If (i) S believes that p, and R believes that not-p, and (ii) S and R know
of this disagreement, and (iii) S and R have formed their beliefs in
rational and informed ways, then S is not justified in a belief that p,
and R is not justified in a belief that not-p.
(E) may be true. But no one could be justified in believing it. (E) itself is the
subject of intractable disagreement—there are informed and rational people
224
Russ Shafer-Landau
who endorse it, and equally qualified agents who reject it. By its own lights,
then, we must suspend judgment about (E). Having done that, however, we are
no longer epistemically forbidden from positively embracing a contested
belief, even if our opponents are as smart as we are.
We can reveal another kind of skeptical self-defeat if we renew our emphasis
on establishing a parity between ethical investigation and philosophical investigations generally. A familiar skeptical line is that there isn’t, really, any adequate
evidence that can be called upon to support our ethical opinions. Unlike
empirical investigations, we haven’t anything tangible that can, at the end of
the day, finally settle a disputed moral question. All the sensory evidence at our
disposal will underdetermine an ethical verdict. And what’s left? Only our
emotional responses and our moral convictions, both of which are traceable to
accidents of birth and upbringing. Their genesis marks them as unreliable indicators of any truth there might be. But there’s nothing else to rely on in ethics.
And therefore our moral views lack justification, one and all.
The problem with such an argument should by now be apparent. There is a
striking equivalence between the nature and source of our evidence in philosophy, and in ethics. We have no choice but to rely on our intuitions and considered judgments in both. What tells us, for example, that many proposed analyses
of knowledge are no good is not some empirical finding that scientists have
unearthed. It is instead our conceptual intuitions about counter-examples. If we
want to know whether determinism is compatible with free will, we will consult
arguments that invariably appeal to our intuitive responses to hypothetical cases.
If such convictions and responses have no evidential credibility, then we should
have to regard all philosophical beliefs as unjustified. Perhaps they are. But then
those of the ethical naturalist, and the moral antirealist, are similarly undone.
6. The causal inefficacy of moral facts
Gilbert Harman (1977: ch. 1) has famously charged that moral facts are
causally inert, and are therefore best construed antirealistically. His brief
remarks have inspired a minor cottage industry, most of whose workers, it
seems to me, are headed in the wrong direction. Rather than try to establish
that point with a survey of the literature, I will try for the larger picture, with
the aim of arguing that Harman’s basic line of attack is misdirected.
Harman doesn’t put things in quite this way, but I think his position, and that
of many who take his lead, can be accurately captured in the following argument:
1. If something exists, and its existence is best construed realistically, then it
must possess independent causal powers.
Ethics as Philosophy
225
2. Moral facts possess no independent causal powers.
3. Therefore either moral facts don’t exist, or their existence isn’t best
construed realistically.
Harman himself believes in moral facts, though he regards them as artefacts of
social agreements. He is an ethical relativist, not a realist.
Since the argument is valid, any realist must choose either or both of the
premises to come in for criticism. I opt for (1), because I suspect that (2) is true.
A property has independent causal powers only if its instantiation by itself has
causal implications, apart from the instantiation of any other properties it may
in an instance depend on or be realized by. I’m not confident that moral facts
possess such powers.
I won’t try to vindicate my lack of confidence here. If it is misplaced, then so
much the better for moral realism. Moral properties would possess independent
causal power, and thereby pass the most stringent test for ontological inclusion.
But let’s instead imagine that my suspicion is correct, and that we are thus
placed in what many have considered a worst-case scenario: trying to defend
the existence of moral properties, realistically construed, while acknowledging
that they are fundamentally different in kind from the properties whose
existence is ratified by the natural sciences. If I am right, then such things as a
benefactor’s generosity, a regime’s injustice, a friend’s thoughtfulness, are causes
(if they are) only by virtue of inheriting the causal powers of the properties that
realize them at a time. Any causal power they have is exhausted by that of the
subvening properties that fix a situation’s moral character. Nothing follows
from this admission unless we are also prepared to insist on a causal test of ontological credibility, of the sort espoused in Harman’s first premise.
Such a test is powerfully motivated, but is ultimately resistible. This test is an
application of Occam’s razor, and is responsible for our having pared down our
ontology in many sensible ways. We’re quite finished with explanations that
invoke Osiris or golems or centaurs, and Occam’s razor is responsible for that.
All that these entities were once invoked to explain can be more parsimoniously
explained by relying on facts whose existence is vindicated through scientific
confirmation. And such confirmation makes essential reference to a putative
entity’s causal powers.
So out with the trolls, the ancient pantheon, and the vampires. That’s not so
bad, is it? Such things aren’t required to explain the goings-on in our world. But
then, by my admission, neither are moral facts. So, by parity of reasoning,
either we keep moral facts, but at the expense of a bloated ontology that
implausibly lets these minor supernatural agents sneak back in, or we abolish
the lot of them. Why should morality get special treatment here, when, as we
226
Russ Shafer-Landau
all agree, the causal test has done its good work in so many other areas? Very
conveniently for me, I don’t have the time in this context to provide the full
answer to this question.⁶ But in lieu of that long story, let me offer a brief reply,
and then a longer one that invokes the ethics–philosophy parallel that I have
already relied on.
The brief reply: application of the causal test has highly counter-intuitive
implications. This is so on two assumptions: first, that only physical properties
possess independent causal powers, and second, that at least most of the properties of the special sciences are not identical to, but only supervenient upon,
those of physics. From these assumptions, allied with the causal test, it follows
that nothing exists but (roughly) atoms and the void. There certainly won’t be
any such things as atmospheres, rock strata, newts, and dandelions, if we grant
that such things are not type-identical to anything referred to in a physics journal. It seems to me that such things do exist, and are multiply realizable by,
rather than identical to, particular physical properties. Thus the causal test
eliminates too much from our ontology.
Suppose that doesn’t faze you—you can live with such a parsimonious ontology,
or you don’t endorse one of the two assumptions that got us there.⁷ Still, we can
invoke the ethics–philosophy parallel in the service of a further argument that
should worry proponents of the causal test. By way of introduction, we can
note that moral facts are a species of normative fact. Normative facts are those
that tell us what we ought to do; they rely on norms, or standards, for conduct
within a given realm. Normative facts cause nothing of their own accord.
We can be helped to see this by comparing ethics, not to philosophy as a
whole, but to one of its close philosophical cousins. In my opinion, moral facts
are sui generis, but they are most similar to another kind of normative fact—
epistemic facts. Epistemic facts concern what we ought to believe, provided
that our beliefs are aimed at the truth. Once one understands the concept of
logical validity, then if confronted with a modus ponens argument, one ought
to believe that it is logically valid. This is a true epistemic principle.
It’s also the case that you oughtn’t believe things that you have no evidence
for, and much evidence against. What does this epistemic truth cause?
Nothing. Nor are particular, concrete epistemic duties—duties had by agents
at a time—at all independently causally efficacious. Epistemic facts have as
their primary function the specification of standards that should or must be
met. We can say, if we like, that such standards are descriptive—they describe
⁶ I try my hand in Moral Realism, 98–114.
⁷ Beware: arguments for rejecting either assumption may well allow moral facts to pass the causal test
(though any sound argument rejecting the second assumption would generate a naturalistic, rather than a
nonnaturalistic, moral realism).
Ethics as Philosophy
227
the conditions under which agents are (e.g.) appropriately sensitive to
evidence, justified in their beliefs, warranted in their views. But, unlike natural
scientific principles and facts, such normative standards may be perfectly
correct even if they are honored only in the breach. The epistemic requirement
that we proportion our beliefs to the evidence can be true even in a world populated wholly by spell-casters and astrologers. The normative facts that specify
the conditions under which we ought to believe the truth, or behave morally,
lack the ability to explain the workings of the natural order. Our epistemic and
moral duties cannot explain why apples fall from trees, why smallpox takes its
victims, why leopards have their spots. But they may exist for all that.
Nor is this failure something specific to the moral or epistemological realms.
Consider prudential or instrumental duties—those that require us to enhance
self-interest or efficiently satisfy our desires. Such normative demands do not
explain what goes on in the world. Alternatively, if they are thought, for instance,
to be powerful enough to explain why agents act as they do, then surely moral and
epistemic requirements are capable of doing so as well. I see no basis for distinguishing the causal powers of any of these normative types from one another.
I don’t mean to suggest for a moment that the causal test is useless. Rather,
I think we should recognize its limits. The causal test fails as a general ontological test: it doesn’t work when applied to the normative realm.
Scientific principles are vindicated, when they are, because they are able to do
two closely related things: cite the causes of past events, and accurately predict
the nature and occurrence of future events. Their claim to be genuinely explanatory depends almost entirely on their ability to discharge these two tasks.
But moral rules are not like that. We can construe rules in this way: Brink,
Sturgeon, and others manage this feat.⁸ But it’s not a very natural way to regard
them. Moral principles aren’t viewed in the first instance as hypotheses that
predict the actions of agents, but rather as requirements that everyone knows
will encounter predictive failures. True, moral principles will reliably predict
the doings of good and bad agents. But that presupposes the reality of moral
properties (goodness and badness), and there’s no reason to make such a concession at this stage, especially given the seriousness of antirealist charges, and the
proper aim (given a naturalistic vantage point) of beginning from a neutral perspective and relying on the causal test as a way to determine the nature of reality.⁹
⁸ See e.g. Brink 1989: 182–97, Sturgeon 1984, Boyd 1988, Railton 1986.
⁹ So in this respect I think that Harman was wrong to concede to his opponents the existence of moral
facts. The proper starting point for an antirealist is one in which we suspend judgment on the existence of
such facts, and demand of the realist some positive arguments for believing in them. Harman instead was
willing to grant the existence of moral facts, but claimed that even so they possessed no independent causal
powers, and so could not be construed realistically.
228
Russ Shafer-Landau
Yes, we can enshrine moral predicates within true counterfactuals, even (in
some cases) counterfactuals of greater generality than those describable at the
physical level. But that is no proof of moral realism, as we can do the same for
the predicates of etiquette and the civil law, which obviously cannot be construed realistically. Moral principles and facts aren’t meant to explain behavior,
or anticipate our actions, but rather to prescribe how we ought to behave, or
evaluate states or events. They don’t cite the causes of outcomes, but rather
indicate what sort of conduct would merit approval, or justify our gratitude, or
legitimate some result. Science can’t tell us such things.
If I am right, then an allegiance to the causal test entirely eliminates the
normative realm. But this is highly implausible. There are reasons to believe
things, reasons to satisfy one’s desires, reasons to look out for oneself. There are
also moral duties to aid others and refrain from harming them, even if doing so
isn’t going to improve one’s lot in life. The standards that supply such reasons
are not capable of causing anything. Nor, it seems, are the reasons or obligations themselves. (Again, if they are, all the better for moral realists.¹⁰) If there
is any such thing as a genuine reason, the test must fail. Alternatively, if the test
is retained, then such reasons must be capable of passing it. And then the causal
argument against moral facts evaporates.
Maybe we can have our cake and it eat, too? Why not retain the causal test,
allow that normative facts exist, but view them, as Harman does moral facts, as
by-products of human choice and election? The causal test is a realist’s test.
Failure to pass it doesn’t mean that a putative fact doesn’t exist. It just means that
the fact cannot be construed realistically. Normative facts may be like this. If so,
we could retain the test, and also retain a global normative antirealism. Perfectly
in keeping with the physicalist leanings of so many of our contemporaries.
The animating spirit behind the causal test is the ontological principle that
the real is limited to what is scientifically confirmable, and the epistemic principle that we have good reason to believe in something only if it impinges on
our experience, or is required in the best explanation of that which does. The
causal test obviously supports, and derives support from, both the ontological
and the epistemic principle. Yet both principles are dubious. The case for the
causal test is considerably diminished once we see why.
The epistemic principle is problematic because it invokes an entity—a good
reason—whose existence is not itself scientifically confirmable. It’s like saying
¹⁰ One natural line of thinking seems mistaken. That is the one that attributes causal power to normative
facts by noting that they, as mediated by an agent’s beliefs and desires, can explain a great deal of why agents
act as they do. That something is seen by an agent as (prudentially, epistemically, or morally) obligatory is
often enough to get one going. But that sort of mediation undermines the independence of the causal
potency that the test requires. (This is one of Crispin Wright’s points (1992: 196).)
Ethics as Philosophy
229
that God sustains a universe that contains no supernatural beings. There’s a
kind of internal incoherence here: the claim discounts the existence of the kind
of thing that is presupposed by the claim itself.
Further, a belief ’s being justified is not the sort of thing that impinges on our
experience. Nor, seemingly, is reference to its epistemic status required to
explain anything that we have ever observed. But then, by the epistemic principle under scrutiny, we have no good reason to think that there is any such
thing as the property of being epistemically justified. But if there is no such
property, then the principle that implies such a thing cannot itself be justified.
And so we can be rid of it.
Here’s another way to get to the same result. We needn’t make essential
reference to this epistemic principle to explain why we see or hear or feel the
things we do. Nor, so far as I can see, is any epistemic principle required in the
best account of why various observable events have occurred in the world. So if
the principle is true, then we lack a good reason for thinking it so. This principle, like normative standards quite generally, seeks to regulate and appraise
conduct, rather than to describe its causal antecedents or powers. If that’s
sufficient to render it unreal, or sufficient to remove any justification we might
have for believing it, then it can’t rightly be used to constrain our epistemic
findings or practices.
And the ontological view? The relevant ontological principle tells us that the
only existential truths there are (i.e. truths about what exists) are those that are
scientifically confirmed. This is certainly false if we are concerned with science
as it stands, as some existential truths have yet to be discovered. Yet the view is
no more plausible if we are envisioning the edicts of a perfected natural science.
Here’s why. Consider this existential claim:
(O) There are no existential truths other than those ratified by perfected
natural sciences.
Either (O) is true or false. If false, let’s drop it: our ontology wouldn’t then be
entirely fixed by the natural sciences. But if it’s true, then it must be false: it’s
self-referentially incoherent. For (O) cannot itself be scientifically confirmed.
If it were true, it would be an instance of a nonscientifically confirmable
existential truth. Thus either way we go, (O) must be false.
(O) is a thesis from metaphysics, not physics. Philosophers, not natural
scientists, are the ones who will end up pronouncing on its merits. This is another
application of the general idea that there are specifically philosophical truths
that escape the ambit of scientific confirmation. There might be abstract entities, or such a thing as conceptual necessity, justified belief, or goodness. Bring
your beakers, your electron microscopes, your calculators and calipers—you’ll
230
Russ Shafer-Landau
never find them. You can’t abolish such things just because they lack
independent causal power, and so escape empirical detection. After all, the
principle calling for such abolition isn’t itself scientifically confirmable.
In the end, the absence of independent causal power is not a good reason to
deny the existence of moral facts, realistically construed. Of course, nothing
I’ve said in this section supplies any argument for thinking that there are such
things. I doubt that causal considerations could do that. But undermining their
role in antirealist arguments can go some ways towards removing a familiar
barrier to justified belief in the sort of nonnaturalism that I find appealing.
7. Conclusion
Once we attend to the fact that ethics is a branch of philosophy, the plausibility of nonnaturalistic moral realism is greatly enhanced. Philosophy is not a
natural science. Basic, fundamental philosophical principles are realistic in
nature. And central ethical principles are philosophical ones. This combination of claims gives us excellent reason to suppose that fundamental ethical
truths are best construed realistically, and nonnaturalistically.
This seems to me to be a very powerful argument that can aid the nonnaturalist realist in replying to a number of perennial criticisms. One such
criticism—that persistent, intractable moral disagreement is best explained
as antirealists would do—can be met once we avail ourselves of the ethics–
philosophy parallel. Moral disagreement shares all structural features with
philosophical disagreement generally, and yet a global philosophical antirealism
is very implausible. Moral disagreement also fails to provide a strong epistemic
defeater for one’s own already-justified moral beliefs. Controversial philosophical beliefs might be justifiedly held; things are no different in the specifically
moral domain. And the causal inefficacy of moral facts can be admitted without threatening moral realism, since the causal test is too restrictive a standard
for ontological credibility. Alternatively, if (contrary to my suspicions) moral
facts do manage to pass that test, then retaining the test will entitle moral facts
to admission into our ontology.
Once we attend to the fact that ethics is a branch of philosophy, a defense of
nonnaturalistic moral realism becomes a bit easier than it otherwise might be.
Which is not to say that it’s at all easy—the many promissory notes scattered
throughout this essay will attest to that. Still, reliance on the ethics–philosophy
parallel enables us to plausibly respond to some of the critical obstacles to the
development of a plausible ethical nonnaturalism. We can hardly hope to vindicate a complex metaethical theory in one fell swoop. We can, if the preceding
Ethics as Philosophy
231
arguments are any good, manage to dust off a neglected view and show that
some of the sources of its unpopularity have been overrated. I hope to have
done that here.
References
Audi, Robert (1997). Moral Knowledge and Ethical Character. Oxford: Oxford
University Press.
—— (1999). ‘Self-Evidence,’ Philosophical Perspectives, 13: 205–26.
—— (2004). Kantian Intuitionism. Oxford: Oxford University Press.
Ayer, A. J. (1936). Language, Truth and Logic. London: Gollancz.
Blackburn, Simon (1993). Essays in Quasi-Realism. Oxford: Oxford University Press.
—— (1998). Ruling Passions. Oxford: Oxford University Press.
Boyd, Richard (1988). ‘How to Be a Moral Realist,’ in Sayre-McCord (1988).
Brink, David (1989). Moral Realism and the Foundations of Ethics. Cambridge:
Cambridge University Press.
Copp, David, and Zimmerman, David, eds. (1984). Morality, Reason and Truth.
Totowa, NJ: Rowman & Littlefield.
Garner, Richard (1994). Beyond Morality. Philadelphia: Temple University Press.
Gibbard, Allan (1990). Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard
University Press.
Hare, Richard (1952). The Language of Morals. Oxford: Oxford University Press.
—— (1963). Freedom and Reason. Oxford: Oxford University Press.
Harman, Gilbert (1977). The Nature of Morality. Oxford: Oxford University Press.
Jackson, Frank (1998). From Metaphysics to Ethics. Oxford: Oxford University Press.
Joyce, Richard (2001). The Myth of Morality. Cambridge: Cambridge University
Press.
Mackie, John (1977). Ethics: Inventing Right and Wrong. New York: Penguin.
Railton, Peter (1986). ‘Moral Realism,’ Philosophical Review, 95: 163–207.
Sayre-McCord, Geoffrey (ed.) (1988). Moral Realism. Ithaca, NY: Cornell
University Press.
Skorupksi, John (1999). ‘Irrealist Cognitivism,’ Ratio, 12(4): 436–59.
Stevenson, Charles L. (1937). ‘The Emotive Meaning of Ethical Terms,’ in
Stevenson (1963).
—— (1948). ‘The Nature of Ethical Disagreement,’ in Stevenson (1963).
—— (1963). Facts and Values. New Haven: Yale University Press.
Stratton-Lake, Philip (ed.) (2002). Ethical Intuitionism: Re-evaluations. Oxford:
Oxford University Press.
232
Russ Shafer-Landau
Sturgeon, Nicholas (1984). ‘Moral Explanations,’ in Copp and Zimmerman
(1984).
—— (2003). ‘Moore on Ethical Naturalism,’ Ethics, 113: 528–56.
Timmons, Mark (1999). Morality without Foundations. Oxford: Oxford University
Press.
Wright, Crispin (1992). Truth and Objectivity. Cambridge, Mass.: Harvard
University Press.
11
The Legacy of Principia
Judith Jarvis Thomson
I
The influence of G. E. Moore’s Principia Ethica on twentieth-century moral
philosophy is largely due to the attractiveness of a certain argument in it, which
came to be known as the open question argument. Let us step back for a
moment. Moore thought that the following thesis was obvious and in need of
no argument:
(Moore’s Premise)
There is such a property as goodness.
Surely there is a property that all and only good things have in common; that is
the property goodness.
Here now is the premise of the open question argument:
(OQA Premise) Whatever natural property NP you fix on, it is an open
question whether things that have NP also have goodness.
The conclusion of the argument is:
(OQA Conclusion) There is no natural property NP such that NP is
identical with the property goodness.
It is a very attractive idea that this argument is valid and, indeed, that its
premise is true.
Moore, of course, accepted the open question argument and, therefore,
rightly, drew the following further conclusion:
(Moore’s Conclusion) The property goodness is a non-natural property.
Alas, the idea that there are non-natural properties looked at best suspect to
many of his readers. But how were they to avoid Moore’s Conclusion? The
234
Judith Jarvis Thomson
readers I will be focusing on accepted the open question argument and
therefore had to, and did, reject Moore’s Premise.¹
Then what is done by a person, say Alfred, who says ‘A is good’? They
thought that if Alfred attributes any property at all to A, then the property he
attributes to it is the property goodness. Having agreed that there is no such
property, they concluded that Alfred does not attribute a property to A. They
therefore concluded that what Alfred says has no truth-value.
They differed, however, in their positive accounts of what Alfred does. I will
focus on one of the positive accounts in particular, namely A. J. Ayer’s.² Ayer
held that in saying what he says, Alfred merely expresses—that is, merely
displays—a favorable attitude toward A, thus he does no more than he would
have done had he instead merely grinned or said ‘Hurrah!’ in response to A.
(See Ayer 1946.)
This idea can helpfully be seen as an idea about why the open question
argument succeeds. Thus: it is an open question whether a thing that has NP is
good because saying that a thing has NP is not expressing a favorable attitude
toward it whereas saying it is good is. I am sure that the fact that the idea can be
so seen contributed to its popularity, since what helps to explain the success of
an attractive argument inherits attractiveness from it.
What we have so far is a pair of ideas about goodness. They are obviously generalizable, however. First, the open question argument is itself generalizable.
Moore had said that for any natural property NP, it is an open question whether
things that have NP also have goodness. So similarly, presumably, for badness.
And for rightness and wrongness. So similarly for all normative properties. It
follows that if there are such properties, they are non-natural properties. That
they are is a suspect idea, however. So we had better conclude that there are no
normative properties. It follows that one who says a normative sentence does
not attribute a normative property to anything. So we had better conclude that
one who says a normative sentence attributes no property at all to anything.
So we had better conclude that what such a person says has no truth-value.
Thus we had better accept:
(No Normative Truth-Value Thesis) Normative sentences have no truthvalues.
¹ In recent years, some people who resist Moore’s Conclusion have instead accepted Moore’s Premise and
rejected the open question argument; these people have come to be known as the Cornell realists. See e.g. the
papers by Sturgeon and Boyd in Sayre-McCord 1988. I return to them only in n. 14.
² I bypass C. L. Stevenson’s account, according to which what Alfred says means ‘I approve of A; do so as
well’ (see Stevenson 1944), since Ayer’s account still has a following, whereas Stevenson’s, I believe, does not.
The Legacy of Principia
235
Then what does a person do who says a normative sentence? We had better
accept:
(Expressivism) What the speaker does is merely to express a favorable or
unfavorable attitude—he does no more than he would
have done had he merely said ‘Hurrah!’ or ‘Boo!’
This pair of ideas was a major presence throughout twentieth-century moral
philosophy. Like them or not, nobody could ignore them. That is the legacy of
Principia.
II
More precisely, the legacy of Principia is the pressure brought to bear by the
open question argument, in virtue of which this pair of ideas seemed attractive
to their friends and a threat to their enemies, and thus could not be ignored.
That pressure continues to be felt, despite the criticism the ideas were subjected
to. I will describe only the two objections that have been taken most seriously;
both are objections to the No Normative Truth-Value Thesis.
The first was made by Peter Geach. (See Geach 1972: 8.1 and 8.2.) Geach
drew attention to the fact that normative sentences appear in what look like
truth-functional contexts, such as ‘If Hitler is a villain, then we ought not
put up a statue to him.’ Indeed, they appear in what look like valid arguments,
such as:
If Hitler is a villain, then we ought not put up a statue to him.
Hitler is a villain.
Therefore, we ought not put up a statue to him.
But an argument is an ordered set of things with truth-values. If normative
sentences have no truth-values, then that is not an argument, and a fortiori it is
not a valid argument.
Friends of the No Normative Truth-Value Thesis responded to this objection by recommending that we relax about what an argument might consist in,
and allow that a set of things without truth-values might be, well, not strictly
speaking a valid argument, but something just as good. (See Blackburn 1984
and Gibbard 1990. Their more recent responses may be found in Blackburn
1998 and Gibbard 2002.) We will return to Geach in the following section.
The second of the two objections emerged from what has come to be called
minimalism about truth—minimalism for short. Roughly expressed, minimalism
Judith Jarvis Thomson
236
tells us that the results of inserting any declarative sentence in for ‘S’ in the
schema
(T-Schema)
‘S’ is true if and only if S
are all true.³ Many people think this idea very plausible. Surely ‘Grass is green’
is true if and only if grass is green, and so on and on.
Here now is the second objection to the No Normative Truth-Value Thesis.
Assume minimalism. The following is obviously true:
(1) ‘Hitler is a villain’ is a declarative sentence.
From minimalism, we can conclude:
‘Hitler is a villain’ is true if and only if Hitler is a villain.
No civilized person denies:
(2) Hitler is a villain.
It follows that ‘Hitler is a villain’ is true and, therefore, has a truth-value. So if
minimalism is true, then the No Normative Truth-Value Thesis is false.
Many Expressivists were moved by this argument and, therefore, gave up on
the No Normative Truth-Value Thesis—but did not give up on Expressivism.
For they pointed to the fact that what entitles us to pass from minimalism to
the conclusion that ‘Hitler is a villain’ is true, and therefore has a truth-value, is
the conjunction of (1) and (2); and they said, rightly, that asserting the
conjunction of (1) and (2) is entirely consistent with asserting Expressivism.
(See, for example, Blackburn 1998 and Timmons 1999.)
But they might have rejected the argument. For they might have drawn
attention to the fact that the statement of minimalism that I supplied above
won’t do at all. It just is not the case that the results of inserting any declarative
sentence in for ‘S’ in the T-schema are all true: a sane minimalist imposes
constraints on candidate replacements for ‘S’ beyond that they must be declarative sentences. Thus a candidate must not contain a pronoun, for no sane
minimalist wishes to be committed to the truth of ‘ “I am a Hungarian” is true
if and only if I am a Hungarian.’ Similarly for all other indexicals, such as
demonstratives (‘this’) and tenses (‘will be’). Similarly for ‘liar sentences’.
Similarly for adjectives like ‘big.’ ‘Alfred is big’ is a declarative sentence,
contains no indexicals, and is not a liar sentence. Yet no sane minimalist wants
³ Minimalism tells us something else as well, namely that there is no property of being true—the truth of
those biconditionals is ‘all there is to truth’. I ignore this further claim in the text: I pretend in the text that
minimalism tells us only about the T-Schema, or, as the point is sometimes put, that truth is a disquotation
device.
The Legacy of Principia
237
to be committed to the truth of ‘ “Alfred is big” is true if and only if Alfred is
big.’ Alfred may be a big strawberry but a small piece of fruit. Is Alfred big?
There just is no answer to this question, for ‘Alfred is big’ is semantically
incomplete.
It emerges that no set of constraints on replacements for ‘S’ is strong enough
unless it guarantees that a candidate replacement has a truth-value. Thus that
the reason why sentences containing indexicals, liar sentences, and sentences
whose grammatical predicate is ‘is big’ are excluded is that, for this or that
reason, they lack truth-values.
The fact that the set of constraints on replacements for ‘S’ must be this
strong could have been seen straightway, for the results of making replacements
in the T-Schema are themselves true only if what follows ‘if and only if ’ in them
itself has a truth-value.⁴
So suppose we instead opt for what might be called constrained minimalism,
namely the thesis that the results of inserting any sentence with a truth-value in
for ‘S’ in the T-Schema are all true. There remains a very good question why
that is so, but that it is, is plain enough.
But constrained minimalism cannot be thought to provide a reason for
thinking that this or that sentence has a truth-value. It can be thought to supply
only an account of the conditions under which a sentence—which is
independently establishable as having a truth-value—is true. So an Expressivist
who wishes to defend the No Normative Truth-Value Thesis can say: your
argument that ‘Hitler is a villain’ has a truth-value relies on the assumption that
that sentence is an acceptable replacement for ‘S’ in the T-Schema. But it is so
only if there is sufficient independent reason to think it has a truth-value. So
your argument is circular, and not only makes no trouble for Expressivism, it
makes no trouble for the No Normative Truth-Value Thesis. (This point was
made by Jackson et al. 1994.)
III
In sum, the objection to the No Normative Truth-Value Thesis from
minimalism fails because minimalism is false; and an objection to the No
Normative Truth-Value Thesis from constrained minimalism fails because
it is circular. Let us return to Geach’s objection to the No Normative
Truth-Value Thesis.
⁴ This fact obviously makes trouble for the further claim made by minimalists that I pointed to in n. 3.
Michael Glanzberg argues in Glanzberg 2002 that that further claim is deeply false.
238
Judith Jarvis Thomson
Geach took it that the fact that normative sentences appear in what look like
truth-functional contexts, such as the conditional ‘If Hitler is a villain, then we
ought not put up a statue to him,’ is itself reason to think that those sentences
have truth-values. Others pointed to the fact that those sentences appear in
non-truth-functional contexts too. ‘I believe that—or wonder whether, doubt
that, regret that, . . . —Hitler is a villain’ are all perfectly respectable sentences.
As the point here is sometimes put, normative sentences are embeddable in
contexts that might be thought to require sentences that have truth-values.
But Geach drew attention, not merely to complex sentences, but also to the
fact that normative sentences appear in what look like valid arguments. The
example I gave above certainly looks like a valid argument. But what looks
like the explanation of its validity is modus ponens, and that means that the
example does not bring out the full strength of Geach’s objection. Consider the
following:
All Greeks are villains.
Hitler is a Greek.
Therefore, Hitler is a villain.
Surely this is yet another valid argument. The validity of this argument,
however, does not turn on the embeddability of normative sentences but,
rather, on what is inside the sentences.
The fact is that the normative grammatical predicate ‘is a villain’ is a logical
predicate: it has the inferential role of a predicate.⁵ An argument now emerges.
From
(P) ‘Is a villain’ is a logical predicate,
we may conclude
(C1) There is such a property as being a villain.
Therefore,
(C2) If Alfred says ‘Hitler is a villain,’ he thereby attributes a property to
Hitler, namely the property ‘being a villain.’
Therefore,
(C3) What Alfred says has a truth-value, namely true if Hitler has the
property and false if he does not.
⁵ To the best of my knowledge, Paul Horwich is the only minimalist who has taken seriously this fact
about normative grammatical predicates like ‘is a villain.’ See Horwich 1993. Not surprisingly, perhaps, for
there is nothing plausibly describable as ‘minimal’ about the fact that a given expression is a logical predicate.
The Legacy of Principia
239
For this argument to reach (C3) is for it to yield that the No Normative
Truth-Value Thesis is false. For it even to reach (C2) is for it to yield that
Expressivism is false, since according to (C2), Alfred, who says ‘Hitler is a
villain,’ does do more than Bert, who says ‘Boo to Hitler!’—Alfred attributes a
property to Hitler and Bert does not.
IV
Does the argument really reach (C3)? That is, are we really entitled to pass from
the fact that a grammatical predicate ‘is P’ is a logical predicate and, therefore,
stands for a property, to the conclusion that ‘A is P’ has a truth-value?
Unfortunately there is reason to think we are not.⁶ Michael Glanzberg argues
(in Glanzberg 2002) that people who suppose that the fact that ‘is P’ is a logical
predicate warrants the conclusion that ‘A is P’ has a truth-value will find
themselves unavoidably enmeshed in the Liar Paradox. So the argument would
need to be enriched by the addition of another premise if it were to reach (C3),
and it is, alas, quite unclear what that new premise should be.
On the other hand, we should be clear that what makes the No Normative
Truth-Value Thesis safe from the argument is not the fact that the argument
concerns itself with a normative predicate in particular: an analogous
argument concerning the non-normative predicate ‘is a banana’ would have
been no more successful. So while an Expressivist who wishes to defend the No
Normative Truth-Value Thesis can say ‘Fine, the job’s done, so far as that
objection goes,’ he was not the one who did it.
V
Does the argument even reach
(C2) If Alfred says ‘Hitler is a villain,’ he thereby attributes a property to
Hitler, namely the property ‘being a villain’?
If it does, then Expressivism is not safe from it. There is room for an objection
to the effect that it does not.
⁶ A less serious reason to think we are not lies in the fact that ‘is a villain’ is on any view vague, and on some
views, ‘Hitler is a villain’ lacks a truth-value if Hitler is a borderline-villain. To bypass this objection, we need
merely add a further premise to the argument, namely the obvious truth that Hitler is not a borderlinevillain.
240
Judith Jarvis Thomson
There are two forms the objection can take: a strong and a weak. In its strong
form, it says that there are no properties at all.⁷ (Or, perhaps, that there is no
sense to be made of the idea that there are such things.) In particular, then,
there is no such property as being a villain. A fortiori, the truth of
(P) ‘Is a villain’ is a logical predicate,
does not justify
(C1) There is such a property as being a villain,
and so does not justify (C2).
We can bypass that idea, however, because no Expressivist would offer it as
an objection to the argument. Expressivists believe that there is a radical difference between the non-normative and the normative: they believe that there are
natural properties, such as being a banana and being red, and that what there
aren’t, are normative properties such as being a villain.
So perhaps an Expressivist would instead make the objection in its weak
form: the fact that ‘is P’ is a logical predicate justifies the conclusion that ‘is P’
stands for a property if and only if ‘is P’ is a non-normative logical predicate. ‘Is
a banana’ and ‘is red’ are non-normative logical predicates, so they do stand for
properties; not so ‘is a villain.’ Once again, therefore, the truth of (P) does not
justify (C1), and so does not justify (C2).
Why this restriction to non-normative logical predicates?⁸ Here the
Expressivist comes into his own: because the use (or function?) of a normative
predicate is to express a favorable or unfavorable attitude. (This way of putting
the point is common.) Or, more precisely, . . . Well, more precisely what?
A number of different precisifications are on offer. Let ‘is F’ be a normative
predicate. Precisification (i) says that if a person says ‘A is F,’ then he expresses—
or anyway is rightly taken to express—a favorable or unfavorable attitude toward
A. (ii) says that someone who says ‘A is F’ does not describe A; he evaluates A.
(iii) appeals to belief and descends from Hume. It says, first, that one who has
the (putative) belief that A is F is thereby motivated to do something, whereas
one who has the belief that A is a banana is not thereby motivated to do
anything. And it says, second, that there are no beliefs such that the person who
⁷ There is good reason to reject the objection in its strong form, lying in the plausibility of what might be
called minimalism about properties. It obviously won’t do to say: if ‘is P’ is a grammatical predicate, then ‘is P’
stands for a property. (Compare my first, rough, expression of minimalism about truth.) So let us take constrained minimalism about properties to say: if ‘is P’ is a logical predicate, then ‘is P’ stands for a property.
What more could it come to for ‘is P’ to stand for a property than that it is a logical predicate? At a minimum,
we can say that a friend of the objection in its strong form has the burden of proof that more is needed.
⁸ The consideration mentioned in n. 7 reinforces the idea that the Expressivist who makes this claim has
the burden of proof.
The Legacy of Principia
241
has them is thereby motivated to do something. It therefore concludes that
there isn’t really any such thing as the belief that A is F.
An Expressivist who opts for one of those three ideas must do two things: he
must show that his claim is true and then go on to show why his claim should
be thought to justify the conclusion that there is no such property as being
F. For brevity, I will bypass (i) and (ii); (iii), anyway, is the one I take to be most
attractive to most contemporary Expressivists.⁹
It is easy enough to see how a friend of (iii) can get from (iii) to the conclusion that there is no such property as being F. If there were such a property as
being F, then nothing would stand in the way of a person’s believing that a thing
has it. So if there is no such belief, then it follows that there is no such property.
What is not so easy to see is why we should accept (iii) itself. According to
the first step taken by friends of (iii), one who has a normative belief (if there
are such things) is thereby motivated to do something. This is so often said as
for it to seem almost rude to ask: motivate the believer to do what? In some
cases, the answer all but suggests itself. If I believe that I ought to mow my lawn,
then what I am motivated to do is to mow my lawn. If I believe that I ought to
pay my taxes, then what I am motivated to do is to pay my taxes.¹⁰ That is
because it is perfectly clear in those cases what acting in accord with the belief
would consist in. Acting in accord with my belief that I ought to do such and
such would consist in doing the such and such; and that is why friends of (iii)
would say that what my belief motivates me to do is the such and such.
But what of my other normative beliefs? Such as, for example, my belief that
Hitler is a villain? Or worse yet for these purposes, my belief that you ought to
mow your lawn? Here it is not at all clear what I am to be thought to be
motivated to do by virtue of having the belief. That is because it is not clear
what, if anything, acting in accord with these beliefs would consist in.
⁹ I return briefly to (i) in n. 11. (ii) appears in Moore’s post-Principia writings, though Moore was
markedly more cautious about it than those who later fixed on it have been. In Moore (1942: 591), he said
that the idea was not clear, and that ‘[t]o make it clear it would be necessary to specify the sense of “describe”
in question,’ and he said he was no more able to do this in 1942 than he had been in 1922, when he first considered the idea. There is a difficulty here, after all, for why should evaluating be thought incompatible with
describing? Indeed, why isn’t describing as good exactly what evaluating as good consists in?
In an earlier work—Harman and Thomson 1996—I had suggested that accepting (iii) is the mark of the
modern Expressivist. The most modern Expressivist, however, evidently prefers something weaker than (iii),
namely (iv): there really is such a thing as the belief that A is F, but it is of the quite special sort that Hume
pointed to, that is, a belief such that a person who has it is thereby motivated to do something. See n. 12.
I have revised the discussion of (iii) that I presented at the conference in order to meet criticism by Allan
Gibbard.
¹⁰ If I have understood him, Allan Gibbard opts for a markedly stronger thesis about first person ‘oughtjudgments’, namely that coming to believe that I ought to do such and such is not merely coming to be
motivated to do so, but is actually deciding to do so. See Gibbard 2002. (I think that Simon Blackburn is
hinting at the same idea in Blackburn 1998: 70.) For criticism, see Hawthorne 2002.
242
Judith Jarvis Thomson
It may well be supposed that if I believe that Hitler is a villain, then if my
town council were to declare a referendum on whether to put up a statue to
him, I would be motivated to vote against doing so. Again, if I believe you
ought to mow your lawn, then if I came to believe that I was situated as you are,
I would be motivated to mow my own lawn. But as things stand now, I do have
those beliefs about Hitler and you, but I am not now motivated to do anything
at all by them.
Let us borrow a figure from Quine. We can think of a person’s normative
beliefs at a given time as forming a net—perhaps that the nodes are his normative
beliefs and that the strands linking them are either principles of logic that
he accepts (for one normative belief may entail another) or statements of fact
that he believes true. At the center are his highly abstract and general normative beliefs, such as (perhaps) the Categorical Imperative. Further out are his
less abstract, less general normative beliefs, such as that Hitler is a villain and
that Jones ought to mow Jones’s lawn. Around the edges are his beliefs to the
effect that he ought to do this and that he ought to do that.
It is only those beliefs around the edges in respect of which it is perfectly clear
what acting in accord with them would consist in. Therefore, it is at most those
beliefs around the edges about which it can plausibly be said that the believer is
now motivated to do something by virtue of having them—the something
being: act in accord with them.
And should we agree that the believer is now so motivated? I believe that I
ought to mow my lawn, yet here I sit, feckless as usual in the face of enterprises
that require effort. Or I am feeling glum and couldn’t care less about returning
the library book I know I ought to return now.
The friend of (iii) can certainly ask the following. Suppose we think we have
a picture of Alfred’s entire normative net. If he is not motivated to do any of the
things that, according to our picture, he believes he ought to do, then what
reason do we have for thinking that our picture is correct? Alfred’s saying that
it is does not count for much. Indeed, his saying that it is counts for nothing
at all if his behavior shows that he is motivated by the normative beliefs around
the edges of a quite different net.
There is something right in the offing here, though it would take some
delicacy to get the point stated properly; after all, we can hardly want to say that
a person’s normative beliefs do not survive through his occasional states of deep
depression in which nothing matters to him at all. Moreover, there would
remain the hard question why it is so, hard because what makes it so is not at all
obvious. In particular, explaining what makes it so by appeal to the idea that
the nodes of a person’s normative net are not really beliefs at all—and therefore
that there are no normative properties—is taking a step that, on the face of it,
The Legacy of Principia
243
seems utterly unwarranted. For the data remain: the grammatical predicate ‘is
a villain’ and its ilk are logical predicates.
In short, the friend of (iii) can certainly have that the question what a person
is motivated to do has an important bearing on what we can plausibly take his
normative beliefs to be. But the conclusion the friend of (iii) wishes to reach
just does not follow from as weak a claim as that is.¹¹
VI
So the idea about belief that we have been looking at does not justify the view
that whereas ‘is P’ stands for a property if it is a non-normative logical predicate, such as ‘is a banana,’ ‘is P’ does not stand for a property if it is a normative
logical predicate, such as ‘is a villain.’ And it therefore does not show that an
argument from
(P) ‘Is a villain’ is a logical predicate,
to
(C1) There is such a property as being a villain,
and therefore to
(C2) If Alfred says ‘Hitler is a villain,’ he thereby attributes a property to
Hitler, namely the property ‘being a villain’
fails to reach its conclusion. But if it does reach its conclusion, then
Expressivism is false.
However, there is something else that might be thought to show that that
argument fails to reach its conclusion, a something that we glanced at only very
briefly in section I. And it may be this that is really or ultimately at work in
many people who are tempted by Expressivism.
Suppose you accept Moore’s open question argument in its most general
form. This is, you believe that normative properties generally, if there are
any, are non-natural properties. If you grant that (P) entails (C1), then you are
¹¹ An Expressivist who opts for precisification (i) may be thought to have an easier time of it with my
belief that you ought to mow your lawn than does the Expressivist who opts for (iii). No doubt I need not be
motivated by that belief, but don’t I anyway have a favorable attitude toward your mowing your lawn? I lack
space to expand on this matter, but merely invite the reader to take the following question seriously: what
favorable attitude? (Blackburn darkly refers to it as ‘the attitude that one ought to’ in Blackburn 1998: 82.) I
make some comments on the ideas in the offing here in Harman and Thomson 1996. The fact that it is hard
to see how the Expressivist is to answer the question ‘What attitude?’ may be what explains why Expressivist
writings seem to slither back and forth between (i) and (iii).
244
Judith Jarvis Thomson
committed to the existence of at least one normative, and thus non-natural,
property, namely being a villain. But surely it is at best a suspect idea that there
are any non-natural properties!
Why is that idea suspect? J. L. Mackie said (in Mackie 1977) that normative
properties, if there were such things, would be queer; but that is largely because
he was relying on the idea that they would be (objectionably) linked to
motivation in the way discussed in the preceding section.
We will look at the sources of this suspicion in the following section. Before
doing so, however, I should concede the unfortunate fact that I do not own the
term ‘Expressivism’. (If I am not mistaken, it was first introduced into philosophy
by Allan Gibbard.) I defined it as I did because I believe that twentieth-century
Expressivism began as a response to Moore’s Principia and, indeed, as that quite
distinctive response to Principia. But it may be that some of our contemporaries who are called, and who would call themselves or allow themselves to be
called, Expressivists would not say what I said Expressivists say—it may even be
that they would accept (C2) in particular. But then what exactly do or would
they say in response to Principia? Each case would have to be examined
individually, and I will not try to survey them.¹²
VII
Let us go back to Principia. Moore invited his readers to accept:
(Moore’s Premise)
There is such a property as goodness.
The open question argument yields:
(OQA Conclusion) There is no natural property NP such that NP is
identical with the property goodness.
¹² I stop for a moment over Simon Blackburn, however, since he is such a puzzling case. He accepts the
description of himself as an Expressivist, but his recent Blackburn 1998 leaves it quite unclear what his
Expressivism consists in.
On the one hand, he says that ‘the expressivist thinks we can say interestingly what is involved for a subject to think that X is good. It is for S to value it, . . .’ (p. 50) and that valuing something ‘is not to be understood as describing it . . .’ (p. 49). Moreover, he says that what we do when we say that something is good
or right is this: ‘We avow a practical state. “Avowal” here means that we express this state, make it public, or
communicate it’ (p. 68). He does not mean that we report that we are in the state.
On the other hand, he tells us later that normative sentences have truth-values, and where ‘S’ is a normative
sentence, that we may come to believe that S. As I said in n. 9, the most modern Expressivist evidently prefers
(iv) to (iii); I had Blackburn in mind. Indeed, we may find out that S and therefore that there is such a fact as
the fact that S (for things do have ‘ethical properties’), which is represented by ‘S,’ and we may therefore come
to know that S.
Is there a consistent view held in both of these two hands jointly?
The Legacy of Principia
245
Moore therefore invited his readers to accept:
(Moore’s Conclusion) The property goodness is a non-natural property.
Alas, the idea that there are non-natural properties looked at best suspect to
many of his readers. Why so? You would find that idea suspect only if you
thought that the natural properties are in a certain way privileged. Which are
the natural properties? And what about them privileges them?
It is of interest that Moore himself thought that readers of chapter I, in
which he lays out the open question argument, would know perfectly well
which the natural properties are, partly because the term was a familiar one,
partly from the examples he gave. That he thought this shows itself in the fact
that he did not trouble to try to say what a natural property is until chapter II.
What he said there is famously unsatisfactory. What was he groping for? In the
preface he began to write for a second edition of Principia, but never finished, he
said that the following indicates the properties he had had in mind: a natural
property is a ‘property with which it is the business of the natural sciences or
Psychology to deal, or which can be completely defined in terms of such’ (Moore
1993: 13). And this seems to have been at least roughly the way in which he was
understood by the readers of Principia that I have been concerned with.
What privileges those properties? Many readers thought: the fact that they
alone are the properties whose possession or non-possession one can find out
about by the methods of the sciences, namely sense-perception and deduction
from sense-perception.¹³ Some of them took that fact about the natural properties to entail that they are the only properties: Ayer, for example, who argued
as follows: If P is a property that is not a natural property, then attributing P to
a thing is asserting a proposition that is not empirically verifiable. But there are
no non-empirically verifiable propositions. Therefore there are no non-natural
properties. (See Ayer 1946.) That too was famously unsatisfactory, but even
those of Moore’s readers who rejected Ayer’s argument thought it mysterymongering to say of a putative property that it really is one, though nobody can
find out by empirical means whether a thing possesses it.
So they thought that Moore’s conclusion had to go, and given the attractiveness of the open question argument, they thought that the best way to get
rid of it was to get rid of Moore’s Premise.
And isn’t there a further, independent, ground for thinking that we should
get rid of Moore’s Premise? My school’s team beats your school’s team at
¹³ Many readers, but very likely not Moore himself, for he thought that although goodness is a nonnatural property, it supervenes on natural properties. So it is arguable that what Moore felt really mattered
about the natural properties—even at the time of writing Principia—was not an epistemological fact about
them but instead that one who attributes them to a thing describes it. (See n. 9.)
246
Judith Jarvis Thomson
basketball. ‘That’s good,’ I say. What property could it be thought that I
attribute to my team’s winning? That it is good for my school? But that is
certainly not what people generally attribute to a thing whenever they say
‘That’s good’ of it. Sometimes they are talking about a fountain pen or a watch
and are not saying about it that it is good for my school. If we are to think that
whenever a person says ‘That’s good’ of a thing he attributes a property,
goodness, to the thing, then this seems to be a property without any particular
content. So there is something additionally objectionable about Moore’s
Conclusion, beyond that it commits us to the existence of an epistemologically
unsatisfactory property: it commits us to the existence of an epistemologically
unsatisfactory property that is also empty.
Two comments are called for. First, this independent ground for thinking
that we should get rid of Moore’s Premise is entirely right, for there is no
property that people attribute to a thing whenever they say ‘That’s good’ of it.
The grammatical predicate ‘is good’ is not a logical predicate; the sentence ‘A is
good’ is semantically incomplete. A person who says ‘A is good’ attributes
something to A only if what he (though not the sentence) means is that A is a
good fountain pen or a good dancer or good for use in making cheesecake or
good for Jones or England or the tree in his backyard. Just as a person who says
‘A is big’ attributes something to A only if what he (though not his words)
means is that A is a big flea or a big strawberry or a big banana.¹⁴
This fact emerges in another way. The sentence ‘A is a red apple’ entails ‘A is
red and A is an apple,’ for ‘A is red’ (like ‘A is an apple’) is semantically complete.
By contrast, ‘A is a good dancer’ does not entail ‘A is good and A is a dancer’—
else ‘A is a good dancer and a bad tennis player’ would entail ‘A is good and A is
bad and A is a dancer and A is a tennis player.’ ‘A is a good dancer’ no more
entails ‘A is good and A is a dancer’ than ‘A is a big flea’ entails ‘A is big and A is
a flea.’ For further discussion, see the Addendum.
One might have thought this an obvious fact, once it had been drawn attention to. But surprisingly enough, it has had almost no impact on normative
theorists, some of whom still set themselves—as Moore did in Principia—to
answer the question which things possess the property goodness, without
noticing that if their question is taken literally, the only possible answer to it is
‘None.’
In any case, those of Moore’s readers who rejected Moore’s Premise were
entirely right to do so. The mistake that the Expressivists made lay in what they
¹⁴ To the best of my knowledge, it was Peter Geach who first drew attention to this fact about “is good,”
in Geach 1956. Others who have done so include Ziff 1960, von Wright 1963, Foot 1985, and Thomson
1997. It is because of this fact about “is good” that I have throughout bypassed the Cornell realists—I mentioned them only in footnote 1.
The Legacy of Principia
247
thought had to be concluded, from the falsity of Moore’s Premise, about what is
done by a person who says ‘A is good,’ namely that such a person merely displays
a favorable attitude toward A—thus that he does no more than he would have
done had he merely said ‘Hurrah!’ in response to A.¹⁵ That was certainly a mistake. A person who says ‘A is good’ does display a favorable attitude toward A.
But he does more than that. He also says that A is a good fountain pen or a good
dancer or good for use in making cheesecake or good for Jones or England or the
tree in his backyard. If the context in which he says ‘A is good’ does not tell his
hearers which of these things he means, then they simply do not know what he
means and have to ask. And what won’t do is for him to reply, ‘No, no, I didn’t
mean anything like that, what I meant was only that A is a good thing!’
I have been in the habit of summarizing this point about the predicate ‘is
good’ in the slogan: all goodness is goodness in a way. Nothing is just, simply,
good. Though many things are good fountain pens and good dancers and so on
and on.
VIII
Something similar holds of ‘is right’ and ‘is wrong.’ Nothing is just, simply,
right, though many things are right answers to such and such questions or
right keys to the front door or right medicines to take to cure a sore throat.
Moreover, ‘Sherman’ is the right answer to the question which general burned
Atlanta; it is a wrong answer to the question which general was most admired
by Lincoln. In short, all rightness is rightness in a way. (There are obvious
differences between ‘is right’ and ‘is good,’ but they do not matter for present
purposes.)
Similarly for ‘is correct’ and ‘is incorrect.’ ‘Sherman’ is the correct answer to
the question which general burned Atlanta; it is an incorrect answer to the
question which general was most admired by Lincoln.
What may seem to have been lost sight of is what has been happily called the
Imperialism of the Moral. An act might have been good for its agent but
morally bad; doesn’t that mean that it was on balance bad? An act might have
been prudentially right but morally wrong; doesn’t that mean that it was on
¹⁵ The idea that there is such a property as goodness has had other baleful effects on moral philosophy
than that of provoking the response made by Expressivists: I argued in Thomson 2001 that it has caused
Consequentialism to seem a more attractive idea than it is. It would be of great interest if those two effects of
the idea were connected otherwise than by their joint cause. I do not see how, but it is a striking fact that
many Expressivists opt for Consequentialism when they turn their attention from metaethics to normative
theory.
Judith Jarvis Thomson
248
balance wrong? If you believe in the Imperialism of the Moral, as I do, you
could opt for accounts of ‘is good’ and ‘is wrong’ that permit saying these
things. But they are not true to usage and not needed. We can make room for
the Imperialism of the Moral without them: we can say that where it shows
itself is where what is in question is what a person ought to do.
For where what is in question is what a person ought to do, what is in question
is what he just, simply, ought to do. Suppose it would be good for a person to
do a thing, but his doing the thing would be morally wrong. We do not conclude that he ought in one way to do the thing and ought in another way not
to, that ending the matter. When a person asks what he ought to do, what he is
asking is what, all things considered, and on balance, he (just, simply) ought to
do. The Imperialism of the Moral says: if it would be morally wrong for him to
do the thing, then he ought not do it, however good for him his doing it would
be. Friends of the Imperialism of the Moral are not committed to thinking that
this is true by virtue of the very meanings of the words (which is a wildly
implausible idea); they are committed only to its being true.
But I will have to leave ‘ought’ aside. There are hard, open questions about
its logic. And how moral considerations enter into fixing what a person ought
to do is too large an issue to enter into here. (I make some suggestions in
Thomson 2001: part 2.)
IX
We were looking at the epistemological considerations that led many of
Moore’s readers to think the natural properties privileged and, therefore, to
reject
(Moore’s Conclusion)
The property goodness is a non-natural property.
In light of the attractiveness of the open question argument, the readers I have
been most concerned with therefore rejected
(Moore’s Premise)
There is such a property as goodness.
Indeed, they had an independent ground for rejecting Moore’s Premise, since
the property goodness, if there were such a property, would have no content. I
said that two comments were called for. The first was that they were right to
reject Moore’s Premise on this independent ground. We should now turn to the
second comment.
Once we take seriously that all goodness is goodness in a way and all wrongness
wrongness in a way, the boundary between normative properties and natural
The Legacy of Principia
249
properties begins to break down. ‘Sherman’ is the right answer to the question
which general burned Atlanta; and its being so consists in the (on any view)
natural fact that it was Sherman who burned Atlanta. It may be that of two
keys, a large and a small, the large is the right key to the front door; and its being
so, if it is, consists in the (on any view) natural fact that it is the large key, not
the small, that opens the front door.
The breakdown also shows itself in the case of goodness. One of Moore’s
paradigm examples of a natural property was being pleasant. (That is because
he was particularly concerned to refute normative theories according to which
goodness just is pleasantness.) No doubt it is a plausible idea that being pleasant is a natural property. But being pleasant is also a way of being good. You say
I really should go see such and such a movie. Why so? You say, ‘It’s a pleasant
little comedy.’ In saying it is pleasant you are praising it. Not just giving a
ground for praise, but praising.
So also for being interesting, absorbing, thrilling, amusing, elegant, charming, witty. Bert’s witty speech at Alfred’s funeral might have been a bad funeral
oration; but there was anyway something good to be said for it, namely that it
was witty. Yet isn’t there as much reason to regard these as natural properties as
there is to regard being pleasant as a natural property? Some or all of these are
‘response-dependent’ properties, but that is compatible with their also being
natural properties. Many people believe—whether rightly or wrongly—that
redness is a response-dependent property; they are not thereby committed to
believing that redness is a non-natural property.
No doubt I may not feel like seeing a pleasant little comedy this evening. If so,
then your praise of the movie does not motivate me to go see it. And it may be false
that I should go see it. But those possibilities are no barrier to the movie’s being
good in a way. If a thing is a good fountain pen or a good watch or a good car then
it is on any view good in a way. That is compatible with its being the case that when
I learn that a thing is good in one of those ways I am not thereby motivated to do
anything in respect of the thing, and with its being false to think that I should.
X
In sum, I think it very plausible that the attractiveness of Moore’s open
question argument itself issued from his fixation on, and his invitation to his
readers to fix on, the pseudo-property goodness. Of course there is no natural
property identical with it! I suspect that if that had been seen for the thoroughly
unexciting fact that it is, Principia would have had a markedly less powerful
impact on twentieth-century moral philosophy than it did.
250
Judith Jarvis Thomson
Addendum
A1. Two further points are worth mention about the analogy Geach draws
between ‘is good’ and ‘is big.’ First, the former has a much richer range of
occurrences than the latter. Whenever a person says ‘A is big,’ then for some K,
what he means is that A is a big K. Not so whenever a person says ‘A is good.’ It
may be that for some K, what a speaker means is that A is a good K. But it may
be that what he means is instead that A is good for this or that or that A is good
for use in doing this or that and so on and on. ‘Good’ is the most general
evaluative term in English, and we evaluate things, not only as K’s, but in other
ways too.
A second contrast is more important for moral philosophy. It has been
suggested that while it may be that ‘A is good’ is semantically incomplete,
perhaps that is not true of ‘A is better than B.’ In any case, the fact that ‘A is
good’ is semantically incomplete does not entail that ‘A is better than B’ is too.
Geach’s ‘is big’ is not specially suited to making this point. For suppose that A
is tall and skinny, and B is short and fat. Which is bigger? Is that to be settled by
comparing their weights? I doubt it. But in any case, ‘is tall’ would have served
Geach’s purposes as well as ‘is big’ did, for whenever a person says ‘A is tall,’ then
for some K, what he means is that A is a tall K. Yet ‘A is taller than B’ is not
semantically incomplete. A person who says ‘A is taller than B’ does not
attribute something to A and B only if what he means is that A is a taller man
than B or A is a taller giraffe than B or A is a taller building than B. A person
who says ‘A is taller than B’ does thereby attribute something to A and B,
whatever A and B are and whatever he thinks they are—whether both men,
both giraffes, one a man and the other a giraffe, and so on.
And if ‘A is better than B’ is not semantically incomplete, then while the only
possible answer to the normative theorist’s question ‘Which things possess the
property goodness?’ is ‘None,’ no such objection can be brought against the
question ‘Which things stand in the relation betterness to which other things?’
(It might pay to make explicit the fact that it would not be possible to get from
the comparative back to the non-comparative. It won’t do at all to say, for example,
that A has the property goodness just in case A stands in the relation betterness
to all other things. Or to most other things. Or to many other things.)
But ‘A is better than B,’ unlike ‘A is taller than B,’ is semantically incomplete.
‘A is a taller man than B’ entails ‘A is taller than B and A and B are men.’ By contrast, ‘A is a better tennis player than B’ does not entail ‘A is better than B, and
A and B are both tennis players’—else ‘A is a better tennis player than B, and B
is a better chess player than A’ would entail ‘A is better than B, and B is better
than A, and A and B are tennis players and chess players.’
The Legacy of Principia
251
A2. Geach had contrasted ‘is red’ with ‘is big’: ‘A is red’ is semantically
complete; ‘A is big’ is not. In section VII above, I accepted that ‘is red’ differs
from ‘is big’ in this way. But in an earlier paper, I had expressed reservations
about Geach’s choice of predicate to contrast with ‘is big.’ Michael
Zimmerman quotes me as having said in a footnote:
‘Red’ was not in fact well chosen for Geach’s purposes, since ‘red’ is heavily context
dependent: what we ascribe to an apple when we say ‘It’s red’ is different from what we
ascribe to the paint in a certain can when we say ‘It’s red.’ Better choices would have
been ‘visible’ . . . or ‘poisonous’ or. . . . (Thomson 1997: 277 n. 5)
And he says:
Now, it seems quite right to say that what’s red as far as apples go may not be red as far
as paints go. But of course this sort of division can be continued. After all, what’s red as
far as Macintosh apples go may not be red as far as Red Delicious apples go. And, contrary to what Thomson seems to imply, what’s visible or poisonous to x may not be visible or poisonous to y. The fact is, very many properties are determinable (to some
extent) rather than (fully) determinate, including all those just mentioned.
(Zimmerman 2001: 22)
And he goes on to use this point to cast doubt on Geach’s conclusion about ‘is
good.’ Let us set aside the use Zimmerman makes of the point and merely fix
on the predicate ‘is red.’
Zimmerman misinterpreted me: what worried me about ‘is red’ is not what
he thinks it was. No doubt I did not make clear enough what it was in the footnote he quotes from.
There are three ideas in the offing here, and I think that it pays to be clear
that they differ. There is the fact that (1) redness is a determinable. That is what
Zimmerman thinks worried me about Geach’s choice of ‘is red.’ Not so. The
contrast Geach drew between ‘is red’ and ‘is big’ is not in the least affected by
the fact that ‘is red’ stands for a determinable, and I was certainly not objecting
to his contrasting ‘is red’ with ‘is big’ on the ground that ‘is red’ stands for a
determinable.
What worried me is what emerges on consideration of the following
contrast: (2) for an apple to be red it is enough that its surface be red, whereas
for the paint in a certain can to be red, it is required that the paint be red ‘all
through’. I concluded that there is no such thing as the property redness: there
are rather two of them, surface-redness and redness-all-through. And therefore
that if someone says ‘it’s red,’ we do not know what he means unless the context tells us which of the two he is attributing to the thing. And therefore that
‘is red’ is in an important respect similar to ‘is big.’ I therefore offered alternatives for which a contrast of kind (2) does not arise.
252
Judith Jarvis Thomson
But I should not have been so worried by that contrast. It is worth taking
note of because we are thereby reminded that a phenomenon familiar from
elsewhere turns up here too. Here is Aristotle’s example: healthy people and
healthy food. Is there such a thing as the property being healthy that healthy
people and healthy food share? Here is another: Alfred caused Bert’s death and
Alfred’s shooting of Bert caused Bert’s death. Is there such a thing as the relation
‘causes’ that Alfred and Alfred’s shooting of Bert both stand in to Bert’s death?
No, of course not. But the one property ‘healthy’ is reducible to the other: for
something to have the one is for eating it to conduce to the eater’s having the
other. Similarly, for Alfred to stand in the one to Bert’s death is for an act of
Alfred’s to stand in the other to Bert’s death. And so also, I should think, for
redness. Surface-redness is presumably fundamental—thus for a thing to
have redness-all-through is presumably for every portion of it to have surfaceredness. (In light of this reducibility of the one property to the other, it is
misleading to say that what we have in these cases is, simply, ambiguity.)
‘Is red’ remains different from ‘is big,’ however, and in just the way Geach
said it does. ‘A is a big flea’ does not entail ‘A is big and A is a flea.’ ‘A is a red
banana’ does entail ‘A is red and A is a banana’—despite the fact that if a person
says these things, we may not know which of the two properties, surfaceredness or redness-all-through, he is attributing to A.
Or so it certainly seems. I said that there are three ideas in the offing here,
and according to the third, that is a mistake: ‘is red’ is exactly like ‘is big’ in that
respect since (3) if Alfred has red hair, then it does not follow that his hair is red!
His hair is orange, not red.
Some years ago, Romane Clark recommended construing ‘is red’ as, like ‘is
big,’ a predicate modifier rather than as a logical predicate. (See Clark 1970, in
which he makes suggestions toward a general theory of predicate modifiers.)
He said that we should find this idea attractive on the ground that ‘[a] very red
chigger need not be very red at all’ (Clark 1970: 334). Not knowing what chiggers look like, I focus instead on red hair.
Concluding that ‘is red’ is a predicate modifier on such a ground is bringing
a big piece of machinery to bear on what must surely be a small local problem.
‘Red hair’ is not unique in this respect: white people aren’t white unless they
have been dead for quite a while. But there aren’t many such cases. On the other
hand, such expressions are of interest, and it might pay to stop for a moment to
ask how they should be understood.
Perhaps ‘red hair’ and ‘white people’ are metaphors?¹⁶ More precisely, perhaps:
unless the context indicates otherwise, a hearer takes them to be metaphors.
¹⁶ I thank Catherine Elgin for this idea and for “white people.”
The Legacy of Principia
253
For after all, the context might indicate otherwise. If we know that Bert has been
wallowing in red paint, we take the child who says ‘Look, Bert now has red hair!’
literally. This is an attractive idea. Opting for it would explain why we do not normally expect a person’s hair to be red when we have been told that he has red hair.
It does not seem entirely satisfactory, however. ‘Purple prose’ is on any view
a metaphor, and we know that for a bit of prose to be purple prose is for it to be
wildly exaggerated. Similarly for ‘black moods’: we know that for a mood to be
a black mood is for it to be a state of gloom. We do not supply a color other than
purple or black. By contrast, we know that for a person’s hair to be red is (normally) for it to be orange, and for a person to be a white person is (normally)
for him to be pinko-gray (as Forster put it).
So perhaps ‘red hair’ and ‘white people’ are merely stable exaggerations. ‘He
turned green when he learned that what he just ate was mashed spider.’ Well,
he didn’t, not really. He only went ever so slightly greenish, and that is all you
thought had happened when I said what I said. What makes an exaggeration
stable is a good question—perhaps of particular interest in the case of ‘white
people’—but that, I think, may be all that is at work here.
A3. Zimmerman says that in any case, all of the preceding is irrelevant to
Principia, for what Moore had in mind there was not goodness, but intrinsic
goodness. And, Zimmerman says, the fact (supposing it a fact) that there is no
such property as goodness is entirely compatible with there being such a
property as intrinsic goodness.
Very well, what is intrinsic goodness? We might have taken the term to be
intended to refer to non-derivative goodness: the goodness a thing has, but not
by virtue of its standing in a suitable relation to (as, for example, its being
conducive to) something else that is good. So understood, however, there is no
such property if there is no such property as goodness.
So what is it then? Zimmerman says that intrinsic goodness is ‘ethical goodness,’ where—if I have understood him—for x, y, z, and so on, to possess
ethical goodness is for it to be the case that ‘there is a moral requirement to favor
them (welcome them, admire them, take satisfaction in them, and so on) for
their own sakes’ (Zimmerman 2001: 24). I mention only one of the difficulties
that arise here. If we are morally required to favor x, y, and z, can that be a sheer,
self-evident datum? Doesn’t it have to be due to their possessing some features
in virtue of which we are? And what feature is that? Not goodness, I hope.
In any case, while this idea, or anyway something like it, does turn up in
Moore’s later writings, I do not find it in Principia, which says that ‘ought’ is
analyzable in terms of ‘good,’ and not the other way round.¹⁷
¹⁷ I thank Tyler Doggett for comments on an earlier draft.
254
Judith Jarvis Thomson
References
Ayer, A. J. (1946). Language, Truth, and Logic. 2nd edn. London: Gollancz.
Blackburn, S. (1984). Spreading the Word. Oxford: Oxford University Press.
—— (1998). Ruling Passions. Oxford: Oxford University Press.
Clark, R. (1970). ‘Concerning the Logic of Predicate Modifiers’. Nous, 4: 311–335.
Foot, P. (1985). ‘Utilitarianism and the Virtues’. Mind 94, reprinted in S. Scheffler
(ed.), Consequentialism and its Critics. Oxford: Oxford University Press.
Geach. P. T. (1956). ‘Good and Evil’. Analysis, 17, reprinted in Philippa Foot (ed.),
Theories of Ethics. Oxford: Oxford University Press, 1967.
—— (1972). Logic Matters. Berkeley and Los Angeles: University of California Press.
Gibbard, A. (1990). Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard University
Press.
—— (2002). ‘Normative and Recognitional Concepts’. Philosophy and
Phenomenological Research, 64: 151–167.
Glanzberg, M. (2003). ‘Minimalism and Paradoxes’. Synthese, 135: 13–36.
Harman, G., and Thomson, J. J. (1996). Moral Realism and Moral Objectivity.
Oxford: Blackwell.
Hawthorne, J. (2002). ‘Practical Realism?’ Philosophy and Phenomenological Research,
64: 169–178.
Horwich, P. (1993). ‘Gibbard’s Theory of Norms’. Philosophy and Public Affairs, 22:
67–78
Jackson, F., Oppy, G., and Smith, M. (1994). ‘Minimalism and Truth Aptness’.
Mind, 103: 287–302.
Mackie, J. L. (1977). Ethics: Inventing Right and Wrong. New York: Penguin.
Moore, G. E. (1942). ‘A Reply to my Critics’. In P. A. Schilpp (ed.), The Philosophy of
G. E. Moore. Evanston, Ill.: Northwestern University Press.
—— (1993). Principia Ethica, ed. Thomas Baldwin. Cambridge: Cambridge
University Press.
Prichard, H. A. (1912). ‘Does Moral Philosophy Rest on a Mistake?’ Mind, 21: 21–37.
Sayre-McCord, G., ed. (1988). Essays on Moral Realism. Ithaca, NY: Cornell
University Press.
Stevenson, C. L. (1944). Ethics and Language. New Haven: Yale University Press.
Thomson, J. J. (1997). ‘The Right and the Good’. Journal of Philosophy, 94: 273–298.
—— (2001). Goodness and Advice. Princeton: Princeton University Press.
Timmons, M. (1999). Morality without Foundations. Oxford: Oxford University Press.
von Wright, G. H. (1963). The Varieties of Goodness. London: Routledge & Kegan Paul.
Ziff, P. (1960). Semantic Analysis. Ithaca, NY: Cornell University Press.
Zimmerman, M. J. (2001). The Nature of Intrinsic Value. Oxford: Rowman & Littlefield.
12
Cognitivist Expressivism
Terry Horgan and Mark Timmons
Expressivism is a position in metaethics that is a descendant of noncognitivism—a
view that was perhaps the dominant metaethical theory for about forty years,
between 1935 and 1975.¹ The basic insight of the noncognitivists was that
language can play a dynamic as well as a descriptive role in interpersonal
interaction, and that moral discourse is a prime example of the dynamic use of
language. According to one dominant strain of noncognitivism, emotivism,
championed by A. J. Ayer and C. L. Stevenson, moral judgments function
primarily to express one’s feelings about some object of evaluation. In an
interpersonal context, such expressions of feeling typically serve the dynamic
function of influencing the attitudes of others. The other main variety of
noncognitivism, prescriptivism, developed by R. M. Hare, took dynamic,
imperatival utterances as a model for moral discourse, and thus emphasized the
directive, action-guiding element of such discourse. Noncognitivism did come
in for its share of criticism, depending on the version under scrutiny, but
the bottom line seemed to be that this kind of view appears incompatible with
what Allan Gibbard calls the ‘objective pretensions’ of moral thought and
discourse, including the idea that moral judgments seem to be beliefs with
assertible, truth-apt content. Emotivists and prescriptivists, because they took
moral language to express noncognitive attitudes, were forced to explain away
such pretensions. So, for instance, Carnap (1935: 25) held that ‘a value
statement is nothing else than a command in misleading grammatical form’.
But noncognitivist attempts to explain away various deeply embedded features
For their help on this paper, we wish to thank Noell Birondo, Paul Bloomfield, David Chalmers, David
Copp, Michael Gill, Uriah Kriegel, Mark Lance, John Tienson, and audiences at the University of
Cincinnati, Harvard University, and Wake Forest University.
¹ Ayer’s Language, Truth and Logic was published in 1936 and contained the first widely influential
presentation of a noncognitivist metaethical theory. As Mary Warnock noted in her Ethics since 1900, ‘It is
part of the measure of the importance of [Ayer’s] book that no sooner was it published than it seemed that
emotivists were everywhere. They had not been converted by the book; it was their creed already’ (1960: 84).
256
Terry Horgan and Mark Timmons
of moral thought and discourse have seemed implausible and indeed unnecessary
to many moral philosophers.
In the late 1970s and on through the 1990s, some moral philosophers were
tempted by moral realism, thinking that with the help of various developments
in philosophy of mind and philosophy of language from the 1960s and early
1970s, one could countenance moral properties and facts and yet remain faithful to philosophical naturalism—the dominant metaphysic of the times. Also,
metaethical history lately has been returning to its Moorean roots with some
moral philosophers boldly defending versions of nonnaturalism.² But just as
the past 100 years of metaethics has seen realism go and come back again, those
working in the tradition of Stevenson and Hare (ourselves included) have
devoted time and effort into reviving the spirit, if not the letter, of older
noncognitivist positions. Recent work along these lines includes Simon
Blackburn’s quasi-realism, Gibbard’s norm-expressivist view, and our own
position, here labeled cognitivist expressivism. In some of our previous writings,
we have either individually or collaboratively tried to make progress in articulating and defending our view. This paper is another installment in a series of
works devoted to this project.
1. Preview of coming attractions
In the days of noncognitivism, the idea that moral judgments are not primarily descriptive of moral properties and facts (nondescriptivism), and the idea
that moral judgments do not express beliefs (noncognitivism) were taken to be
mutually entailing. Nondescriptivism and noncognitivism were a package
deal. And the views are mutually entailing if one embraces the following claim,
which we call the semantic assumption:
SA All cognitive content (i.e. belief-eligible, assertible, truth-apt
content) is descriptive content. Thus, all genuine beliefs and all
genuine assertions purport to represent or describe the world.³
The semantic assumption has been widely taken for granted in metaethics; it
has framed much of the philosophical debate, and has constrained the range of
² We have in mind Shafer-Landau (2003) and Audi (2004).
³ For present purposes we use ‘represent’ and ‘describe’ (and their variants) interchangeably. (This use of
‘describe’ is at work in metaethics when the label ‘descriptivism’ is deployed; it is wider in scope that the use
in which ‘descriptive’ and ‘normative’ operate as contrast-terms.) Beliefs and assertions that ‘purport to represent or describe the world’ include ones with theological content, and can also include ones about abstract
entities like numbers; so ‘world’ does not just mean ‘physical world’ or ‘spatiotemporal world’ or ‘world of
concrete particulars’.
Cognitivist Expressivism
257
options on the standard menu of competing positions. But this deeply
entrenched piece of orthodoxy, we maintain, is false. Its falsity would mean
that some metaethical space opens up (potentially anyway) for the combination
of nondescriptivism and cognitivism.
We defend just such a view, claiming that although moral judgments are
genuine beliefs, their overall content is not descriptive content. Moral judgments count as beliefs, despite being nondescriptive, because they possess
enough of the key, generic, phenomenological, and functional features of belief
(as well as satisfying the relevant platitudes governing the concept of belief ) to
qualify as genuine beliefs. We defend the claim that moral judgments are
genuine beliefs in section 6 below.
In our previous writings, we have used different labels for our position,
sometimes calling it ‘assertoric nondescriptivism’, sometimes calling it ‘nondescriptivist cognitivism’. Gibbard (1990: 7–8) uses the term ‘expressivism’ to
refer to metaethical views that take the primary role of moral judgments to be
expressive of attitudes that do not purport to represent or describe some moral
reality. So expressivism is committed to nondescriptivism about moral judgments and utterances. Expressivism thus subsumes old-time versions of
noncognitivism. But it leaves open the possibility of a cognitivist construal of
moral thought and discourse. Because the term ‘expressivism’ has taken hold,
we are here calling our view cognitivist expressivism.
Cognitivist expressivism is very similar in spirit to Blackburn’s more recent
presentation of his quasi-realism—it is a metaethical project that embraces an
austere irrealist moral metaphysics and yet attempts, in its semantic construal
of moral terms and the concepts, to account for the deeply embedded features
of moral thought and discourse. The main differences between our view and
Blackburn’s have to do with philosophical execution.⁴
In this paper we will not spend time explaining why we reject versions of
moral realism, moral constructivism, moral relativism, the error theory, and
noncognitivism. We have done that elsewhere.⁵ Rather, we plan to articulate
and defend our evolving metaethical view in a way that repackages key ideas
from our prior writings while also going beyond our previous work in two
important respects. First, we will dwell on matters of moral phenomenology—
the ‘what-it’s-like-ness’ of experiences involving moral judgment; we will
⁴ Blackburn (1984: 167–71) refers to any philosophical view that regards the judgments of some
discourse as ‘expressive’ as opposed to descriptive, an ‘expressive theory’ about that discourse. So his quasirealism is a form of expressivism. And in more recent writings (1996) he explicitly distances his ethical
expressivism from forms of noncognitivism. See also Blackburn (1998).
⁵ For a very brief overview of the problems for such theories with references to our other work, see Horgan
and Timmons (2005).
258
Terry Horgan and Mark Timmons
argue on one hand that this phenomenology supports the cognitivist
contention that moral judgments are genuine beliefs, and on the other hand
that such cognitive phenomenology also comports with the denial that the
overall content of moral judgments is descriptive. Second, we will offer a more
detailed account than we have provided before of how to accommodate certain
crucial generic features of the psychological role of belief-states (and corresponding features of the interpersonal role of moral utterances)—features
involving the embeddability of moral content within logically complex beliefs
and sentences, and inferences employing beliefs and sentences with embedded
moral content.
We begin with an insight to be found in G. E. Moore’s metaethical reflections
and, with Moore’s guidance, we then proceed to develop and partially defend
our view.
2. Moore on methodology
In Principia Ethica, Moore famously defended the idea that goodness is a ‘simple, indefinable, unanalysable object of thought’ (1903: 72). He also thought
that ‘real’ definitions of terms—definitions that reveal the essential nature of
their referent—are possible only when the term to be defined refers to something
complex. Since the property of goodness is simple, having no parts, Moore
claimed that ‘good’ cannot be defined.
The most important sense of ‘definition’ is that in which a definition states what are
the parts which invariably compose a certain whole; and in this sense ‘good’ has no
definition because it is simple and has no parts. It is one of those innumerable objects
of thought which are themselves incapable of definition, because they are the ultimate
terms by reference to which whatever is capable of definition must be defined.
(1903: 61)
Often, the term ‘irreducible’ is used in connection with the idea that in some
important sense it is not possible to analyze or define fundamental moral
concepts and the terms that express them. Put in these terms, Moore thought
that because the most fundamental concept in ethics, namely, goodness, refers
to something simple, the concept (and the term expressing the concept) is
irreducible.⁶
⁶ It is worth noting that Moore himself thought that talk of what is intrinsically good is equivalent in
meaning to talk about what ought to exist (1903: 33–4, 68, 166). So, roughly, the kind of reduction Moore
was anxious to rebut would involve defining basic moral terms and concepts by using nonmoral terms and
concepts.
Cognitivist Expressivism
259
We do not accept Moore’s moral realism, and we do not accept his view that a
moral term like ‘good’ refers to a property. But we do adhere to what we call
‘Moore’s methodological maxim’: moral terms and concepts are irreducible.
However, to say this does not mean that that nothing philosophically illuminating
can be said about them (in addition to their being irreducible). R. M. Hare, for
instance, had quite a lot to say about the semantics of moral terms and concepts,
which, if true, is very illuminating. However, in making use of imperatives to
understand moral language, Hare insisted that ‘it is no part of my purpose to
“reduce” moral language to imperatives’ (Hare 1952: 2).⁷ Rather than offer a
reductive analysis of moral terms and concepts which would, in effect, express
those terms and concepts in some sort of nonmoral idiom, Hare proposed to
understand terms like ‘ought’ by exploring the ‘logical behavior’ of these words in
ordinary language via similarities between such moral language and imperatives.
Of course, by taking imperative sentences as a model for the ‘logical behavior’
of moral terms and concepts, Hare did deny in effect that moral judgments are
genuine beliefs and that moral utterances are genuine assertions. On his imperatival model, the declarative grammatical form of moral sentences is misleading,
since declarative sentences normally are employed to make assertions and to
express beliefs. So his metaethical position can be called weakly reductive in its use
of nonassertoric, nondeclarative, sentences as a model of moral thought and language, even though it eschews the strongly reductive claim that moral utterances
are synonymous with, or semantically interchangeable with, imperatives.
Like Hare, we propose to explore the meaning of moral terms like ‘ought’ by
considering how such terms function in thought and language. But we plan to
give more weight than did Hare to moral thought (as opposed to moral
language), and we will give specific emphasis to matters of phenomenology.
Also, again like Hare, we will argue that sentences with certain distinctive
grammatical features are a useful model for understanding moral thought and
discourse (namely, sentences in a specific sort of formal language, described in
section 6). But our linguistic model will not be ‘reductive’ in even weak senses,
because it fully accommodates the cognitivist claim that moral judgments are
beliefs and moral utterances are assertions. In eschewing any sort of ‘reductive’
account of moral terms and concepts, we will be respecting Moore’s (antireductive) methodological maxim.
In articulating and defending our view, we set for ourselves three main tasks
that we will proceed to take up in order in the following sections.
Task I:
Describe some key generic features of beliefs, and argue that
moral judgments exhibit these features.
⁷ See also Hare 1952: 180–1.
260
Task II:
Task III:
Terry Horgan and Mark Timmons
Explore some key distinctive characteristics of moral judgments
in particular, as contrasted with ordinary nonmoral beliefs.
Set forth a theoretical account of belief that simultaneously
(i) treats moral judgments as genuine beliefs, (ii) treats moral
judgments as not descriptive in their overall content, (iii) accommodates the key distinctive characteristics of moral judgments in
a way that renders these features consistent with the claim that
moral judgments are genuine beliefs, and (iv) accommodates the
key generic features of belief in a way that is consistent with the
denial that the overall content of moral beliefs is descriptive.
In the course of pursuing the first two tasks, certain specific challenges will
emerge that will need to be faced in addressing the third task: some of the distinctive features of moral judgments threaten the idea that these judgments are
genuine beliefs, whereas some of the generic features of belief exhibited by
moral judgments threaten the idea that these judgments are nondescriptive in
their overall content. These challenges will be noted as the first two tasks are
pursued, and will be addressed in the course of dealing with the third task.
By completing these tasks we will provide a presumptive case in favor of cognitivist expressivism, but there remain further issues and challenges that our
view must meet. Late in the paper, we consider some of the most pressing of
these, and we briefly explain how our view attempts to meet them.
3. Terminological preliminaries
Before proceeding, it will be useful to make some explicit remarks about matters of terminology. We begin with some observations about our use of ‘moral
judgment’. First, we use this term in a metaethically neutral way to refer to those
psychological states whose contents are expressible by a moral sentence. Thus,
calling this sort of state a judgment leaves open whether it is a belief, a desire,
an intention, or some other psychological state. Below we argue that moral
judgments are most plausibly understood as beliefs. Second, like Mandelbaum
(1955: 46), our use of the term is intentionally broad in another way: what we
are calling moral judgments need not be psychologically inferential; they
might be psychologically spontaneous as when one just ‘sees’ that some action
is obligatory. Third, ‘judgment’ allows for process/product ambiguity in its
usage, i.e. between an episode of judging and being in the psychological state
resulting from a judging episode. Context should make clear how we are using
the term.
Cognitivist Expressivism
261
We have been using, and will continue to use, the currently widespread
term ‘content’ in connection with moral judgments and utterances. We discuss
content of various kinds: the overall content of a judgment or utterance (the
content of the ‘that’-clause employed to describe the state, or the sentence
employed to express it), cognitive content (the kind of content that is belief-eligible
and truth-apt), descriptive content (the kind of content that represents, or
constitutes, a way the world might be), and nondescriptive content.
Although ‘content’-talk is extremely natural and convenient in metaethics
and in other branches of philosophy (as was ‘meaning’ talk in earlier decades),
we stress that in relation to cognitivist expressivism, this terminology needs to
be taken with a metaphysical grain of salt. On our account, talk of nondescriptive content is to be understood as not really positing any such items as overall
contents or cognitive contents; likewise for generic talk of overall content and
of cognitive content, construed as encompassing nondescriptive as well as
descriptive content. (We will take up ‘descriptive content’ presently.) Rather,
such talk is both syncategorematic and pleonastic.
It is syncategorematic in the sense that one can use such talk only in a fairly
restricted, fairly specific, range of syntactic-grammatical contexts. One can talk
about a psychological state or a declarative sentence as ‘having cognitive content’, about its ‘having cognitive content that is not descriptive’, and so forth.
One can even use definite descriptions like ‘the cognitive content of the belief ’,
but only in certain specific kinds of sentential contexts (e.g. contexts like ‘The
overall content of his belief is that Jones ought to apologize’.) But such talk is
not rightly construed as positing some entity that is a nondescriptive cognitive
content. On our picture, there is no such entity.
Generic content-talk is pleonastic in this sense: it normally functions as a way
of saying something that could pretty much be said some other way (although
the other way needn’t be outright synonymous with the original way). For example,
saying that moral judgments, and the utterances expressing them, ‘have
cognitive content’ is often, in effect, a way of saying that the judgments are
genuine beliefs and that the utterances are genuine assertions. Likewise, saying
that a psychological state has ‘overall content’ is often, in effect, a way of saying
that it is a state describable via a mentalistic characterization employing a
‘that’-clause construction.
Although such syncategorematic, pleonastic usage is very convenient,
one can employ it as we do to facilitate communication and exposition even if
one denies, as we do, that there are any such in-the-world items as nondescriptive contents. (Compare: One can sincerely utter the sentence ‘He has a loud
voice’, and this sentence can be true, even though one’s usage does not incur
any ontological commitment to such in-the-world items as voices.) The claim
262
Terry Horgan and Mark Timmons
that this terminology is ‘ontologically lightweight’ will receive theoretical
justification from within our own metaethical position, as articulated below.
Talk of descriptive content, on the other hand, is a different matter, from our
theoretical perspective. In this paper we will assume, at least provisionally, that
there are such items as descriptive contents—ways the world might be, and or
ways that some thing (or things) might be. Below we will invoke two kinds:
(i) propositional descriptive contents, expressible in English via closed sentences
like ‘Snow is white’, and (ii) non-propositional descriptive contents, expressible
in quasi-English via open sentences like ‘x is white’. We will leave it open how
exactly to construe such items metaphysically.⁸
We now proceed to our three tasks.
4. Task I: key generic features of belief
We begin with some prototypical, generic, features of belief—features associated
respectively with the phenomenology, the semantic assessability, and the
functional role of beliefs. We begin with the former. By ‘phenomenology of
belief ’ as used here, we mean to refer to the sort of subjective, what-it-is-like
experiential aspect of occurrent beliefs, something that unfortunately has
largely been ignored by contemporary philosophers of mind.⁹ In making
various observations about the phenomenology of belief, we will be discussing
ordinary nonmoral descriptive beliefs that most clearly purport to represent or
⁸ Perhaps even talk of descriptive content ultimately should be construed as pleonastic, syncategorematic,
and not ontologically committal. But even if this turns out to be so, such talk is still importantly different
metaphysically from talk of nondescriptive content: sentences and psychological states ‘with descriptive content’ have in-the-world truth-makers of some sort, whereas utterances and psychological states ‘with nondescriptive content’ do not. That difference would need to be respected and reconstructed, within any
ontological approach that treats talk of descriptive content as being syncategorematic and pleonastic itself.
⁹ That occurrent beliefs have a distinctive phenomenology was recognized by Hume who thought this
aspect of beliefs was crucial for understanding the difference between an occurrent belief with a certain
content and various nonbelief states having the same content—states such as entertaining without believing.
Hume distinguishes between the ‘ideas’ (roughly contents) and the manner of considering those ideas, and
explains:
[B]elief consists not in the nature and order of our ideas, but in the manner of their conception, and in their
feeling to the mind. I confess ’tis impossible to explain perfectly this feeling or manner of conception. We
make use of words, that express something near it. But its true and proper name is belief, which is a term that
every one sufficiently understands in common life. And in philosophy we can go no farther, than assert, that
it is something felt by the mind, which distinguishes the ideas of the judgment from the fictions of imagination. It gives them more force and influence; makes them appear of greater importance; infixes them in the
mind; and renders them the governing principles of all our actions. (1739: 629)
Hume, of course, attempts to use the features of force and vivacity to explain the nature of belief. Although
we agree with Hume that belief has a distinctive phenomenology, we do not follow him in trying to make
sense of this phenomenology (and associated functional role) in terms of force and vivacity.
Cognitivist Expressivism
263
describe some aspect of the world. But we will also be discussing occurrent
moral ought-judgments, calling attention to those phenomenological features
that they share with nonmoral descriptive beliefs and which (together with
considerations of semantic assessability and functional role) we claim qualify
them as genuine beliefs.
The phenomenology of belief
Here, then, is a list of five interrelated features of what we are calling the
phenomenology of occurrent belief. This what-it-is-likeness typically involves:
(1) psychologically ‘coming down’ on some issue, in a way that (2) classifies
(sometimes spontaneously) some ‘object’ of focus as falling under some
category, where one’s classificatory coming down is experienced (3) as
involuntary, (4) as a cognitive response to some sort of consideration that is
experienced (perhaps peripherally in consciousness) as being a sufficient reason
for categorizing as one does, and (5) as a judgment that is apt for assertion and
hence is naturally expressible in public language by a sentence in the declarative
mood. Each of these elements of phenomenology requires comment.
Occurrent belief involves the experience of ‘coming down’ on some issue,
where this manner of coming down may be preceded by some amount of
reflection, as when one is looking up at the October night sky and taking a close
look at a very bright celestial object and, after staring for a moment, comes to
believe that it is Mars. But very often belief comes about as an automatic
response to one’s surroundings, as when one catches a glimpse of a passing car
while watering the lawn. In this kind of case, by spontaneously and unreflectively taking it to be a car, one spontaneously psychologically comes down with
respect to the object-kind instantiated by a moving object passing through
one’s field of vision.
Turn now to all-things-considered occurrent ought-judgments. One
obvious feature of so judging is that one experiences this activity as a matter of
psychologically ‘coming down’ on whatever issue is under consideration. As
with descriptive beliefs, coming down with respect to some moral issue may be
preceded by deliberation. After listening to various opposing opinions about
the morality of same-sex marriages, Jones finds himself accepting the view that
such marriages are not morally wrong and that therefore a government ought
not to make such unions illegal. Jones, we are supposing, is initially undecided
about this issue, but later and in light of reflecting on various facts about
same-sex marriages, finds himself ‘coming down’ on one side of the issue. Of
course, many of our moral judgments, like descriptive beliefs, are formed spontaneously: they are triggered immediately in experience. Harman’s (1977: 4)
264
Terry Horgan and Mark Timmons
famous example of seeing some hoodlums igniting a cat and, without deliberation,
coming to think that what they are doing is wrong, is a case in point. In such
cases, one’s moral judgment is experienced as a fairly unreflective and immediate ‘coming down’ morally with respect to what one sees. Indeed, here people
often talk about just ‘seeing’ an action as wrong.
Clearly, making decisions and forming intentions are likewise experienced
as coming down with respect to some issue. Since intentions are not beliefs (we
are supposing), the experience of psychologically coming down is not alone a
sufficient phenomenological indicator of belief. The kind of psychological
coming down that seems fairly distinctive of belief, as opposed to intentionforming and the like, involves experiencing perceived or contemplated items as
falling under categories. Things get sorted experientially, when one comes
down on a matter in the belief-wise way. In the case of descriptive beliefs, things
get sorted experientially into descriptive categories, via category-concepts that
purport to represent some worldly object-kind or some property. Likewise, in
the case of a moral judgment that some action ought not to be done, a particular act-type or act-token gets sorted experientially into the category of those
actions that one is not to perform. We claim that the sorting/categorizing
aspect is central to the phenomenology of belief.
Another salient phenomenological feature of beliefs is their involuntariness.
One looks out the window and spontaneously and involuntarily believes that
the sun is shining. Sometimes reflection precedes one’s involuntarily coming
down on some issue. After mulling over various bits of evidence concerning a
mechanical failure, a trained mechanic involuntarily comes to believe that the
failure was due to a faulty oil pump. After inspecting the evidence, she just
sees what must have caused the failure. Regarding ought-beliefs, we have
already mentioned Harman’s case of unreflectively and spontaneously coming
to have a moral belief about the hoodlums, as well as a case in which Jones
comes to have a belief about same-sex marriages preceded by some amount of
reflection.¹⁰
Related to the involuntariness of beliefs is the fact that they are experienced
as possessing a kind of rational authority, consisting in their being grounded by
reasons. In the case of ordinary nonmoral perceptual beliefs about objects
and their properties, this experienced authority issues from one’s perceptual
experiences. In moving my hand across the desk, I feel its smoothness and
spontaneously come to believe that it is smooth. The confidence I experience
in having this belief is arguably part of the overall experience of occurrently
¹⁰ The spontaneity involved in both moral and nonmoral expertise is a major theme in Dewey (1922),
and is stressed in Dreyfus and Dreyfus (1990) and Varela (1992).
Cognitivist Expressivism
265
believing and partly explains why involuntarily psychologically coming down,
rather than just entertaining a corresponding hypothesis, is appropriate.
Similarly, the experience of morally coming down on some issue includes
experiencing this state as possessing a kind of rational authority—an authority
grounded in factual considerations that are experienced as rationally grounding
one’s moral judgment. In matters of morality, it is not ‘up to’ an individual what
to think about some matter of moral concern, just as it is not ‘up to’ someone
what to think about some nonmoral factual matter of concern. Sometimes this
feature of phenomenology is put in terms of the idea that judgments of moral
obligation have an ‘external’ authoritative source, just as do nonmoral factual
beliefs. It is by dwelling further on the phenomenology of moral authoritativeness that one can begin to distinguish ought-beliefs from nonmoral descriptive
beliefs, which, as we shall see, leads to considerations of the distinctive functional
role of such beliefs. We save further discussion of these matters for the next section.
Finally, the categorizing manner of psychologically coming down, in
occurrent belief, is experienced as a mental affirmation that is apt for assertion,
and thus is experienced as being naturally and spontaneously expressible in
public language by a sentence in the declarative mood.¹¹ Declarative-mood
sentences are the standard public-language vehicles employed for the speech
acts of asserting.
To sum up so far: There are a number of phenomenological features characteristic of belief as a distinct state type. These experiential features are possessed
by moral judgments, and so there is good prima facie reason for claiming that
such judgments are beliefs.
Prima facie, these very features can easily seem to require a construal of
moral judgments and moral utterances as being descriptive in their overall
content—or at least as purporting to be descriptive, even if there are no
in-the-world moral properties or facts. How, one might wonder, can one make
good sense of the classificatory, involuntary, reason-based, coming-down-ish,
phenomenology of moral judgments—including the experience of such
judgments as being apt for assertion—except by supposing that their overall
content is descriptive in purport? Addressing this challenge will be part of Task III.
Semantic assessability
It is grammatically permissible, and also common in practice, to ascribe truth and
falsity to moral judgments and statements. Such judgments and statements thus
appear to be semantically assessable. Furthermore, truth ascription seems entirely
¹¹ This aspect of belief is a case where the phenomenon we are pointing to is plausibly understood as
involving both phenomenological and functional role features.
266
Terry Horgan and Mark Timmons
natural and appropriate, given the belief-ish features of moral phenomenology
lately noted. If one definitively ‘comes down’ on the apartheid issue by finding
oneself having formed a judgment expressible as ‘Apartheid is wrong’, and one
expresses this moral judgment by asserting that apartheid is wrong, then one will
also think, and will be ready to assert, that it’s true that apartheid is wrong.
Semantic assessability presents an obvious challenge to moral irrealists:
making sense of truth ascription and falsity ascription within a general
approach that treats moral judgments, and the declarative sentences expressing
them, as having overall content that is not descriptive content. We need to deal
persuasively, within our cognitivist expressivism, with the line of thought that
says, ‘Well, if moral judgments lack descriptive, way-the-world-might-be, content, then their content cannot correspond or fail to correspond with how things
really are; so, such judgments cannot really be either true or false, even though
people often apply the terms “true” and “false” to them in ordinary discourse’.
Addressing this challenge too will be part of Task III.
Functional role
Beliefs are often characterized functionally as being psychological states that combine in distinctive ways with other psychological states to rationally-inferentially
yield further content-appropriate states as well as action. This is at least a partial characterization of their typical functional role, which is clearly illustrated
in cases where one’s belief combines with one’s desires leading one to form
intentions (assuming that intentions represent a distinct kind of psychological
state)—both long term and short term. Intentions in turn lead to action. My
desire to eat a taco for lunch together with my belief that Taco del Cielo is
around the corner (as well as a host of other content-appropriate beliefs and
desires) combine to yield an intention to go to Taco del Cielo in a few minutes.
The generic rational-inferential functional role of beliefs is a matter that
needs to be approached with some care, in our view. Although descriptive
beliefs do typically generate intentions and actions only in combination with
conative states like desires, moral beliefs—if such there be—might very well
have somewhat different prototypical functional roles in human mental life.
(More on this in section 6.) However, one crucially important and fully generic
aspect of the rational-inferential role of beliefs is this:
The overall content of any given belief can occur as an embedded
content-constituent of other, logically more complex beliefs; and beliefs with
such embedded content-constituents interact with other beliefs to generate
new beliefs via logical relations among their contents (specifically, via
logical-entailment relations).
Cognitivist Expressivism
267
We will call this feature logical embeddability. A familiar example is the following
piece of moral reasoning:
If it’s wrong to steal, then it’s wrong to get your little brother to steal.
It’s wrong to steal.
Therefore, it’s wrong to get your little brother to steal.
Logical embeddability has often been posed as a challenge to noncognitivist
positions. If what one is really doing in making a moral utterance is something
like expressing an attitude of disapproval or issuing a command, then how
exactly can one make sense of utterances in which moral content occurs
embedded, like the first premise in the argument just above? And how can one
make sense of the judgments expressed by such utterances?
Whatever one might think of the prospects for dealing with this challenge
within traditional noncognitivism, our own challenge is a different one, namely,
to make sense of the relevant, logically complex psychological states as fullfledged beliefs, and to make sense of logical-inferential relations involving such
beliefs, without supposing that moral belief-content is descriptive content.
This challenge is closely related to the preceding one about semantic
evaluability. Inferential connections among beliefs (and among sentences) are
normally explained in terms of truth and falsity: if the beliefs that are the
premises of an argument are true, then the belief that is the conclusion must
be true (in the case of logical entailment) or must be made likely to be true by
the premises (in the case of cogent inductive reasoning). Accommodating
logical embeddability thus goes hand in glove with accommodating truthaptness. Yet another burden for Task III.
This completes our first task of exploring the key generic features of belief
while at the same time calling attention to the fact that moral judgments
exhibit these features. As we have been saying, this creates at least a strong
prima facie case for genuine moral belief—while also posing various challenges
needing to be addressed by our cognitivist expressivism. We now turn to
various phenomenological and associated functional role characteristics that
are distinctive of moral judgments, with specific attention to moral oughtjudgments. (Hereafter we will explicitly call these judgments beliefs.)
5. Task II: the distinctive nature of ought-beliefs
As noted already, beliefs generally, and moral beliefs in particular, are experienced as grounded by a kind of rational authority. Talk of ‘rational authority’
268
Terry Horgan and Mark Timmons
while suggestive is quite vague, at least so far. But we can begin to illuminate
this idea by dwelling further on matters of moral phenomenology.
In characterizing the experienced authority attendant to first-person oughtbeliefs, we follow the lead of Maurice Mandelbaum who, in his 1955 The
Phenomenology of Moral Experience, distinguished what he called ‘direct’ from
‘removed’ judgments of moral obligation. Direct moral beliefs (judgments) are
first-person ought-beliefs, formed in some context in which the agent experiences a ‘felt demand’ on her own behavior, while removed ought-beliefs are
about what someone else (including one’s past self ) ought to do or ought to
have done. For the time being, we will focus on direct ought-beliefs, making
passing references to removed ought-beliefs. Later, in section 8, we will return
to the latter type of ought-belief.
Mandelbaum characterizes the felt demand involved in direct moral oughtbeliefs as a complex phenomenon involving the experience of an origin and a
direction. In judging that I ought to perform some action—that the action is
morally obligatory upon me—I experience a felt demand placed upon my
choices and consequent action issuing from the circumstances that I confront
(at least as I believe them to be). Thus, phenomenologically, one experiences
the demand as having an origin that is ‘external’ to oneself. It is this element of
the kind of felt demand characteristic of judgments of moral obligation that
distinguishes moral demands from the felt demands that are experienced as
having one’s own ‘internal’ desires or preferences as their origin. So, whereas
one’s own desires may issue forth in a kind of ‘internal’ felt demand for their
satisfaction by the agent who has them, the demandingness associated with a
judgment of moral obligation is ‘external’, coming from, as it were, the morally
relevant facts of the circumstances in which one finds oneself on some occasion. This aspect of one’s moral experience constitutes, then, the particular
manner in which moral ought-judgments are experienced as being grounded
in ‘objective’ reasons. Such reasons are factual considerations confronting
the agent that she takes to be morally significant. These considerations are
experienced as themselves demanding a certain course of action; they are, as
Mandelbaum puts it, the ‘origin’ in the phenomenology of felt demandingness.
The so-called direction of a felt demand has to do with whom, relative to
the judger, the obligation is directed against. In direct ought-judgments, the
obligation is directed against the judger herself; this is what Mandelbaum calls
a ‘reflexive demand’. By contrast, removed judgments of moral obligation
are experienced as directed against someone other than the individual making
the judgment. This element of the phenomenology of the experience of
direct ought-beliefs brings us to issues having to do with their distinctive
functional role.
Cognitivist Expressivism
269
Certain kinds of moral belief—direct ought-beliefs in particular—typically
play a motivationally ‘hot’ functional role in human mental life: they have
motivational force in and of themselves, apart from any pre-existing desires or
other ‘pro-attitudes’. The thought that first-person ought-beliefs are more
directly action-oriented than are ordinary nonmoral descriptive beliefs is what
inspires various forms of ethical internalism. This thought seems right and
important, even though a proper characterization of internalism—one that
allows, for instance, for the conceptual possibility of an ‘amoralist’ with moral
beliefs that have no motivational force at all—is a delicate matter.
Also, even though ought-beliefs are typically hot cognitions and thus need
not combine with a pre-existing desire in order to provide sufficient motivation to action, a sensible internalism should allow that, at least in atypical cases,
such ought-beliefs can work in concert with desire to produce action. Suppose,
for instance, that normally Joe’s judging that he ought to do something is
sufficient to move him to action. However, on some occasions, where Joe is
perhaps suffering from mild depression, his having a direct ought-belief will
only move him if he has certain content-appropriate desires which serve to
provide a supplemental motivational spark strong enough to move Joe to do his
duty. Perhaps Joe has a strong desire to maintain a certain self-image, so strong
that even in a state of mild depression his focusing on this desire of his, together
with his belief that failing to do what he ought would seriously damage his selfimage, moves him to act. So, even though direct ought-beliefs have as part of
their typical functional role a direct motivational force independent of any
pre-existing desires, such psychological states are still capable of combining
with prior desires to yield intention and subsequent action.
To sum up: we saw in the previous section that general phenomenological
considerations support the claim that moral judgments are genuine beliefs:
they involve an involuntary, categorizing, way of psychologically coming down
on some issue of moral concern, on the basis of considerations that are
experienced as rationally requiring the judgment—where this judgment is
experienced as truth-apt and hence as naturally expressed in thought and
language by sentences in the declarative mood. Moreover, the overall contents of
moral judgments can occur as embedded constituents in logically complex judgments, which then can combine with other judgments to logically-inferentially
generate new judgments. These are key generic features of belief. The
presumptive case for moral belief is thus strong.
However, moral judgments and, in particular, direct ought-judgments
exhibit some distinctive phenomenological and functional role characteristics
too; in particular, typically they are motivationally hot. So yet another
challenge to be addressed as part of task III is to explain how such judgments,
270
Terry Horgan and Mark Timmons
despite sometimes being motivational all by themselves, nonetheless can be
full-fledged beliefs. This brings us to our third task—the task of sketching a
framework for belief that can accommodate what we have been saying about
belief in general and moral belief in particular and which treats moral beliefs as
nondescriptive in their overall content.
6. Task III: a framework for belief ¹²
Earlier we noted that one deeply entrenched assumption of metaethics and
philosophical enquiry generally—the semantic assumption, as we call it—is that
all beliefs are descriptive beliefs. We reject this assumption. Having made our
prima facie case for moral belief, we turn now to the task of providing a general
framework for understanding belief (and also assertion) which incorporates
beliefs (and assertions) whose overall content is not descriptive—i.e. does not
purport to represent the world as being a certain way.
Two logically fundamental belief types: is-commitment and oughtcommitment
We begin with the logically most basic kinds of belief, leaving embeddability
matters until later. On the account we recommend, a belief is a certain kind of
commitment state—an affirmatory commitment—with respect to a descriptive
content that we call the belief ’s core descriptive content. There are two logically
fundamental belief types: is-commitment with respect to a core descriptive
content, and ought-commitment with respect to a core descriptive content. For
example, the belief that it is the case that Bush is US president, and the belief that
it ought to be the case that Bush is US president, are respectively an is-commitment
and ought-commitment vis-à-vis the same core descriptive content, namely,
that Bush is US president. (In the case of the ought-commitment, the core
descriptive content, that Bush is US president, differs from the belief ’s overall
content, that it ought to be the case that Bush is US president; and this overall content is not itself descriptive. In the case of the is-commitment, on the other
hand, the core descriptive content coincides with the overall content, namely,
that Bush is US president.¹³)
¹² This and the following two sections derive partly, but with significant refinements, from sections I and
III–V of Horgan and Timmons (2000a).
¹³ The ‘that’-clause ‘that Bush is US President’ expresses a way the world might be, as does the ‘that’-clause
‘that Gore is US president’. Neither clause says that the world is the way expressed (or that the world ought
to be the way expressed), because ‘that’-clauses by themselves don’t have assertoric force.
Cognitivist Expressivism
271
These two species of belief, involving two distinct ways of mentally
affirming a core descriptive content, are both sui generis: neither type of mental state is reducible to the other, and neither type is reducible to some kind of
nonbelief state such as an attitude of approval or a mental state appropriately
expressible via an imperative sentence. The way to understand the nature of the
beliefs we are calling ought-commitments is not by trying to reduce them to
something else, but rather by delineating their key features—both the generic
features they share in common with descriptive beliefs (is-commitments), and
the distinctive features that set them apart from descriptive beliefs. This is just
what we have lately been doing, in the course of addressing Task I and Task II.
Recall that Task III was formulated as follows: Set forth a theoretical account
of belief that simultaneously (i) treats moral judgments as genuine beliefs, (ii)
treats moral judgments as not descriptive in their overall content, (iii) accommodates the key distinctive characteristics of moral judgments in a way that
renders these features consistent with the claim that moral judgments are
genuine beliefs, and (iv) accommodates the key generic features of belief in a
way that is consistent with the denial that the overall content of moral beliefs is
descriptive. Part (i) is addressed by what we have just said, in conjunction
with our treatment of Task I in section 4 above: a moral judgment is an oughtcommitment with respect to a core descriptive content, and ought-commitments
are a species of belief because they possess the key generic features of belief
described in section 4.
Concerning part (ii), the crucial point is that ought-commitment is a
fundamentally different kind of affirmatory mental stance toward a core
descriptive content than is-commitment. An ought-commitment is not a mental state whose overall content is descriptive, representing a way the world
might be; hence it is not a state of mentally affirming that the world is such a
descriptively represented way. To construe moral beliefs in this manner is to
mistakenly assimilate them to descriptive beliefs, i.e. to is-commitments.
Rather, an ought-commitment is a distinct kind of mental affirmation vis-à-vis
a core descriptive content. Although there is a certain temptation to assimilate
ought-commitments to nonbelief states of the sort expressible linguistically by
nondeclarative utterances like ‘That Bush is US president, boo!’, or ‘US citizens,
do not elect Bush as US president!’, this temptation should be resisted—and can
be, once one gives up the semantic assumption. Ought-commitment is a sui
generis type of mental state, while also being an irreducible species of belief.
Although the overall content of ought-commitments is nondescriptive, nevertheless these states exhibit the key generic features that qualify them as beliefs.
Concerning part (iii), the motivationally ‘hot’ psychological role typically
played by first-person ought-judgments now gets smoothly accommodated.
272
Terry Horgan and Mark Timmons
Although this feature makes trouble for the idea that moral judgments
are beliefs insofar as one assumes that all beliefs have overall content that is
descriptive, it makes no trouble at all for us, because our framework rejects the
semantic assumption SA and treats ought-commitments as a distinctive species
of belief whose overall content is nondescriptive. Motivational hotness is an
important aspect of what constitutes ought-commitment in typical cases—
although we think that an adequate moral psychology also should allow for
cases in which the typical motivational force of moral belief is outweighed, or
suppressed, or even silenced altogether by other psychological states of the
agent.
Part (iv) of Task III will require more extensive treatment. We will segment
the discussion into several subsections that address respectively the three
challenges noted earlier in discussing Tasks I and II: accommodating the phenomenology of belief, accommodating truth-aptness, and accommodating the
key generic functional-role feature of belief, namely, inferential embeddability.
Accommodating the phenomenology of belief
In section 4 we described some key generic phenomenological features of
belief, and we argued that moral judgments exhibit these features. Is the
possession of these characteristics consistent with the contention that moral
judgments are ought-commitments whose overall content is nondescriptive?
Indeed it is. One can experience an occurrent ought-commitment as an
involuntary, classificatory, coming-down state vis-à-vis some descriptive
content, even though the overall content of this state is not descriptive.
The phenomenologically classificatory aspect of this coming-down state,
expressible in language via moral terminology (e.g. via ‘ought’), need not be a
matter of experiencing oneself to be mentally attributing some putatively inthe-world moral property to some act, agent, or state of affairs. Rather, it
can perfectly well be the experiential manifestation of the specific mode of
affirmatory commitment that the agent now instantiates with respect to the
given descriptive content—namely, ought-commitment.
Likewise, an occurrent ought-commitment can be based psychologically
upon descriptive considerations that are experienced as rationally grounding
this commitment state independently of the morally judging agent’s pre-existing
desires, even though the overall content of the commitment-state is not
descriptive.
Furthermore, an occurrent ought-commitment can be experienced as apt
for assertion by virtue of its involuntarily classificatory phenomenology and its
experienced authority, despite not being descriptive in its overall content.
Cognitivist Expressivism
273
So, since the declarative grammatical mood is the appropriate linguistic vehicle
for assertion, and since a sincere assertion is normally a belief-expressing speech
act, it is no surprise that occurrent ought-commitments are experienced as
psychological states appropriately expressible linguistically via declarative
sentences.
In short, the generic phenomenological features in virtue of which moral
judgments count as beliefs can perfectly well be present even if, as we maintain,
moral beliefs are ought-commitments whose overall content is nondescriptive.
These phenomenological features do not, in and of themselves, build descriptivity into moral judgment.
It remains possible even so, however, that the phenomenology of moral
judgment does include descriptivity, even though the phenomenological
features described in section 4 do not themselves entail it. That is, it remains
possible that the full what-it’s-like of classificatory coming-down, in moral
judgment, includes the experience of predicating a putative in-the-world moral
property. Well, does it in fact include such an experiential dimension?
Introspection, we submit, yields no ready or obvious answer to this question.
Moral belief is experientially much like descriptive belief, to be sure: both kinds
of state involve the experience of involuntary, classificatory, coming-down that
is based upon considerations as grounding-reasons. But beyond this patent and
powerful phenomenological similarity in the two kinds of belief, can one also
detect introspectively that moral belief is clearly like—or clearly unlike—
descriptive belief with respect to descriptivity per se? It seems not, either way.
Rather, the presence or absence of descriptivity as an aspect of the phenomenology of moral belief is a subtle question about which introspection does not
deliver a confident judgment.
This being so, wider theoretical considerations weigh in on the matter. For
one thing, there is no particular reason why moral phenomenology should
include descriptivity, given its functional role in human cognitive economy
and given the sociological role of moral discourse in human social intercourse.
On the contrary: since there are powerful theoretical reasons for denying the
existence of in-the-world moral properties and facts, the presence of descriptivity within moral phenomenology would constitute a built-in experiential
error with respect to the nature of external reality. Since descriptivity would be
a gratuitous and erroneous aspect of moral phenomenology, probably it is not
really an aspect of moral phenomenology at all.
Another theoretical consideration in support of this conclusion is the fact
that first-person moral judgments typically are motivationally hot, despite
qualifying as genuine beliefs. Descriptivity, though, would not comport
smoothly with this distinctive motivational role. Why should a belief concerning
274
Terry Horgan and Mark Timmons
putative in-the-world facts of any kind, qua factual belief, be intrinsically
motivating? Surely it comports better with the motivational role of moral judgments to say that they are beliefs of a different, non-factual, kind: namely,
non-descriptive ought-commitments.
These theoretical considerations are not conclusive, admittedly. The question whether moral phenomenology includes an aspect of descriptivity strikes
us as ultimately empirical—albeit an issue whose investigation would need to
be methodologically very subtle, and might well need to incorporate especially
careful introspection. But meanwhile, two key points need emphasis.
First, we take it that the dialectical burden is on those who would claim that
moral phenomenology does include descriptivity; they need to make a case for
this claim, given that simple introspection delivers no clear verdict either way.
Second, even if it should turn out that descriptivity is indeed an aspect of
moral phenomenology, this would not be terribly damaging to our metaethical
position. Although we would then be forced to claim that there is an erroneous
element in moral phenomenology, this would not undercut our contention that
moral judgments already qualify as beliefs anyway, for reasons independent of
their descriptivity—the reasons set out earlier in this paper. Nor would it
undercut our claim that moral judgments are a distinctive species of belief,
namely, ought-commitments. Qua ought-commitments, moral judgments
would be a sui generis kind of belief already, even if they turned out also to be
erroneous is-commitments that mistakenly predicate putative in-the-world
moral properties. And they would still play a crucial and legitimate actionguiding role in human psychology.¹⁴
One final point. Moral phenomenology may very well be susceptible to
influence by higher-order beliefs about the nature of morality itself. Certainly
many people believe that there are objective moral facts—a belief that can
easily be instilled, for instance, through the persistent intertwining of religious
¹⁴ J. L. Mackie’s version of error theory is much more philosophically problematic than would be the kind
of error theory just described, because Mackie in effect construed moral judgments as being only descriptive
beliefs whose contents involve putatively in-the-world moral properties. If that’s all there is to being a moral
belief, and if there are no such properties, then it becomes hard to see why moral thought and moral discourse
are not so hugely mistaken that they should be abandoned altogether. But if moral beliefs are also oughtcommitments with respect to core descriptive contents, then moral beliefs and moral assertions have important, legitimate, and indeed indispensable psychological and sociological roles to play even if they also include
an erroneous aspect of descriptivity.
Also worth noting is that moral realists too could accept our contention that moral judgments are oughtcommitments, and could accept that ought-commitments are a distinctive, sui generis species of belief. A
moral realist could claim that a moral judgment is both an is-commitment with respect to its overall content
and an ought-commitment with respect to its core descriptive content. This approach would have the advantage, for moral realists, of allowing them to acknowledge the internalist aspects of moral judgment. For further discussion of this theme, with explicit attention to the reasons for preferring our own cognitivist
expressivism to such a view, see Horgan and Timmons (2000a), especially section V.
Cognitivist Expressivism
275
instruction with moral education. For those who believe (perhaps only implicitly)
in objective moral facts, there may well arise a derivative kind of moral
phenomenology—induced by the interaction of this higher-order belief with the
more universal aspects of moral experience—that does include descriptivity.
But even if such erroneous moral phenomenology sometimes occurs by virtue
of the permeating effects of false beliefs about the metaphysics of morals, we
contend that the more fundamental, more universal, kind of moral experience
does not include an aspect of phenomenological descriptivity.
Accommodating semantic assessability
The concepts of belief, assertion, and truth are interconnected by a battery of
platitudes. For example, to sincerely assert some claim is to express one’s belief
regarding that claim; to believe a claim is to take that claim to be true; and
so forth. Since, on our view, moral judgments are genuine beliefs and moral
utterances are genuine assertions, our position must be able to make sense of
attributions of truth and falsity to moral judgments and utterances. Since we
advocate a robust form of irrealism in ethics, we claim there are no in-the-world
moral facts that could serve as truth-makers for moral beliefs and assertions.
Moreover, since we are nondescriptivists about moral thought and discourse,
we claim that moral beliefs and assertions lack overall descriptive content and
so we maintain that they are not in the business of purporting to represent or
describe the world: we are not error theorists. So the challenge for us is to make
sense of truth in ethics.
Our fundamental contention, in addressing this challenge, is that truth
ascriptions to statements and judgments with moral content are morally
engaged semantic appraisals—i.e. appraisals in which semantic evaluation is
‘fused’ with moral evaluation. These truth ascriptions thus are not descriptive,
because the overall content of the first-order judgments and utterances to
which they are applied is not descriptive. That the notion of truth should be
employable in a nondescriptive, morally engaged way is to be expected (given
our irrealist construal of moral concepts), since ordinary uses of the truth predicate normally operate in accordance with schema T. Since first-order moral
judgments and utterances have overall content that is not descriptive, the
accompanying truth ascriptions governed by schema T inherit this same
feature.
Although the point just made suffices to accommodate semantic assessability within our cognitivist expressivist position, we think there is quite a lot
more to say about the notion of truth in general and about its various uses with
respect to matters moral. Here we will briefly sketch some further views of ours
276
Terry Horgan and Mark Timmons
on these matters that are developed at greater length elsewhere (cf. Horgan
2001 and other papers cited therein; Horgan and Timmons 2002, 2006;
Timmons 1999: ch. 4). We mention these additional claims partly because we
think they speak to various potential concerns regarding our position, and
partly because we think they smoothly situate the morally engaged form of
truth-ascription within an independently plausible general conception of
how the notion of truth operates.
First, although in many contexts it is appropriate to employ the truth predicate in a morally engaged way in which one’s truth ascriptions run in tandem
with one’s first-order moral beliefs and assertions, there are also contexts in
which it instead becomes appropriate to engage in morally detached semantic
assessment. Under this usage, truth is a matter of correspondence to the
world, and falsity is a matter of noncorrespondence: a judgment or assertion is
true or false if and only if it has objective truth conditions, and otherwise
it lacks truth value. (It is true if the world satisfies those truth conditions—this
is correspondence—and it is false if the world fails to satisfy them—
noncorrespondence.) When the notion of truth is being employed in the
morally detached, correspondence-requiring manner, the proper thing to say
about moral judgments and assertions is that they are neither true nor false.
Second, first-order moral judgments and assertions, and likewise morally
engaged truth ascriptions, are typically categorical in content. Although they
are made from within a morally engaged stance in which one brings one’s own
moral standards to bear, they are not implicitly relativized to those standards
themselves. Relativism, about first-order moral claims and/or about truth
ascriptions to such claims, seriously misconstrues their content. In effect, relativism of this sort conflates morally engaged and morally detached usage, by
mistakenly treating engaged usage as a form of detached usage that incorporates implicit relativization to some specific set of moral standards.
Third, we advocate a general approach to concept–world and language–world
relations, and to the notion of truth, that we call contextual semantics. Some
leading ideas are these. (1) Truth is semantically correct affirmability, under
contextually operative semantic standards; falsity is semantically correct
deniability, under such standards. (2) Numerous concepts and terms, including the concept of truth itself, are governed by contextually variable semantic
standards of correct affirmability/deniability—where contextual variation can
occur not only across different subject matters, but even within thought or
discourse about a specific subject matter (e.g. morals). (3) Sometimes the
contextually operative semantic standards are tight, i.e. these standards conspire with how things are in the world to render a given judgment or statement
correctly affirmable or correctly deniable. (4) Thought and discourse governed
Cognitivist Expressivism
277
by semantically tight standards is descriptive. (5) Uses of the truth predicate
(and the falsity predicate) governed by semantically tight standards express
correspondence (or noncorrespondence) to the world. (6) Sometimes the contextually operative semantic standards are not tight. (7) Thought and discourse
governed by semantically non-tight standards is nondescriptive in its overall
content. (8) Uses of the truth predicate (and the falsity predicate) governed by
non-tight standards do not express correspondence or (or noncorrespondence)
to the world; when applied to moral thought and discourse, such truth ascriptions
are fused semantic/normative appraisals.¹⁵
Fourth, given the general framework of contextual semantics, it would be far
too crude to say that the detached and the engaged forms of semantic appraisal
involve ‘different concepts of truth’ or ‘different meanings of the word “true” ’.
Contextual variation in operative semantic standards, both at the first-order
level and at the level of truth ascription, is a much more subtle matter. The
same concept and the same meaning persist across such variation despite
identity-preserving changes from one context to another, in something like
the way a single person persists through time even while undergoing identitypreserving changes.
Logical embedding: accommodating the generic functional role of belief
How is logical embedding to be explained, within cognitivist expressivism? We
will now sketch the leading ideas of the account we recommend. We provide
further elaboration of technical details, plus further commentary, in the
Appendix.
Suppose that an agent, Tom, has a logically complex belief—say, the belief
that either Dick is cheerful or Harry ought to apologize. This belief does not
¹⁵ Another key thesis of contextual semantics is that tight semantic standards often operate in such a way
that a statement can be correctly affirmable under such standards—i.e. true—even though there are no
in-the-world objects or properties answering directly to the statement’s singular terms, predicates, or
unnegated existential quantifications. Such indirect correspondence, as we call it, does involve in-the-world
truth-makers (and falsity-makers), but the truth-making conditions need not include items in the world
answering directly to the given statement’s referential apparatus. (For instance, the statement ‘Mozart
composed 27 piano concertos’ can be true even if the objectively correct ontology does not include such
items as piano concertos.) Indirect correspondence is very important with respect to the ontological commitments of statements governed by tight semantic standards. But for present purposes it is secondary,
because we claim that the semantic standards at work in the engaged use of moral concepts are non-tight; this
means that moral statements lack even the indirect kind of correspondence (or noncorrespondence) to
reality. For a treatment of truth and objectivity with substantial similarity to our own contextual semantics,
see Wright (1992). Wright, however, seriously flirts with the idea of reducing truth to an idealized form of
epistemically warranted affirmability that he calls ‘superassertibility’, whereas we maintain that semantically
correct affirmability cannot be reduced to any form of epistemically warranted affirmability however
idealized (cf. Horgan 1995, 1996).
278
Terry Horgan and Mark Timmons
contain an ‘embedded is-commitment’ or an ‘embedded ought-commitment’,
whatever those might be. In holding this belief, Tom is not is-committed to the
content that Dick is cheerful, and is not ought-committed to the content that
Harry apologizes. Rather, the belief state is a logically complex commitment state,
one that obtains with respect to these core descriptive contents jointly. It is a
disjunctive commitment state.
Crucial and essential to such a logically complex commitment state, on the
view we are here suggesting, is its distinctive constitutive inferential role in the
cognitive economy of a cognitive agent (insofar as the agent does not exhibit
lapses in logical competence). Such a state is one that is poised to interact with
other potential beliefs to inferentially generate yet further beliefs that are inferentially ‘in the offing.’ For instance, if Tom occurrently has the disjunctive
belief about Dick and Harry, and Tom also occurrently has the belief that
Dick is not cheerful, these two beliefs together should inferentially generate
(insofar as Tom does not exhibit a rationality failure) the occurrent belief
that Harry ought to apologize. (Note that such inferential processes have a
phenomenological aspect too; there is something that it’s like to consciously
recognize such logical connections, and there is something that it is like for such
inferences to occur in conscious experience.) The constitutive inferential role of
logically complex beliefs also includes combining with other beliefs to yield certain implicit, logically grounded, further commitment states—even if these
remain implicit and perhaps fail to be psychologically operative in the agent.
Constitutive inferential role is a matter of logical consequence relations
among various beliefs. So we need to provide a way of construing the logical
consequence relation among belief-commitments of the various kinds:
is-commitments, ought-commitments, and logically complex commitments.
We will do so in three steps: first, describing a formal language whose logical syntax overtly models the various types of belief-commitment posited by our account;
second, providing formal semantics for sentences of this language, including a
definition of the relation of logical consequence; and third, explaining how this
formal semantics comports with our above-described treatment of the semantic
assessability of moral beliefs and assertions. (This three-step account is sketched in
this section and then developed more fully in the Appendix.)
Step 1. The formal language we propose employs two affirmatory-force operators, I[ ] and O[ ], whose respective analogues in English are the constructions
‘It is the case . . .’ and ‘It ought to be the case . . .’ The familiar atomic formulas
of predicate logic here are construed not as sentences but rather as closed nonsentential formulas. The natural analogues in English are ‘that’-clauses. So, for
example, letting Pb symbolize ‘that Bush is US president’, the formal sentence
I[Pb] says It is the case that Bush is US president, whereas O[Pb] says It ought to
Cognitivist Expressivism
279
be the case that Bush is US president. Grammatically, the operators I[ ] and O[ ]
are thus sentence-forming (and more generally, sentential formula forming)
operators.
Logically complex non-sentential formulas, including open ones containing
free variable-occurrences, are constructable within the formal language in the
standard ways described in predicate logic. Because they all are nonsentential formulas, however, the closed ones (i.e. those without free variable-occurrences)
are not sentences. A sentence (or open sentential formula) results from application of the syntactic operation of inserting a nonsentential formula into the
bracketed slot of I[ ] or O[ ]. For example, letting ‘Fx’ symbolize ‘x is a
Fraternity member’ and and ‘Gx’ symbolize ‘x takes out the Garbage’, the
formal version of the sentence It ought to be the case that some fraternity member
takes out the garbage results from insertion of the closed nonsentential formula
(᭚x)(Fx & Gx) into the bracketed slot of the operator O[ ], to yield the sentence O[(᭚x)(Fx & Gx)]. Although nonsentential formulas that can be
inserted into the bracketed slots of the operators I[ ] and O[ ] can be arbitrarily
complex, they all have descriptive content (insofar as the formal language is
semantically interpreted). Closed nonsentential formulas have propositional
descriptive content, and open ones have non-propositional descriptive content.
Turn now to logically complex commitment-types, like the disjunctive one
involved in Tom’s belief that either Dick is cheerful or Harry ought to apologize.
Within the formal language, such logically complex commitments are explicitly reflected by complex sentence-forming operators, which are built from the
primitive operators I[ ] and O[ ] by way of operator-forming connectives and
operator-forming quantifiers. For vividness, we use different symbols for these
connectives and quantifiers than for the connectives and quantifers that are
used to construct logically complex nonsentential formulas; we also use
boldface for all these operator-forming connectives and quantifiers. There is a
whole hierarchy of logically complex commitment-types, corresponding to the
various logically complex sentence-forming operators.
For instance, Tom’s disjunctive belief that either Dick is cheerful or Harry
ought to apologize involves a logically complex commitment-type expressible by
the complex, disjunctive, sentence-forming operator (I[ ] o O[ ]). If the closed
nonsentential formulas ‘Cd’ and ‘Ah’ respectively symbolize that Dick is cheerful and that Harry apologizes, then Tom’s disjunctive belief about Dick and
Harry is formally expressible by the sentence that results from respectively
inserting these formulas into the operator’s respective slots, thus: (I[Cd] o
O[Ah]). Ontologically, Tom’s belief is a logically complex commitment-state
of type (I[ ] o O[ ]), with respect to the pair of propositional contents that Dick
is cheerful and that Harry apologizes.
280
Terry Horgan and Mark Timmons
For an example of a complex operator with quantification, suppose that Tom
believes that there is a specific fraternity member who ought to take out the garbage.
This belief involves a logically complex, quantificational, commitment-type
whose structure is reflected by the sentence-forming operator (⌺ )(I[ ] O[ ]).
A formal sentence expressing Tom’s belief results from inserting a variable into
the existential-quantificational slot of this complex operator and inserting
open sentential formulas into the operator’s bracketed slots, thus: (⌺x)(I[Fx]
O[Gx]). Ontologically, Tom’s belief is a logically complex commitment-state
of type (⌺ )(I[ ] O[ ]), with respect to the pair of non-propositional descriptive contents that x is a fraternity member and that x takes out the garbage.
Step 2. In the Appendix we employ the idea of a valuation, i.e. an assignment
of the truth values T and F to some (but not necessarily all) of the sentences and
closed nonsentential formulas of the formal language, and we then use this idea
to define the relation of logical consequence.
Step 3. Nothing in the formal semantics set forth in the Appendix requires
that the notion of truth employed in the definition of valuation be understood
as operating in the morally detached ‘correspondence’ manner. On the contrary, insofar as the formal language is construed as an interpreted language
rather than an uninterpreted formal calculus, the truth values assigned to sentences by a given valuation can perfectly well be those that reflect a given agent’s
morally engaged truth assessments. (These run in tandem with the agent’s
morally engaged first-order beliefs, in accordance with schema T.) Thus, the
definition of logical consequence in the Appendix can likewise be understood
as reflecting logical relations among an agent’s various morally engaged beliefs
(including logically complex beliefs), and as also reflecting the implicit beliefcommitments logically generated by these beliefs.
The notion of valuation is defined so as to allow for valuations in which certain sentences and/or nonsentential formulas are assigned neither T nor F. This
is because, insofar as a valuation reflects a given agent’s nonmoral and moral
beliefs, certain sentences might be ones whose overall content the agent holds
neither true nor false (even when using the truth predicate in a morally engaged
way), but instead is agnostic about.
On the other hand, there will also be a valuation that assigns truth values in
accordance with a morally detached correspondence-usage of ‘true’, and that
furthermore assigns truth and falsity based on whether or not a given sentence
actually corresponds to how things are or not (rather than on the basis of any
particular agent’s beliefs). In such a valuation, all sentences of the form O[A]
will be assigned neither T nor F.
In short: In the formal language we have described, there is a hierarchy of
sentence-forming operators with logical structures that explicitly reflect the
Cognitivist Expressivism
281
various kinds of commitment, both simple and logically complex, that constitute
distinct types of belief. The nonsentential formulas, corresponding to ‘that’clauses in natural language, have descriptive content but not assertoric force. A
sentence, constructed by inserting the respective members of a sequence of
nonsentential formulas into the respective bracketed slots of a sentence-forming
operator (and inserting suitable variables into the respective quantificational
slots, to bind all free variable-occurrences in the inserted formulas), reflects the
ontological structure of the belief it expresses: the belief is a certain type
of commitment-state with respect to a sequence of propositional or nonpropositional descriptive contents—where the commitment-type is expressed
by the relevant sentence-forming operator, and the respective descriptive
contents are expressed by the closed and/or open nonsentential formulas
inserted into this operator’s respective bracketed slots. The notion of a valuation—a logically permissible assignment of the truth values T and F—can be
defined for this formal language, and the relation of logical consequence can be
defined using this notion. This definition of logical consequence comports
well with the contention that truth ascription to sentences whose overall content is nondescriptive is a matter of morally engaged semantic evaluation in
which the evaluator’s moral and semantic standards are fused. The definition of
logical consequence thereby comports well with cognitivist expressivism.
In addressing Task III, we have described belief as affirmatory commitment
with respect to one or several core descriptive contents. There are two logically
fundamental belief-types: is-commitment and ought-commitment. Although
an ought-commitment with respect to a core descriptive content is indeed a
species of belief, its overall content is nondescriptive; nevertheless, it does have
the involuntary, classificatory, coming-down phenomenology that is distinctive of occurrent beliefs. Truth ascription to beliefs and assertions with moral
content is a morally engaged fusion of moral and semantic evaluation, and thus
comports with the fact that moral content is not descriptive. There are also
logically complex belief-types: kinds of affirmatory commitment expressible
by logically complex sentence-forming operators in the formal language we
have proposed as modeling the ontological structure of beliefs. Beliefs bear logical consequence relations to one another, whether or not they have overall
content that is descriptive.¹⁶
We have mainly been dwelling on matters of moral psychology, arguing that
moral judgments are beliefs whose overall content is nondescriptive. Similar
points can be made about moral utterances. These are assertoric speech acts
¹⁶ In the Appendix we explain how the descriptive/nondescriptive distinction applies to beliefs involving
logically complex commitments, and to the sentences expressing such beliefs.
282
Terry Horgan and Mark Timmons
that play a distinctive sociolinguistic role—a role in interpersonal dynamics.
An assertion, we claim, is best understood as a stance-taking speech act, an act
through which one overtly comes down on some issue and thereby expresses
an is-commitment, or an ought-commitment, or a logically complex commitment with respect to one or more core descriptive contents. In so doing one
positions oneself within the context of sociolinguistic interaction, vis-à-vis the
core content(s). A stance is therefore an orientation that one occupies in an
interpersonal situation. An ought-stance in particular is typically an actionguiding stance with respect to some core descriptive content. Some oughtstances are more directly action-guiding than others, but they are all distinctively
action-oriented—just as removed ought-beliefs are still tied to action,
although less directly so than direct ought-beliefs. Moreover, just as moral
beliefs typically involve a responsiveness to reasons, so engaging in a moral
stance-taking speech act normally involves occupying a sociolinguistic role
involving the preparedness to give reasons for one’s moral stance on some issue.
For instance, by asserting that Jones ought not to lie to his neighbor, one
thereby takes a moral stance in which one signals one’s willingness to engage in
interpersonal reason-giving with respect to one’s own ought-commitment,
and to defend one’s commitment against objections, or else give up one’s
commitment.
So sincere utterances of declarative sentences with moral content are speech
acts of assertion, even though the overall content of such sentences is not
descriptive. Of course, sincere utterances of sentences with descriptive overall
content are assertions too. But it should be noted that token descriptive sentences
also can be construed as making belief-independent assertions (as we will call
them)—i.e. assertions that are independent of any specific asserter(s). For
instance, an inscription on a subway wall of the sentence ‘There are no weapons
of mass destruction in Iraq!’ asserts—in and of itself, apart from its author(s)—
that there are no weapons of mass destruction in Iraq. According to cognitivist
expressivism, however, sentence tokens whose overall content is not descriptive
cannot rightly be construed as making belief-independent assertions. If an
inscription of ‘Abortion ought to be outlawed!’ is written on a subway wall, then
it can be correctly regarded as an assertion only insofar as one interprets it as
expressing an ought-commitment of some person(s), known or unknown.
7. More on methodology
Having completed our three main tasks in setting out our metaethical view, let
us pause to reflect a bit more on matters of metaethical methodology. Then,
Cognitivist Expressivism
283
following this section, we will briefly consider various additional tasks that a
view like ours must eventually tackle.
Recall from section 2 that we take a page from Moore in refusing to
‘reduce’ moral thought and discourse to any other type of discourse. We
can now be more explicit in explaining what we are refusing to do and in
explaining what we think can be done by way of illuminating such thought
and discourse.
First, we refuse to set forth truth conditions for moral beliefs and assertions
except, of course, non-substantive ones such as, for example, ‘Slavery is wrong’
is true if and only if slavery is wrong. The typical expectation in giving some
sort of semantically illuminating set of truth conditions for a form of statement
is (in effect) to set forth a set of non-trivial substantive truth-makers for
such claims. On our view, it is a mistake to suppose that there are substantive
truth-makers for moral beliefs and assertions; our view is robustly irrealist in
this respect.
Second, we refuse to engage in any sort of paraphrase of moral ‘ought’ judgments that would effectively ‘reduce’ them to some other type of judgment, or
to some combination of those. We have in mind any sort of ‘analysis’ that
would construe an ‘ought’ judgment of the form ‘S ought to do A’ as equivalent in meaning to (say): ‘S, do A!’ directed toward oneself or another. We also
eschew any weakly reductive strategies, such as those that treat certain nondeclarative linguistic constructions (e.g. imperative sentences) as models for
moral thought and discourse—models that supposedly provide semantic
and metaphysical illumination without necessarily yielding translations or
paraphrases.
But these denials do not mean that we are quietists about the possibility of
illuminating the meaning of moral thought and discourse. Like Hare, quoted
earlier in section 2, we think that proper illumination comes from understanding the distinctive point and purpose of moral ‘ought’ judgments in moral
thought and discourse as well as their associated phenomenology. So, there is a
methodological component to our brand of cognitivist expressivism that we
may formulate as follows:
Moral thought and language does not admit of any kind of semantic
‘reduction’; rather it is sui generis. Moreover, a proper understanding of
such thought and discourse involves understanding the distinctive
phenomenology and associated functional roles of such psychological
states and sociolinguistic speech acts.
In articulating the key elements of our view, we have been illustrating our
Moorean nonreductive methodology.
284
Terry Horgan and Mark Timmons
8. Remaining tasks—brief progress report
There is a battery of challenges that any view like ours must meet. In the
remainder of this paper we can only indicate our current thinking about how
to go about addressing some of these; dealing with them fully will have to be
left as remaining tasks. The challenges we have in mind concern: (1) extending
our account of direct ought-beliefs to other types of moral belief, (2) moral
progress, (3) reasons, (4) moral seriousness, (5) lingering adherence to the
descriptivist conception of belief. We now proceed to take them up in order.
Extending the account
In order to fully defend cognitivist expressivism, we need to extend our account
of direct ought-beliefs to encompass removed ought-beliefs, beliefs about value
(goodness), and moral beliefs that employ ‘thick’ moral concepts such as
benevolence, courage, and malice. For present purposes, we will only consider
removed ought-beliefs, again following the lead of Mandelbaum.
According to Mandelbaum, removed judgments of moral obligation, like
direct judgments, are a response to one’s experiencing some action as ‘fitting’
vis-à-vis the circumstances confronting an agent. But as distinct from direct
judgments, removed judgments are: (1) made from a third person, spectator’s
point of view, (2) typically expressed by sentences of the form ‘S ought to
do/have done A’ (where ‘S’ might refer to one’s past self as well as to another
person), and (3) are connected with an agent’s motivation relatively indirectly
and are thus further removed from an agent’s motivation to act accordingly.
Phenomenologically, and in contrast to direct judgments, there is obviously
no felt reflexive demand, upon the individual engaged in this kind of judging,
to perform or not perform the action whose performance or avoidance is
judged to be obligatory. However, such judgments often ground attitudes of
disinterested approval or disapproval toward the action being evaluated. Here
is how Mandelbaum describes this basic contrast:
Removed moral judgments, as we have seen, involve attitudes which may be denominated as ‘selfless’ or ‘detached’; they are ‘contemplative’ rather than being states of the
self. On the other hand, in a direct moral judgment the element of reflexive demand
evokes emotion; this emotion, like fear or anger, is experienced as a state of the self and
is directly related to action. Thus, the stirredupness and pressure which are present in
direct moral judgments have no counterpart in removed moral judgments. In the latter
we approve or disapprove, or we may feel admiration, disgust, contempt, or loathing;
but even when these stronger affective states are present they appear as by-products of
our acts of moral apprehension, and not as direct manifestations of what are experienced
to be motivational forces. (1955: 127)
Cognitivist Expressivism
285
So with removed ‘ought’ judgments, in contrast to direct ought-commitments,
the element of felt reflexive demand is absent and thus the motivational role of
the judgment differs from direct moral judgments.¹⁷
So, direct and removed ought-beliefs differ in that beliefs of the former type
involve a felt reflexive demand whose role is directly action-guiding, whereas
the latter type of belief lacks these characteristics. Further exploration of the distinctive phenomenology and functional role of these types of commitment-state
cannot be undertaken here. Nevertheless, both types of ought-commitment
state are genuine beliefs—they both involve involuntarily coming down on
some matter of moral concern, and they both are typically grounded in reasons
and thus possess a kind of felt authority.
Moral progress
For a metaethical descriptivist-realist, intellectual moral progress is made when
one’s moral beliefs come to better approximate the moral facts. For a descriptivistconstructivist, intellectual moral progress is made when one’s moral beliefs
come to better accord with the moral norms that would be accepted by
individuals under certain idealized conditions. Indeed, for a cultural moral
relativist, there can be individual moral progress in the sense that one’s moral
beliefs come to better approximate the moral norms (either actual or ideal,
depending on the version of relativism) of her culture. But if moral belief and
assertion are not to be understood as purporting to describe or represent substantive moral facts, then how can one make sense of genuine moral progress?
Furthermore, if there is no metaphysical anchor for moral thought and
discourse, then why take it seriously, why not construe moral dispute and
discussion as being more like disputes over fashion in clothes and matters of
taste generally?
These challenges focus on our irrealist moral metaphysics, and we consider
them to be among the most difficult for any robust moral irrealist. Here, then,
is an indication of how we would respond to these challenges, though they certainly deserve more attention than it is possible to give them here.
On our view, moral progress is not to be understood as a matter of bringing
one’s beliefs into closer proximity to some realm of moral facts. Instead, we
propose thinking of moral progress as something to be judged from within a
¹⁷ However, this is not to say that removed moral judgments do not have an important tie to motivation.
Hare (1952), for instance, maintained that because moral judgments are grounded in reasons and thus commit one to a moral principle which expresses those reasons, in making a removed moral judgment, one is
committing oneself to act in a certain way were one to face the circumstances in which the agent being judged
is placed. In this way such judgments are at least indirectly action-guiding.
286
Terry Horgan and Mark Timmons
committed moral outlook: when one makes judgments about moral improvement,
one does so from an engaged moral perspective. So, for example, in judging that
one’s current view about the morality of euthanasia is better (more correct)
than one’s former view about this issue, one is not simply registering the fact
that one’s view on this matter has changed—mere change is not equivalent to
progress! Rather, one is making, based on what one experiences as an improved
understanding of the morally relevant aspects of euthanasia (including various
types of cases this practice covers), a moral judgment—a judgment backed by
reasons. In other words, judgments about moral progress are morally engaged
themselves, to be understood according to the general metaethical picture we
have been sketching.
Reasons
According to our characterization of moral beliefs, such commitment-states
are typically grounded in what one experiences as reasons for the belief or assertion in question. We are irrealists about moral properties and facts, but what
about moral reasons? Is our view committed to realism about moral reasons? If
it is, then our overall metaphysical view countenances normative properties
and facts—properties and facts having to do with reasons. But then, why be
irrealists about normative properties and facts such as goodness and rightness?
On the other hand, if we are reasons irrealists, how do we propose to make
sense of such claims as, ‘The fact that her lie was motivated by pecuniary
self-interest is a reason for concluding that her lie was wrong’?
Here, we embrace the spirit if not the letter of C. L. Stevenson’s way of
dealing with this issue on behalf of his version of noncognitivism. According to
Stevenson, claims about some nonmoral fact R constituting a reason for or
against some moral judgment are themselves moral claims made from within a
morally engaged outlook. Here is how Stevenson put the point:
So, the general situation is this: when we claim that the factual reason, R, if true, would
justify or help to justify the evaluative conclusion, E, we are in effect making another
value judgment, E⬘, of our own—the latter serving to evaluate the situation that we
shall have if the facts of the case include those that R purports to describe. (1963: 89)
Thus, an enquiry into those nonmoral factual considerations that serve as good
reasons for accepting or denying moral statements is what moral thinking is all
about. And various normative moral theories have been proposed (versions of
consequentialism, deontology, virtue ethics, and so forth) in answer to questions about reasons in ethics. Moral reasons claims, then, express substantive
moral beliefs and are themselves to be understood (depending on the type of
Cognitivist Expressivism
287
reasons claim in question) in the general manner set forth in our version of cognitivist expressivism.
Moral seriousness
How, on our view, can sense be made of the fact that people take their moral
views with utter seriousness? If there are no moral facts to which our moral
beliefs and assertions must answer, then why care so much about morality?
In response, we think that the sort of challenge being posed is best construed as
a moral challenge: why ought people to take their moral views seriously? And the
appropriate response is to give moral reasons—reasons that, for instance, will
likely appeal to the important role that morality plays in people’s lives. Such moral
reasons are not hard to find. And here again, our understanding of the challenge is
to take it as appropriately dealt with from within an engaged moral outlook.
Beliefs as descriptive
A final challenge worth mentioning will come from those who continue to
embrace the semantic assumption we described in section 1, and who insist
that genuine, full-fledged beliefs must be psychological states whose overall
content is descriptive.
We have two points to make in response. First, we have offered a battery of
arguments supporting both the claim that moral judgments are genuine beliefs
and the claim that the overall content of a moral judgment is nondescriptive,
and we have offered responses to various challenges that such a position faces.
Given these arguments and responses, there is a substantial burden of proof
upon those who wish to insist nonetheless that real beliefs must be descriptive.
The fact that the semantic assumption has traditionally been so widely
accepted does not alone suffice to justify its acceptance.
Second, even if it should turn out that genuine beliefs really must be descriptive in their overall content, a variant of our cognitivist expressivism would still be
available, and would constitute a significantly novel, nontraditional, version of
noncognitivism. This variant position, which might be called quasi-cognitivist
expressivism,¹⁸ would deny that ought-commitments are a species of belief, but
otherwise it would look very much like our own view. It would embrace the
claims (1) that these states are sui generis and irreducible, (2) that they share
with beliefs certain key phenomenological features involving involuntary,
reason-based, classificatory ‘coming-down’ on an issue, (3) that they are subject
¹⁸ This label was suggested by Uriah Kriegel.
288
Terry Horgan and Mark Timmons
to morally engaged semantic assessment, and (4) that there are logical
consequence relations among (i) ought-commitments, (ii) beliefs whose overall content is descriptive, and (iii) logically complex commitment-states.
In section 2 we pointed out that traditional forms of noncognitivism are at
least weakly reductive, by virtue of their reliance on certain nondeclarative
grammatical constructions (e.g. universal imperatives) as providing a putative
model of moral discourse—even though some prominent advocates of such
theories (e.g. Hare) repudiated the kind of strong reductionism that asserts the
outright synonymy or semantic equivalence of moral utterances with such
nondeclarative utterances. Quasi-cognitivist expressivism, on the other hand,
is not reductive even in the weak sense, because it shares with our own position
a reliance on certain declarative grammatical constructions as providing a
model of moral discourse and of the ontological structure of moral judgments—namely, sentences containing the ought-operator O[ ], in the formal
language we have described. Eschewing weak reductionism is thus another
important respect in which the fallback view we call quasi-cognitivist expressivism is similar to our own position and is different from traditional forms of
noncognitivism.
9. Conclusion
We have covered quite a lot of ground, sketching a general phenomenology of
belief, developing a general framework of belief (and assertion) that treats some
beliefs and assertions as having overall content that is not descriptive, and arguing that there is reason to construe moral beliefs and assertions as nondescriptive. A cognitivist version of expressivism thus emerged. After saying a bit more
about our methodology, we concluded by considering a battery of likely challenges often raised against expressivist views in metaethics, and briefly
explained how our view answers them. We maintain that our brand of cognitivist expressivism is superior to its metaethical competitors and is well worth
developing further.¹⁹
Appendix
We here describe in detail the formal language we propose, with sentences whose
syntactic structure models the ontological structures that we claim are possessed by
¹⁹ Our further explorations will be recorded in our works in progress, Moral Phenomenology and Moral
Theory and Expressivism’s Progress.
Cognitivist Expressivism
289
various kinds of belief. We also set forth formal semantics for this language, including
a definition of the logical consequence relation. We then add a number of observations
by way of commentary.
Syntax
Primitive symbols
Non-boldface: name letters, predicate letters, individual variables, connectives ⬃, ν,
&, ⊃, ⬅, quantifier symbols ᭙ and ᭚, parentheses, commas.
Boldface: operator symbols I, O, Bel, connectives ¬, , , →, ↔, quantifier symbols ⌸ and ⌺, parentheses, brackets.
Non-sentential formulas (nsf ’s):
If P is an n-ary predicate letter or a complex n-ary predicate, and T1, . . . ,Tn are terms
(not necessarily distinct) each of which is a name letter or a variable, then PT1, . . . ,Tn
is a nonsentential formula. (Complex n-ary predicates are defined below.)
If A is a nonsentential formula and X is a variable, then ⬃A, (᭙X)A, and (᭚X)A are
nonsentential formulas.
If A and B are nonsentential formulas, then so are (A ν B), (A & B), (A ⊃ B), and
(A ⬅ B).
Nothing else is a nonsentential formula.
If A is a nonsentential formula with no free variable-occurrences, then it is closed.
Otherwise it is open.
Sentential-formula forming operators (sff operators):
I[ ] and O[ ] are sff operators. (The slots in these operators are bracketed slots.)
If ⍀ is an sff operator, then ¬ ⍀, (⌸ )⍀, and (⌺ )⍀ are sff operators. (The indicated
slots are quantificational slots.)
If ⍀ and ⌬ are sff operators, then so are (⍀ ⌬), (⍀ ⌬), (⍀ → ⌬), and (⍀ ↔ ⌬).
Nothing else is an sff operator.
Sentential formulas:
If ⍀ is an sff operator containing n quantificational slots and m bracketed slots, and
Q is a sequence of n variables (not necessarily distinct), and F is a sequence of m nonsentential formulas (not necessarily distinct), then ⍀//(Q, F ) is a sentential formula.
(Notation: ⍀//(Q, F) is the expression obtained by inserting the respective elements
of Q into the respective left-to-right quantificational slots in ⍀ and inserting the
respective elements of F into the respective left-to-right bracketed slots in ⍀.)
Nothing else is a sentential formula.
A sentential formula containing no free variable-occurrences is a sentence.
290
Terry Horgan and Mark Timmons
Predicate-forming-operator forming operators (pfo forming operators):
Bel[ ] is a pfo forming operator.
Bel( )[ ] is a pfo forming operator.
Nothing else is a pfo forming operator.
Predicate-forming operators (pfo’s):
If ⍀ is a sff operator, then Bel[⍀] and Bel( )[⍀] are predicate-forming operators.
Nothing else is a predicate-forming operator.
Complex predicates:
If ⍀//(Q, F) is a sentence, then Bel[⍀]//(Q, F) is a complex 1-place predicate.
If ⍀//(Q, F) is a sentential formula containing free occurrences of all and only the
distinct variables X1, . . . , Xn, then Bel(X1, . . . , Xn)[⍀]//(Q, F) is a complex (n ⫹ 1)ary predicate. (The operator Bel( )[ ] binds the initial occurrences of X1, . . . , Xn in
Bel(X1, . . . , Xn)[⍀]//(Q, F), and also binds all occurrences of these variables that are
free within ⍀//(Q, F).)
Nothing else is a complex predicate.
Semantics
A nonsentential valuation N is an assignment of the truth values T and F to some (but
not necessarily all) closed non-sentential formulas, in conformity with the following
conditions.
No closed nonsentential formula is assigned both T and F by N.
If A is a closed nonsentential formula, then
N assigns T to A iff N assigns F to ⬃A.
N assigns F to A iff N assigns T to ⬃A. (So N assigns neither T nor F to A iff N
assigns neither T nor F to ⬃A.)
If A and B are closed nonsentential formulas, then
N assigns T to (A ν B) iff either N assigns T to A or N assigns T to B.
N assigns F to (A ν B) iff N assigns F to both A and B.
N assigns T to (A & B) iff N assigns T to both A and B.
N assigns F to (A & B) iff either N assigns F to A or N assigns F to B.
N assigns T to (A ⊃ B) iff either N assigns F to A or N assigns T to B.
N assigns F to (A ⊃ B) iff N assigns T to A and N assigns F to B.
N assigns T to (A ⬅ B) iff either N assigns T to both A and B or N assigns F to both
A and B.
Cognitivist Expressivism
291
N assigns F to (A ⬅ B) iff either N assigns T to A and F to B, or N assigns F to A
and T to B.
If A is a nonsentential formula in which the only free variable-occurrences are occurrences of X, then
N assigns T to (᭙X)A iff for every name letter N, N assigns T to A(X/N).
(Notation: A(X/N) is the result of replacing every free occurrence of X in A by N.)
N assigns F to (᭙X)A iff for some name letter N, N assigns F to A(X/N).
N assigns T to (᭚X)A iff for some name letter N, N assigns T to A(X/N).
N assigns F to (᭚X)A iff for every name letter N, N assigns F to A(X/N).
A closed nonsentential formula A is a nonsentential consequence of a set of closed nonsentential formulas {B1, . . . ,Bn} iff (1) every nonsentential valuation that assigns T to
each of B1, . . . Bn also assigns T to A, and (2) every nonsentential valuation that does
not assign F to any of B1, . . . Bn also does not assign F to A.
A valuation V is a pair ⬍ N, S⬎ such that N (the nonsentential element of V) is a nonsentential valuation and S (the sentential element of V) is an assignment of truth values
to some (but not necessarily all) sentences in conformity with the following conditions:
No sentence is assigned both T and F by S.
If A is a closed nonsentential formula, then
S assigns T to the sentence I[A] iff N assigns T to A.
S assigns F to the sentence I[A] iff N assigns F to A.
S assigns T to the sentence O[A] only if for every closed nonsentential formula B
that is a nonsentential consequence of A, S assigns T to O[B].
S assigns F to the sentence O[A] only if for every closed nonsentential formula B
such that A is a nonsentential consequence of B, S assigns F to O[B].
If ⍀//(Q, F) is a sentence, then
S assigns T to ¬⍀//(Q, F) iff S assigns F to ⍀//(Q, F).
S assigns F to ¬⍀//(Q, F) iff S assigns T to ⍀//(Q, F).
If ⍀//(Q1, F1) and ⌬//(Q2 , F2 ) are sentences, then
S assigns T to (⍀ ⌬)//(Q1, F1, Q2, F2) iff either S assigns T to ⍀//(Q1, F1) or S
assigns T to ⌬//(Q2, F2). (Notation: (⍀ ⌬)//(Q1, F1, Q2, F2) is the expression
obtained by inserting the respective elements of Q1 and F1 into the respective
quantificational and bracketed slots of the ⍀ segment of (⍀ ⌬), and likewise
inserting the respective elements of Q2 and F2 into the slots of the ⌬ segment of
(⍀ ⌬).)
S assigns F to (⍀ ⌬)//(Q1, F1, Q2 , F2) iff S assigns F to both ⍀//(Q1, F1) and
⌬//(Q2 , F2).
Terry Horgan and Mark Timmons
292
S assigns T to (⍀ ⌬)//(Q1, F1, Q2, F2) iff S assigns T to both ⍀//(Q1, F1,) and
⌬//(Q2, F2).
S assigns F to (⍀ ⌬)//(Q1, F1, Q2, F2) iff either S assigns F to ⍀//(Q1, F1) or S
assigns F to ⌬//Q2, F2).
S assigns T to (⍀ → ⌬)//(Q1, F1, Q2, F2) iff either S assigns F to ⍀//(Q1, F1) or S
assigns T to ⌬//Q2, F2).
S assigns F to (⍀ → ⌬)//(Q1, F1, Q2, F2) iff S assigns T to ⍀ //(Q1, F1) and S
assigns F to ⌬//(Q2, F2).
S assigns T to (⍀ ↔ ⌬)//(Q1, F1,Q2, F2) iff either S assigns T to both ⍀ //(Q1, F1)
and ⌬//(Q2, F2) or S assigns F to both ⍀ //(Q1, F1) and ⌬//(Q2, F2).
S assigns F to(⍀ ↔ ⌬)//(Q1, F1,Q2, F2) iff either S assigns T to ⍀ //(Q1, F1) and
F to ⌬//(Q2, F2) or S assigns F to ⍀ //(Q1, F1) and T to ⌬//(Q2, F2).
If X is a variable and ⍀//(Q, F) is a sentential formula in which the only free variableoccurrences are occurrences of X, then
S assigns T to (⌸X)⍀//(Q, F) iff for each name letter N, S assigns T to ⍀//(Q,
F)///(X/N). (Notation: ⍀//(Q, F)///(X/N) is the expression obtained from
⍀//(Q, F) by replacing all free occurrences of the variable X by the name letter N.)
S assigns F to (⌸X)⍀//(Q, F) iff for some name letter N, S assigns F to ⍀//(Q,
F)///(X/N).
S assigns T to (⌺X)⍀//(Q, F) iff for some name letter N, S assigns T to ⍀//(Q,
F)///(X/N).
S assigns F to (⌺X)⍀//(Q, F) iff for each name letter N, S assigns F to ⍀//(Q,
F)///(X/N).
A valuation V assigns T (or F) to a sentence or nonsentential formula ⌽ iff either the
sentential element or the nonsentential element of V assigns T (or F) to ⌽.
A valuation is complete iff it assigns T or F to every closed nonsentential formula and
every sentence. Otherwise it is partial.
A sentence or closed nonsentential formula A is a logical consequence of a set of sentences or closed nonsentential formulas {B1, . . . ,Bn} iff (1) every valuation that assigns
T to each of B1, . . . ,Bn also assigns T to A, and (2) every valuation that does not assign
F to any of B1, . . . Bn also does not assign F to A.
Commentary
1. We can now state how the descriptive/nondescriptive distinction applies to
sentences generally, including sentences constructed by insertion of nonsentential
formulas (nsf ’s) into logically complex sentence-forming operators. A sentence A has
descriptive overall content just in case there is a partial valuation V such that (1) for
every closed nsf B, V assigns neither T nor F to O[B], and (2) V assigns T or F to A. The
idea is that there is some valuation that assigns a truth value to the sentence A while also
assigning no truth value to any ought-sentence.
Cognitivist Expressivism
293
2. The semantics involves a substitutional rather than objectual treatment of the
quantifiers. This is for simplicity, but one could instead formulate the semantics
model-theoretically with an objectual construal of the quantifiers, rather than the
truth-theoretic way with substituational quantifiers.
3. Truth values are assigned by a valuation not only to sentences but also to closed
nsf ’s, because the latter too have propositional descriptive content and hence can be
true or false. Likewise in English, a that-clause can be true or false.
4. As explained already, some sentences can be assigned neither T nor F by a valuation. Falsity conditions thus needed to be built into the notion of valuation, rather than
a stipulation that F is assigned to any sentence or closed nsf not assigned T.
5. We depart from familiar approaches to formal semantics, which construe a
valuation as an assignment of truth values just to the logically simplest sentences and
then define ‘truth in a valuation’ recursively. This is because we take it that not all
sentences have their truth values uniquely determined by the truth values assigned by
a valuation to logically simpler ones. For instance, suppose that for every name symbol
N, a valuation V assigns F to the sentence O[GN], where ‘Gx’ symbolizes ‘that x takes
out the garbage’. Such a valuation might yet assign either T or F to the sentence
O[(᭚x)Gx]. Even though no particular individual is such that he/she ought to take out
the garbage, perhaps nonetheless it ought to be the case that somebody takes out the
garbage; or perhaps not. Logically, both possibilities remain open.
6. The fundamental semantic principles governing the operator I[ ] are the ones saying that a valuation assigns T (or F) to I[A] iff it assigns T (or F) to A. The idea is that a
given descriptive content should be assigned the same truth value by a valuation as is
assigned to the sentence that makes an is-the-case assertion with respect to that
content. As we pointed out in addressing semantic assessability in section 6, a valuation
can serve either of two roles: first, reflecting an agent’s beliefs, and second, describing
the truth values the sentences possess when they are construed as belief-independent
assertions. Consider these two construals of a valuation, in turn. First, if an agent is iscommitted with respect to the content expressed by A, i.e. has a belief of type I[ ] with
respect to A, then a valuation reflecting the agent’s beliefs will assign T both to A itself
and to I[A]. Second, if sentences of the type I[ ] are being construed as making beliefindependent assertions (rather than belief-expressive assertions), then again a sentence
I[A] should be assigned T by V just in case V also assigns T to A itself. If a certain
descriptive content is true under a valuation, then so is the sentence that is-asserts that
content; and conversely.
7. The fundamental semantic principles governing the operator O[ ] are the ones
saying that V assigns T to O[A] only if for every B that is a nonsentential consequence
of A, V assigns T to O[B]; and that V assigns F to O[A] only if for every B such that A
is a nonsentential consequence of B, V assigns F to O[B]. It is because these are if/then
constraints, rather than biconditionals, that the truth values assigned by a valuation
to statements of the type O[A] are not uniquely determined by the truth values
assigned to logically simpler statements. For instance, if for every N, V assigns F to
O[FN], then V may assign either T or F to O[(᭚x)Fx]. (See the example in comment
5 above.)
294
Terry Horgan and Mark Timmons
8. The fundamental semantic principles governing the operator O[ ] are quite weak.
One could consider strengthening them in various ways, even while retaining their
if/then form. But we doubt that there is adequate theoretical motivation for doing so.
Also, a general reason to avoid stronger principles is the need to avoid various wellknown deontic paradoxes. Consider, for instance, this candidate semantic principle: If
V assigns T to both O[A] and O[(A ⊃ B)], then V assigns T to O[B]. Building this closure principle into the formal semantics would allow the generation of the ‘contrary to
duty imperative paradox.’ Suppose, for example, that Andy steals the money, that he
ought to be punished for doing so, but that he is otherwise undeserving of punishment.
Letting ‘Sa’ and ‘Pa’ respectively symbolize ‘that Andy steals the money’ and ‘that Andy
is punished’, these four claims are all true: O[Pa], O[⬃Sa], O[(⬃Sa ⊃⬃Pa)], (O[⬃Pa]
→¬(O[Pa]). Given the closure principle under consideration, these claims would
jointly entail the contradiction (O[Pa] ¬(O[Pa]).
9. Iteration of the operator symbols I and O and is not permitted in this formal language, as we have specified its syntax. Allowing iteration of I would be entirely pointless, as far as we can tell. If desired, however, one could permit iteration of O by
modifying the syntax to say that if ⍀ is an sff operator then so is O⍀. (One might also
modify the formal semantics too, to lay down certain constraints on how a valuation’s
assignment of T or F to a sentence of type O⍀ must be related to the truth values the
valuation assigns to various other sentences. Alternatively, one might not build in any
such constraints; perhaps logic alone—or anyway, nonmodal logic alone—does not
impose any formal constraints upon the iteration of ‘ought’.) Although it is not obvious that moral thought and moral discourse really need iterated ‘ought’ constructions,
a prima facie case can be made for this claim. Suppose, for instance, that Andy has
stolen the money, and let ‘Sa’ and ‘Pa’ respectively symbolize ‘that Andy steals the
money’ and ‘that Andy is punished’. Arguably, although O[Pa] is true, OO[⬃Pa] is
also true (because O[⬃Sa & ⬃Pa] is true).
10. Suppose that a sentence is constructed (by insertion of nonsentential formulas,
plus perhaps insertion of variables and prefixing of quantifiers) from a logically complex
sentential-formula forming operator, rather than being constructed directly from
either of the logically simple sff ’s I[ ] or O[ ]. The assertoric force of such a sentence is
borne by the entire logically complex sff operator, and not by its proper constituents.
Likewise, for a belief expressed by such a sentence, the belief ’s affirmatory force is
borne by the entire logically complex commitment-type of which the belief is an
instance. For example, letting ‘Cd’ and ‘Ah’ respectively symbolize ‘that Dick is cheerful’ and ‘that Harry apologizes’, the assertoric force of the disjunctive sentence (I[Cd]
o O[Ah]) is borne by the entire operator (I[ ] o O[ ]) from which this sentence is
constructed, and not by either of this operator’s proper constituents I[ ] or O[ ]. Likewise,
the affirmatory force of the belief that either Dick is cheerful or Harry ought to apologize
is borne by the entire logically complex commitment-type of which this belief is an
instance—a commitment-type expressed by the complex operator (I[ ] o O[ ]).
11. In the definition of a valuation, the clauses governing logically complex
sentential formulas work semantically exactly the same way they would work if these
Cognitivist Expressivism
295
sentential formulas were constructed syntactically in the more familiar manner—
namely, directly from logically simpler sentential formulas via connectives and quantifiers (rather than by insertion of nsf ’s into the slots of logically complex operators). But
it bears reiteration that the purpose of our own syntactic construction rules is this: the
syntactic structure of the sentences thereby reflects the ontological structure (on our
account) of the relevant beliefs expressible by these sentences. The key point to appreciate is that formal semantics does not mandate construing logically complex sentences
as being built up syntactically in ways familiar from standard symbolic logic; on the
contrary, the logical syntax can perfectly well be as we have specified.
12. The formal semantics is naturally construed as yielding two complementary
classes of logical relations, applicable to two distinct domains of relata. First is the
domain of propositions and nonsentential formulas. A proposition can be construed as a
set of ways the world might be, and is expressible linguistically by a closed nonsentential
formula. (Such a formula expresses the proposition without asserting it.) Propositions,
and the closed nonsentential formulas that are their linguistic vehicles, bear logical
relations to one another characterizable in terms of the notion of a nonsentential
valuation—notably, the relation of nonsentential consequence. Second is the domain
of beliefs, assertions, and sentences—with sentences being the linguistic vehicles for
expressing beliefs and for making assertions. Beliefs, assertions, and sentences bear
logical relations to one another (notably the relation of sentential consequence)
characterizable in terms of the notion of valuation. Not all beliefs and assertions are
ones whose overall content is propositional content; this is reflected in the fact that
the notion of valuation is broader than that of nonsentential valuation.
13. Consider beliefs of logically complex type—i.e. beliefs other than those
belonging to the two logically simple commitment-types, is-commitments and oughtcommitments. For any belief B of logically complex type: if B is descriptive (and thus
is expressible by a descriptive sentence —cf. comment 1), then B is logically equivalent
to some belief B* that is a logically simple is-commitment (and thus is expressible by a
descriptive sentence constructed by inserting some closed nonsentential formula into
a single occurrence of the operator ‘I[ ]’). Take, for instance, the belief that either it is
the case that Dick is cheerful or it is the case that Harry apologizes, expressible symbolically
via the descriptive sentence (I[Cd] o I[Ah]). This belief is logically equivalent to the
belief that it is the case that either Dick is cheerful or Harry apologizes, expressible symbolically via the descriptive sentence I[(Cd v Ah)]. The former belief is a commitment
of logically complex type (I[ ] o I[ ]) with respect to the pair of descriptive contents that
Dick is cheerful and that Harry apologizes, whereas the latter belief is a logically simple
is-commitment with respect to the logically complex descriptive content that either
Dick is cheerful or Harry apologizes.
14. For any belief B of logically complex type: if B is not descriptive (i.e. is
not expressible by a descriptive sentence), then B is not logically equivalent to any iscommitment. Take, for example, the belief that either it is the case that Dick is cheerful
or it ought to be the case that Harry apologizes, expressible symbolically via the sentence (I[Cd] v O[Ah]). This belief, which is a commitment of logically complex type
296
Terry Horgan and Mark Timmons
(I[ ] v O[ ]) with respect to the pair of descriptive contents expressed respectively by the
nonsentential closed formulas ‘Cd’ and ‘Ah’, is not logically equivalent to any iscommitment whatever.
15. The asymmetry revealed in points 13 and 14 is to be expected, given our irrealism
about the ontology of morals: an is-commitment (or an ought-commitment) can only
obtain with respect to a way the world might be, and according to our moral irrealism,
only descriptive that-clauses express genuine ways the world might be. (All closed
nonsentential formulas of the formal language correspond to descriptive that-clauses of
natural language.)
16. The formal language includes belief-attributing predicates. Once again, we have
specified the syntax so that these constructions too reflect, in their syntactic structure,
the ontological structure of the beliefs they attribute. For each kind of commitmentstate that is a type of belief, including the logically complex ones, there is a predicateforming operator whose structure mirrors that belief-type. A belief-attributing
predicate expressing a specific belief is constructed by inserting, into the operator’s
bracketed slots, nsf’s expressing the respective descriptive contents that are the descriptivecontent relata of the given belief. For instance, Tom’s belief that either Dick is cheerful
or Harry ought to apologize involves a commitment-type expressed by the predicateforming operator Bel[(I[ ] O[ ])]. Letting ‘Cd’ and ‘Ah’ respectively symbolize
‘that Dick is cheerful’ and ‘that Harry apologizes’, a formal predicate expressing Tom’s
belief-property is obtained by inserting into the operator’s respective bracketed slots
the closed sentential formulas expressing the two propositional-content relata of the
belief, thus: Bel[(I[Cd] O[Ah])]. So, letting ‘t’ refer to Tom, the logically complex
nonsentential formula that Tom believes that either Dick is cheerful or Harry ought to
apologize is symbolized this way: Bel[(I[Cd] O[Ah])]t. Likewise, Tom’s belief that
that there is a specific fraternity member who ought to take out the garbage involves a
commitment-type expressed by the predicate-forming operator Bel[(⌺ )(I[ ] O[ ])].
A formal predicate expressing the belief-property is obtained by inserting a variable
into the operator’s quantificational slot and inserting into its bracketed slots open
sentential formulas expressing the two non-propositional descriptive-content relata of
the belief: Bel[(⌺x)(I[Fx] O[Gx])]. Thus, the nonsentential formula that Tom
believes that there is a specific fraternity member who ought to take out garbage becomes:
Bel[(⌺x)(I[Fx] O[Gx])]t.
17. The point of the belief-attributing predicates employing the variable-binding operator Bel( )[ ] is to accommodate ‘de re’ belief constructions, in addition to de dicto ones.
So, letting ‘r’ and ‘o’ refer to Ralph and Ortcutt respectively and letting ‘S’ symbolize ‘is a
spy’, the de dicto belief predicate ‘believes that Ortcutt is a spy’ is symbolized as
Bel[I[So]], and the de dicto nonsentential formula that Ralph believes that Ortcutt is a
spy is symbolized as Bel[I[So]]r—whereas the de re belief predicate ‘believes of Ortcutt
that he is a spy’ is symbolized as Bel(x)[I[Sx]]o, and the de re nonsentential formula
that Ralph believes of Ortcutt that he is a spy is symbolized as Bel(x)[I[Sx]]o, r.
18. Our treatment of the logical syntax of de dicto and de re belief predicates is
adapted from a proposal of Quine (1960, 1970), further discussed and developed in
Cognitivist Expressivism
297
Horgan (1989). Quine, though, construed his approach as a way of dispensing with
so-called ‘propositional objects’ of belief, whereas here we are instead construing beliefs
as relations linking a believer to one or more descriptive contents—namely, the contents expressed by the nonsentential-formulas that get inserted into the bracketed slots
of the complex predicate-forming operator expressing a particular belief-type. Thus,
Ralph’s de dicto belief is a relation between Ralph and the propositional descriptive
content expressed by ‘that Ortcutt is a spy’; and Ralph’s de re belief is a relation linking
both Ortcutt and Ralph to the nonpropositional descriptive content expressed by ‘that
x is a spy’.
19. Our account of morality-involving beliefs—both logically simple oughtcommitments and logically complex commitments like those of the logical type
expressed by the predicate-forming operator Bel[(I[ ] o O[ ])]—is extendable in a fairly
natural way to other kinds of morality-involving psychological states too. Consider, for
instance, the state entertaining that either Dick is cheerful or Harry ought to apologize.
This can be construed as ‘hypothetically trying on’ the state of believing that either
Dick is cheerful or Harry ought to apologize. Ontologically, it is a logically complex
attitude type with respect the pair of core descriptive contents that Dick is cheerful and
that Harry apologizes. And this attitude type can be expressed within the formal
language in an ontologically perspicuous way by introducing the predicate-formingoperator forming operators Ent[ ] and Ent( )[ ], analogous to the belief operators Bel[ ]
and Bel( )[ ]. Or consider the state suspecting that either Dick is cheerful or Harry ought
to apologize. Ontologically, this too is a state of ‘hypothetically trying on’ the corresponding belief, but with an accompanying disposition to embrace that belief. On the
general picture we are here advocating, logically complex psychological attitudes of
entertaining, suspecting, doubting, fearing, etc. all are ways of ‘hypothetically trying
on’ a logically complex belief—and hence all can be accommodated by fairly natural
extensions of the ontological/syntactic/semantic treatment we recommend for
logically complex, morality-involving, beliefs.
References
Audi, R. (2004). The Good in the Right: A Theory of Intuition and Intrinsic Value.
Princeton: Princeton University Press.
Ayer, A. J. (1936). Language, Truth and Logic. 2nd edn. New York: Dover.
Blackburn, S. (1984). Spreading the Word. Oxford: Oxford University Press.
—— (1996). ‘Securing the Nots’. In W. Sinnott-Armstrong and M. Timmons (eds.),
Moral Knowledge? New Readings in Moral Epistemology. New York: Oxford
University Press.
—— (1998). Ruling Passions. Oxford: Oxford University Press.
Carnap, R. (1935). Philosophy and Logical Syntax. London: Kegan Paul, Trench,
Trubner & Co.
298
Terry Horgan and Mark Timmons
Dewey, J. (1922). Human Nature and Conduct. Carbondale: Southern Illinois
University Press.
Dreyfus, H., and Dreyfus, S. (1990). ‘What is Morality: A Phenomenological
Account of the Development of Ethical Expertise’. In D. Rassmussen (ed.),
Universalism vs. Communitarianism. Cambridge, Mass.: MIT Press.
Gibbard, Allan (1990). Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard
University Press.
Hare, R. M. (1952). The Language of Morals. Oxford: Oxford University Press.
Harman, G. (1977). The Nature of Morality. Oxford: Oxford University Press.
Horgan, T. (1989). ‘Attitudinatives’. Linguistics and Philosophy, 12: 133–65.
—— (1995). ‘Critical Study: Crispin Wright, Truth and Objectivity’. Nous, 29: 127–38.
—— (1996). ‘The Perils of Epistemic Reductionism’. Philosophy and
Phenomenological Research, 56: 891–7.
—— (2001). ‘Contextual Semantics and Metaphysical Realism: Truth as Indirect
Correspondence’. In M. Lynch (ed.), The Nature of Truth: Classic and Contemporary
Perspectives. Cambridge, Mass.: MIT Press.
—— and Timmons, M. (2000a). ‘Nondescriptivist Cognitivism: Framework for a
New Metaethic’. Philosophical Papers, 29: 121–53.
—— —— (2000b). ‘Copping Out on Moral Twin Earth’. Synthese, 124: 139–52.
—— —— (2002). ‘Conceptual Relativity and Metaphysical Realism’. Philosophical
Issues, 12: 74–96.
—— —— (2005). ‘Morality without Moral Facts’. In J. Dreier (ed.), Contemporary
Debates in Moral Theory. Oxford: Blackwell Publishers.
—— —— (2006). ‘Expressivism, Yes! Relativism, No!’ In R. Shafer-Landau (ed.),
Oxford Studies in Metaethics. Oxford: Oxford University Press.
Hume, D. (1739). A Treatise of Human Nature. Oxford: Oxford University Press.
Mandelbaum, M. (1955). The Phenomenology of Moral Experience. Glenco, Ill.: The
Free Press.
Moore, G. E. (1903). Principia Ethica. Cambridge: Cambridge University Press.
Quine, W. V. (1960). Word and Object. Cambridge, Mass.: MIT Press.
—— (1970). Philosophy of Logic. Englewood Cliffs, NJ: Prentice-Hall.
Shafer-Landau, R. (2003). Moral Realism: A Defense. Oxford: Oxford University Press.
Stevenson, C. L. (1963). Facts and Values. New Haven: Yale University Press.
Timmons, M. (1999). Morality without Foundations: A Defense of Ethical
Contextualism. New York: Oxford University Press.
Varela, F. (1992). Ethical Know-How. Stanford, Calif.: Stanford University Press.
Warnock, M. (1960). Ethics since 1900. London: Oxford University Press.
Wright, D. (1992). Truth and Objectivity. Cambridge, Mass.: Harvard University Press.
13
Truth and the Expressing in Expressivism
Stephen Barker
0. Introduction
In the beginning expressivism—otherwise known as emotivism—was the
thesis that utterances of value-sentences are mere expressions of desire and that
consequently they are not genuine assertions that manifest genuine beliefs or
judgements, or are open to truth evaluation.¹ This first wave of expressivism
was a highly revisionist doctrine, since value-statements really appear to have
all these features. Latter-day, or new wave, expressivism renounces this
revisionism. Its essence is distilled in the following claim:
In asserting a value-sentence, then at least with respect to the value-content
of that utterance, the speaker is not reporting any fact—natural or nonnatural—but is rather expressing a motivational state of some kind. Thus
value-sentences, at least with respect to their value-content, are not made
true by reality.²
In short, the essence of new wave expressivism is rejection of the principle that
all truth-apt sentences have truth-makers—parts of reality that make them
true. Somehow, motivational states are part of the content of value-sentence
utterances, but the latter are not reports about the possession of such states by
speakers.
Clearly some discourse is reportive and subject to truth-making. For
example, true logically simple sentences, like Snowy is black, are, it seems, made
true by reality; a dog’s being black. So within the domain of truth accessible
discourse there is, if new wave expressivism is right, a division between
¹ A paradigm first-wave expressivist is Ayer (1946). Blackburn (1984) is a more recent proponent. See
Hale (2002) and van Roojen (1996), for criticisms of Blackburn’s theory. Gibbard (1990) portrays himself
as an expressivist, but it is not clear at all that he is. See Koelbel (2002).
² I make the proviso about value-content because expressivism is consistent with the position that valuesentences have dual contents; reportive non-value-content and expressive value-content. See Hare (1952).
300
Stephen Barker
reportive and expressive assertion. But what is expressing in the expressivist’s
required sense?³
I think there is a broad set of ideas that the reporting/expressing distinction
might be understood as the distinction between representational and nonrepresentational discourse. A programme that sets out to realize this idea is the
Minimalist Programme or MP:
MP (Minimalist Programme)
A: Truth-aptness and truth are to be understood minimally. Truth-apt
discourse is discourse meeting certain syntactic and use-conditions. The
essence of our concept of truth is given by our commitment to affirm all
instance of the T-schema: ‘S’ is true iff S (or it is given by this commitment and avowal of certain further minimal norms).
B: Some areas of discourse meeting the above conditions for truth-aptness
involve contents that, because of additional facts about assertoric
practice, function as representations. Other areas meeting the same
conditions for truth-aptness involve contents that do not function as
representations.
C: Both representational and non-representational truth-apt discourse is
equally truth-apt. Because truth is defined minimally and not linked
essentially to representation in any substantial sense, the truth-maker
principle is not generally valid. Value-sentences are instances of this kind
of discourse: truth-apt but lacking truth-makers.
Price (1990), Wright (1992), Blackburn (1998), Horgan and Timmons (2000)
are examples of theorists carrying out MP in different ways. Wright does not
equate his treatment of value-sentences with expressivism, but that is, in part,
because he takes expressivism to be first wave, revisionist expressivism.
For Horgan and Timmons the emphasis is on beliefs and judgements either
possessing or not possessing representational content, and only secondarily on
sentences possessing such content.
Does MP represent the right form for second wave expressivism? I don’t
think so. One doubt is that MP buys into minimalism about truth and the
T-schema. I argue in Barker (2003) that the T-schema is not valid. Natural
languages contain conventional implicature operators like, ‘even’, ‘but’,
‘nevertheless’, ‘therefore’, and ‘if ’. These operators are non-truth-conditional.
³ The word ‘expressing’ has a technical use in this paper, that suited for the value-expressivist’s needs. I will
not be considering expressing in the sense that expletives or utterances of thanks express emotion or assertions express beliefs.
Truth and the Expressing in Expressivism
301
Languages containing conventional implicatures are ones where the truthpredicate lacks the disquotational property. For example, take the two
sentences:
(1) ‘Even Mother Teresa was pious’ is true.
(2) Even Mother Teresa was pious.
If minimalism about truth is right, then (1) and (2) have the same linguisticrule-based commitments. However, the rule-based commitment of (2) and (1)
differ: (2)’s is to (2)’s truth and its implicature content being satisfied; (1)’s is to
(2)’s truth only. Now, implicature-operators like ‘even’ embed in logical compounds and semantically interact with sentential operators. Thus ‘Even Mother
Teresa was pious’ is true if and only if even Mother Teresa was pious is a well-formed
instance of the T-schema. But the formal, rule-based ground for this instance
of the T-schema is not one we accept since its right-hand and left-hand sides
differ with respect to rule-based commitments. If so, not all instances of the
T-schema are correct, and consequently, minimalism about truth must be
rejected.
I don’t expect people to take this argument seriously, partly because its conclusion disputes a principle long held as a part of philosophical lore. Also its
appeal to linguistic facts is non-standard. So I will not press it here. I have
another objection to MP. The second objection is that the division between
the representational and non-representational is flawed since discourse is
essentially representational. In uttering sentences we are just in the business
of representing how things are, either with the world, or with ourselves. The
representational/non-representational distinction then cannot ground the
reporting/expressing distinction. I cannot defend this strong claim here, but I
can argue for a weaker one that is just as damaging: two utterances can be
equally representational but one is reportive and the other expressive. Consider
the sentences:
(3) I believe strongly that it is raining.
(4) It is probably raining.
(3) involves a report about the speaker U herself possessing a credence state,
namely [C(It’s raining) ⫽ high]. We judge (3) true if and only if we think U has
this state. (4) is not a report of a credence state but an expression of one.⁴ We
judge (4) true if and only if we think it probable that it is raining, that is, if and
only if we have the credence state. However, utterances of (3) and (4) by U are
⁴ See Barker (2004) for some detailed argument that (4) cannot be construed as a report about an objective probability or a subjective state.
302
Stephen Barker
representational in the same way. As uttered by U, both sentences involve U
representing her possession of the credence state. Both (3) and (4) are signs of
that state. In each case, audiences take away the information that U has [C(It’s
raining) is high]. (3) and (4) have the same assertability conditions for the
speaker. There is a rule-based regularity that it is permissible for U to utter
either when she believes that she has [C(It’s raining) is high]. In short, they have
the same representational content. But they differ in truth-conditions. One
reports a state and is true if and only if U has [C(It’s raining) is high]. The other
does not.
I think a similar line of argument can be applied to value-sentences, assuming
expressivism about value. Analogous to (3) and (4), we shall find sentence
pairs, the first of which—the analogue of (3)—is used to report possession of a
motivational state ⌬, and has the form ‘I have ⌬’, and the second—the analogue of (4)—is a sentence V used to express possession of ⌬. Both sentences
will have the representational content that U has ⌬. But one is reportive and
the other expressive.
To conclude, expressive discourse is not non-representational discourse; MP
can’t be right. I don’t expect the reader to be convinced by this argument.
Nevertheless, let us suppose it is correct. Where would the project of expressivism
about value go from here if it was? Let us suppose that discourse is essentially
representational. It might seem that making that admission would be fatal to
expressivism, because it would mean that something like the semantic conception
of truth or SCT would be correct:
SCT: Truth-bearers are sentences with representational content.
If SCT were correct, truth would be correct representation, and the truth-maker
principle would generally hold, blocking expressivism. In fact, the admission
that discourse is representational does not commit us to SCT at all. We can
deny that truth is essentially about representation by proposing that it is not
merely representation; truth-bearers are sentences with representational
content used in a certain way. More precisely, my thought is the pragmatic
concept of truth or PCT:
PCT: The truth-bearers are sentences with representational content that
are also used with an assertoric purpose.
On this conception, truth-bearers are assertions, and assertions are sentences
carrying representational content used with a certain assertoric communicative
purpose. That purpose, as will be revealed, is one of defending a commitment
property: a cognitive or conative state that the speaker is in. Assertions are exercises
in defending commitment properties. Assertions come in varieties: reportive
Truth and the Expressing in Expressivism
303
and expressive. The distinction between reportive and expressive assertions is
not that between representational and non-representational discourse—as
MP—would suggest, but between sentences with representational content
used with one assertoric communicative purpose as opposed to another.
In this conception of truth-bearers, truth-claims are not reports about representation of or correspondence to mind-independent reality, or reports about
coherence, peer agreement, or possession of the disquotational property, but
expressions of a kind of intersubjective agreement about the instantiation of the
commitment properties associated with assertions. Reportive and expressive
assertions are both capable of genuine truth, in precisely the same sense.
Nevertheless, true reportive assertions correspond to, and are made true by,
reality whereas this doesn’t hold for true expressive assertions.
In sections 1 to 3 below, I sketch the basic idea. In section 4, I lay out the
conception of normativity and objectivity. In section 5, I explain the case of
value-sentences and provide thereby a new framework for expressivism about
value.
1. Preliminaries: assertion
We are supposing that all discourse is representational. In giving George an
order to jump, I represent how things are—I desire George to jump. In asserting
that I want George to jump I represent how things are—I desire George to
jump. But why is my second utterance truth-apt but not the first? Because the
second involves assertion? But what is assertion? Assertion cannot simply be
uttering sentences intending to represent how things are, for the same goes on
in orders—this is just repeating the point about representation.
Appeals to truth and belief won’t help us identify assertions either.
Assertions cannot be analysed simply as acts of uttering sentences with intentions to utter truths, because we need to know what a truth-apt sentence is. We
cannot analyse assertions as acts in which speakers express their beliefs, since
speakers express beliefs in other illocutionary acts.⁵
What then is the peculiar activity that we call assertion? Assertion is a kind
of stance taking (Alston 2000 and Horgan and Timmons 2000). I interpret this
⁵ Brandom (1983) contends that U asserts that S iff (i) U undertakes to justify S, if asked to, and (ii) permits
speakers to use S as a premise in arguments. But one can make a case that orders meet both conditions. Re
(i), audiences can challenge the correctness of an order. Re (ii), one can certainly make inferences based on
orders, e.g. inferring that Fred is sick from the order to cure him. Of course, one might object that we cannot write up Cure Fred! So he is sick, as an argument. But why not, given that we can make inferences based
on orders? I think we provide an answer to this question through a prior account of assertion.
304
Stephen Barker
idea in the following way. To make a stance is to defend turf. The turf is a
cognitive or conative state of some kind possessed by the speaker. Assertion
involves (potentially) intersubjective engagement: acceptance, dispute, dialogue with another about what is defended. In asserting that Snowy is black, I
defend a belief, but we could also say, I defend a commitment to representing
the world as being a certain way, namely, that Snowy the dog is black. My audience will engage with me, and agree with my assertion, if they too defend the
same kind of thing: they defend a commitment to representing things thus and
so. An audience H judges my assertion correct/true if and only if she is willing
to defend that kind of turf. In short, assertion is to defend a commitment, and
judgement of truth is to express in some way that that commitment is accepted.
How does this help us distinguish orders from assertions? In orders and
assertions speakers represent how things are. The difference is in purpose. In
assertions, speakers defend a commitment to a cognitive/conative state,
whereas in orders they are not defending anything. Assertions are democratic;
they are a form of stance-taking dialogue. Orders are autocratic. In an order, a
speaker represents how things are—that they desire an audience to do such and
such—where the speaker’s purpose is that the audience recognize this desire.
That’s it. There is no invitation to dispute defence of commitment to that
desire. The speaker is not defending anything. In contrast, assertions are about
defence. The truth-predicate is used to express acceptance of what the speaker
defends. That’s why orders are not truth-apt! This is despite the fact that in
both assertion and ordering, discourse is representational.
Communicative purposes then enable us to isolate truth-apt discourse
within the sphere of representational discourse. It also allows us to distinguish
the reportive from the expressive within the sphere of the truth-apt, that is,
within the sphere of assertions. Consider (3) and (4). In uttering (3), the
speaker U defends a commitment to representing the world as one in which she
has [C(It’s raining) is high]. An audience H judges (3) true if and only if she, H,
accepts the commitment. But what of (4)? In uttering (4), U is not defending
a representational intention. She is defending something else. We can see what
by looking at how a potential audience H truth-accesses U’s utterance. H does
not judge (4) true if and only if she thinks U has the state [C(It’s raining) is
high], but if and only if she, H, thinks she, H, has the state. If so we can
conclude that what U defends in uttering (4) is acceptance of [C(It’s raining)
is high].
The difference between reportive and expressive assertions is then this: in
reportive assertions speakers defend commitments to representational intentions; in expressive assertions speakers defend commitments to states—such as
credence states—whose possession they have in fact represented. This accords
Truth and the Expressing in Expressivism
305
with the schema PCT outlined above. In illocutionary acts, be they orders,
reportive or expressive assertions, speakers represent how things are. What distinguishes the illocutionary acts from each other is purpose. Truth is pragmatic
in that it is not a semantic property like representation; the truth-bearers are
assertions: representational acts used with a specific kind of purpose.
2. The Pragmatic conception of truth (PCT)
To get this account up and running we need more details about assertion
and semantic content. First, assertion and some finessing. I have spoken of illocutionary acts in terms of representing and intentions to represent. But the
reality is more subtle. We really need to talk of speakers advertising intentions
to represent. By advertising intentions I mean that U intentionally engages in
the behaviour characteristic of a person who, following certain rules, has the
intentions. So in uttering ‘Snowy is black’ U advertises an intention to represent how things are, that a fact of a certain form obtains. Advertising is the
common element in sincere, insincere, and fictional assertion. If U sincerely
asserts that Snowy is black, U really has the fact-representing intention she advertises. If U is lying, then she lacks the intention she advertises, but intends her
audience to believe she has it. And if U is engaging in fictional make-believe, she
lacks the intention she advertises, and intends her audience to recognize this fact.
The first component of an assertion then involves uttering a sentence and
advertising a fact-representing intention. But assertion involves more than
representation. It also involves a communicative purpose of defending a
commitment to a property. This again involves advertising. In asserting a
speaker advertises an intention to defend a certain cognitive or conative
property—call it ⌸—and in fact really has that intention. So an assertion has
the form:
Assertion:
(i) U utters S and advertises intentions:
(a) to represent a fact of the form ⬍P⬎; and
(b) to defend commitment to a property ⌸.
(ii) U possesses the intention (ib).
So in reporting that Snowy is black, U utters a sentence, say ‘Snowy is black’,
advertises an intention (a) to represent a fact ⬍snowy, black⬎ and (b) to
defend a commitment to intending to represent the obtaining of that fact.
That U has this communicative purpose is indicated, conventionally, in the
case of simple sentences like ‘Snowy is black’, by the indicative mood of the
sentence.
306
Stephen Barker
Whether the assertion is a reportive or an expressive assertion depends on
what ⌸ is. The assertion about Snowy is a case where ⌸ is a representational
intention. But if the assertion is expressive, ⌸ may be a cognitive or conative
property like having a belief or a subjective probability state, or an intending or
desiring something. For example, we can now give the official analysis of the
assertions of (3) and (4), given in section 1, which are, respectively, a report and
expression of a mental state:
P1:
(i) U utters ‘I believe strongly that it is raining’ advertising intentions:
(a) to represent a fact of the form ⬍U has [C(It’s raining) is
high]⬎;
(b) to defend commitment to the representational intention
in (a).
(ii) U has the intention in (ib).
P2:
(i) U utters ‘It is probably raining’ advertising intentions:
(a) to represent a fact of the form ⬍U has [C(It’s raining) is
high]⬎;
(b) to defend commitment to [C(It’s raining) is high].
(ii) U has the intention in (ib).
In both P1 and P2, U produces a sentence with the same representational
content—U performs the act (a) of uttering a sentence advertising an intention to represent a fact concerning acceptance of a credence state. What
differs in P1 and P2 are the advertised communicative intentions. In P1,
it is the property of defending the representational intention displayed, the
(a)-intention. In P2, it is the defence of the commitment to the probability
state represented.
That difference in communicative intention means that (3) and (4) have different truth-conditions for an audience. In uttering (3), U performs P1,
defending a representational-intention. H accepts U’s utterance, judges it true,
just in case H accepts commitment to that representational intention in her
own case, i.e. just in case H believes U has the probability state. In uttering (4),
U performs P2, defending commitment to a subjective probability state. H
accepts U’s utterance, judges it true, just in case H accepts the property U
defends, that is, the probability state: [C(It’s raining) is high]. In short, the different truth-conditions of (3) and (4), that differing applicability conditions of
the truth-predicate, are determined by the distinct communicative intentions
associated with them.
Truth and the Expressing in Expressivism
307
We are now in a position to answer the question: what is a truth-bearer? Just
what truth-bearers amount to is elicited by considering a dialogue (5),
(5) Smith: It’s probably raining.
Jones: True. It’s probably raining.
Smith and Jones are in agreement about a truth. But what is the object of which
they both predicate truth?⁶ According to PCT, the truth-bearer is an assertiontype defined by a commitment property. That type is P2, where ‘U’ is treated
as a variable. Both Smith and Jones are in agreement about instantiating the
property, [C(It’s raining) is high]. But this also translates into agreement about
a truth. Both are disposed to perform instances of P2, which is an assertiontype. In short, the thought that both contemplate, and of which they predicate
truth, is defined by this assertion-type and this commitment property.
Truth-bearers are identified not by propositions—states of affairs—but by
commitment properties.
How does the truth-predicate function here? Utterance of ‘S is true’ is not a
report that one agrees with one’s interlocutors; that would be to introduce a
kind of relativism. Nor is it a report that S has a certain property such as correspondence with reality, being believed-by-such-and-such-group, cohering
with some system of beliefs, or even the disquotational property.⁷ Rather,
the truth-predicate is used to express commitment to accepting certain
commitment properties ⌸ pertaining to an assertion. In uttering ‘S is true’, U
expressively asserts her commitment to accepting the ⌸-property pertaining to
assertion of ‘S’. That is:
True: (i) U utters ‘S is true’ advertising intentions:
(a) to represent a fact of the form ⬍U accepts the commitment property of assertion of S⬎;
(b) to defend commitment to the state in (a), that is, accepting the commitment property of assertion of S.
(ii) U has the intention (ib).
According to this theory, true sentences are not necessarily made true by reality. They will be, if the assertion judged true is reportive, but not if the judged
assertion is expressive. In both cases, the truth-predicate functions in the same
way: as the expressive predicate described here.
⁶ According to SCT, the semantic concept of truth described in section 0, the truth-bearer is a sentence
with representational content. No truth-bearer in this sense can be found for (5), since there is no state of
affairs in the world that both speakers are representing. See Barker (2004: ch. 1.2).
⁷ Since, as already suggested, the truth-predicate lacks the disquotational property. See Barker (2003) for
the argument.
308
Stephen Barker
3. Objectivity, defence, and naturalism
The key component of this account is defending, but I have said little about it.
I now offer an account of it. U defends a commitment to ⌸ just in case U
dialectically engages with her audience. This occurs when:
(i) H recognizes that U has ⌸; and
(ii) H either accepts, confirms, or disputes a commitment to ⌸ in her own
case.
In (ii), dispute means either rejection or doubt. So in assertion U intends to
get H to believe that she accepts ⌸ and for H to respond by confirmation,
acceptance, doubt, or rejection.⁸
It might be thought that commitment, acceptance, and rejection are normative states. They are internally connected to having reasons. So expressivism
about value or subjective probability statements rests on a prior normativity:
that of reason. Even if this were so, it would not undermine the project
of expressivism about moral and aesthetic value, and expressivism about
probability statements or, say, if-sentences. But in fact there is no intrinsic
normativity built into these states.
In Barker (2004), I give naturalistic treatments of acceptance, rejection, and
commitment. Underpinning the capacity to form speech-acts is a natural
system of representation with simple cognitive and conative states. That system
comprises logically simple representational states, laws regulating relations
between such states, dispositions and propensities to token representations,
constraints on not tokening them, etc. The system itself does not contain
thoughts of the kind that are realized through use of natural language sentences, since the system is not intersubjective. Thoughts, as we know them, are
communicative: they feature communicative intentions. This pre-linguistic
system is the precondition for having such intentions.
In terms of this system, being committed to ⌸ is just a cognitive state, to
the effect that given normal functioning of the system it would produce
instantiation of ⌸ as one of its states. Rejection is an attitude in which U is
not disposed to instantiate ⌸ where, in addition, U is constrained, by internal
cognitive conditions, to cease all epistemic search procedures for justifying
that commitment. A state of doubt is one that is neither acceptance nor
rejection.
⁸ Perhaps the claim that in assertion U intends to get H to believe U accepts ⌸ is too strong. A better
approximation might be that U hopes that H will believe that she accepts ⌸. Can one assert something without any interest at all that any audience responds? I suggest not. If a depressed Robinson Crusoe soliloquizes,
indifferent to the attention of anyone else, he still hopes to get things right by his own lights. He monitors
his own speech. The producer hopes to satisfy the monitor, both being, in this case, within the same speaker.
Truth and the Expressing in Expressivism
309
In saying these states are natural I am not proposing a reduction of cognitive
norms to natural facts about agents. No such reduction is possible. Rather I am
saying there are natural facts about cognitive systems to which we have normative
attitudes. The normative attitude is simply the fact that we are prepared to defend
the cognitive regularities that in fact we instantiate. In asserting that Snowy is
black, U displays herself as committed to representing a certain fact. Her commitment, which is simply a dispositional state, is open to scrutiny, in the sense that the
cognitive laws that produce it can in turn be defended. Hence we find a kind of
boot strapping; U defends acceptance of ⌸ through her cognitive and inferential
dispositions. That defence has normative force because of U’s disposition to
defend these cognitive states and inferential dispositions in turn, and so on.
Another way of putting this is in terms of reason. Acceptance, rejection, etc.
are intimately connected to possession of reasons; one accepts ⌸ because one
has a reason, say, accepting ⌸⬘. In the expressive framework proposed here, the
normativity of reason is itself expressively analysed. That is, assertions of
sentences R, below,
R: Accepting ⌸ is a reason for accepting ⌸⬘
are assertions where we express commitment to certain cognitive regularities of
the form: instantiating ⌸⬘ is a causal ground for instantiating ⌸. So, assertions
of R have the structure:
R:
(i) U utters R and advertises intentions:
(a) to represent a fact of the form ⬍For U, instantiating ⌸⬘ is a
causal ground for instantiating ⌸⬎; and
(b) to defend commitment to a property in (a).
(ii) U possesses the intention (ib).
An audience H will judge an assertion of R true if and only if she herself
experiences the cognitive causal law as something compelling in her case; she
instantiates the state. Thus an assertion of R is not a claim that certain causal
law holds for certain cognitive systems—that is a crude and untenable naturalistic reduction—but is an expressive assertion in which such a state is defended.
In short, the intersubjective engagement of defence presupposes that agents
embody certain natural regularities in their cognitive and conative systems.
These regularities must be reasons—they must have normative rightness—but
their having such status just resides in the fact that normative assent is always
possible with respect to them. This means that we are disposed to defend
the cognitive/conative regularities through intersubjective engagement. Of
course, we cannot defend every regularity of our system in one go.
The PCT-account uses the intersubjectivity of defence to underpin normativity and objective truth. But it might seem that it cannot be right. What of the
310
Stephen Barker
lone thinker, Robinson Crusoe, who does not engage with any one? Crusoe
makes assertions to himself that are objectively true. But Crusoe is not engaged
intersubjectively with others. Of course, we can allow that Crusoe monitors his
own speech. However, in describing Crusoe, we, as authors of a certain fiction,
take up a stance towards our fictional creation. If we think that Crusoe has said
something true it is because we express a commitment to the ⌸-property he
defends.
Another objection to the PCT-account is that it over-intellectualizes
assertion. In assertion U intends, tries to bring it about, that H recognizes her
possession of the state, and either accepts or disputes it. It might seem that this
is a rather complex intention, and a higher-level intention at that. In advertising
an intention to defend ⌸, U intends to engage in the behaviour characteristic
of a speaker who intends that H recognize that U accepts ⌸. But, this, I suggest,
is ultimately just causal complexity. Speakers do not need explicit theories
themselves about what they are doing any more than they need a theory of
grammar. So there is no over-intellectualization.
4. Embedding and belief
According to PCT, truth-bearers are assertion-types or tokens. This entails
that unasserted sentences—though fully contentful—don’t have truth-values.
But is that plausible?⁹ There are three reasons to think not. (a)In locutions of
the form ‘It is true that S’ we assert truth of something, but one might claim it
is unasserted. (b)The sentences embedded in logical compounds like negations
and disjunctions have truth-values, but are unasserted. (c) Generally what is
believed is open to truth-evaluation, but it’s not an assertion.
I deal with these objections in turn. Clearly assertions don’t embed. What
embeds is something prior to assertion. In uttering S embedded, U performs
component (i) of the illocutionary act of assertion: the act of advertising intentions to represent and defend. That U so advertises is a conventional matter. U
advertises these intentions, but of course lacks them. I call this act of advertising
intentions proto-assertion. It is proto-assertions that embed in constructions
like ‘It is true that . . .’ and logical compounds. Although falling short of a
full-fledged assertion, proto-assertions, nevertheless, determinately indicate
assertions. So with respect to objection (a), my reply is that in ‘It is true that S ’
U performs a proto-assertion with S, but ‘that S’ denotes an assertion-type
⁹ One piece of evidence that non-asserted declarative sentences are not truth-apt is the fact that declarative sentences used to perform orders are not truth-apt. See Barker (2004: ch. 1.2).
Truth and the Expressing in Expressivism
311
corresponding to this proto-assertion. We predicate truth of this assertion-type.
With respect to objection (b), I say that in a compound ‘It is not the case that
S’ U performs a proto-assertion with S and the whole is judged true just in case
the corresponding assertion-type is judged false. And so on.
According to PCT, truth-bearers are assertion-types and assertions are
either reportive or expressive. Are then logical compounds reportive or expressive assertions? I propose treating them as expressive assertions. Thus in uttering ‘It is not the case that S’ U expresses a commitment to rejection of the
⌸-property of A(S ), the assertion of S. That is,
Neg:
(i) U utters ‘It is not that the case that S’ U performs a proto-assertion
of S, and in addition advertises intentions:
(a) to represent a fact of the form ⬍U rejects ⌸ of A(S )⬎;
and
(b) to defend commitment to a property: rejection of ⌸ of
A(S).
(ii) U possesses the intention (ib).
Disjunction can likewise be characterized. In uttering ‘either S1 or S2’, U
expresses her rejection of commitment to rejecting both disjuncts, S1 and S2.
As expressive assertions, logically complex assertion are evaluated with respect
to truth or falsity just as other expressive assertions are. An audience H employs the
truth- or falsity-predicates to express her acceptance, or rejection, of the commitment states U expresses. Such sentences lack special truth-makers. So for example,
negations are not made true by negative facts, disjunctions by disjunctive facts, etc.
The third objection (c) given above is that beliefs can be truth-apt but are not
assertion-types. In reply, I emphasize that beliefs and judgements are not entities that lie behind assertions, as wholly distinct from them. Judgement just is
assertion. Thus what I call assertions may be private affairs involving purely
internal episodes—sentences in the head or word qualia—associated with
intentions. I reject the view of assertions as public and relatively superficial
manifestations of underlying states that are the real cognitive action.
Secondly, I note that there are belief-attributions that are best explained as
cases where the object of belief is an assertion-type. Take attribution of probabilistic beliefs—as in ‘Jones believes that it is probably raining.’ Jones is related
to a content. That content, I suggest, is simply the expressive assertion-type,
given in P2. There is no other candidate.
It is not my claim here, however, that belief-objects are always and essentially
assertion-types. As already noted, underpinning an ability to perform assertions—
verbal judgements—is a natural cognitive/conative representational system.
Some of the states of this system are logically simple belief-like representational
Stephen Barker
312
states. I call these doxastic tracking states. Perhaps non-verbal creatures
unequipped to perform illocutionary acts have such states as what are essentially belief-states with non-linguistic objects. They have doxastic tracking
states. However, doxastic tracking states are not truth-apt. Truth is not the
property of correctly representing reality.
What is believing when its object is an assertion-type? Is it a disposition to
sincerely assert? An objection to this is that people can sincerely assert that S
even though all other evidence indicates that they believe not-S. There are two
ways to go here. First, we can propose that the speaker has contradictory beliefs.
She believes that S, based on verbal production, and consciously, and she
believes that not-S unconsciously, based on behaviour. An alternative is that we
put stronger conditions on belief: it is not merely being disposed to perform an
assertion sincerely, it is also required that the underpinning doxastic grounds in
fact commit U to assertion.
5. Value
That completes my sketch of PCT and a theory of expressive discourse that is
an alternative to MP.¹⁰ We now turn to value-sentences. My hypothesis is that
the concept of expressing articulated in the last few sections is precisely
the notion of expression required for the value expressivist. Consider oughtstatements like (6) on a deontic reading:
(6) Bill ought to confess.
If we were objectivists about value, we might treat utterance of (6) as a report
about a putative special fact of value. Expressivism rejects this analysis; utterance
of (6) is an expression of attitude. We can now give a linguistic model of what
that is. Utterance of (6) is, I submit, an expressive assertion with a structure
analogous to P2 above, the structure of a subjective probability statement. The
only difference is that the state that U is expressing is a desire state rather than an
epistemic state. So the assertion-type corresponding to (6) is something like:
VO:
(i) U utters ‘T ought to ⌽’ advertising intentions:
(a) to represent a fact of the form ⬍U approves of T ’s
⌽-ing⬎; and
(b) to defend commitment to approval of T’s ⌽-ing.
(ii) U has the intention in (ib).
¹⁰ See Barker 2004 for a detailed treatment of PCT, and especially the issue of compositionality.
Truth and the Expressing in Expressivism
313
Evidently, the kind of approval that U is defending is of the special ethical kind.
It is an approval based on the judgement that certain kinds of properties are
instantiated, ones we identify as ethically relevant. An expressivist treatment of
value needs to provide some kind of account of what sorts of properties these
are—that would be a theory of what distinguishes moral attitudes from other
kinds. I won’t attempt to provide any such account here.
On the current view, ‘T ought to ⌽’ has the content of ‘It is desirable that T
⌽s’ or ‘It is preferable that T ⌽s,’ where these sentences are used in expressive
assertions in exactly the same sense. It is quite consistent with this theory that
U asserts, ‘I do not want to do ⌽, but I ought to do ⌽.’ U simply has conflicting desires about ⌽-ing. ⌽-ing has properties which make it desirable but also
undesirable, where those properties that make it desirable are properties that
stimulate moral attitudes.
In the case of ethical musts, such as ‘Fred must not kill Jane,’ it seems that the
attitude expressed is stronger than desire or preference. Perhaps the state
defended in ‘Fred must ⌽’ is something more like:
(7) [x would be shocked, horrified, filled with deep disapproval, etc., if
Fred were not to ⌽]
Ethical ‘may’ as in ‘Fred may/is permitted to ⌽’, are cases in which U expresses
her rejection of any commitment to desiring that Fred not ⌽ or her rejection of
a state like (7). States of desire can be rejected since there are rational connections between belief and desire sets and desires. (More of that in a moment.)
Truth and objectivity
The PCT-analysis applied to value-sentences enables us to explain how valuesentences, assuming value-expressivism, are truth-apt, but have no truth-makers.
Assertion of value-sentences is guided by subjective states, but they are not
reports about those states or any facts in the world that might have given rise to
them. Consider the dialogue (8):
(8) Smith: Bill ought to confess.
Jones: True. Bill ought to confess.
Let us assume VO above. Smith and Jones in (8) defend approval states about
Bill’s confessing. They are not merely reporting their own states; their utterances are not true by virtue of the obtaining of those states of approval—there
is no subjectivist relativism here. In fact, their utterances are not true by virtue
of anything. The common truth-bearer about whose truth Jones and Smith
agree upon is the assertion-type whose commitment property is possession of
314
Stephen Barker
the state of desiring that Bill confesses. Both Smith and Jones accept this
⌸-property. Jones’s judgement of truth is his expression of acceptance of
that ⌸-property. Our truth judgement as third parties is expression of acceptance of the ⌸-property; or if we judge the assertion false, our falsity judgement
is an expression of rejection of that ⌸-property.
I submit that PCT-account, applied to value-statements, can explain the
objectivity that attends ought-statements. That objectivity resides in the facts
that: (a) what ought to be, or may be, etc., is not a matter of desire; (b) the truth
of value-sentences is something that we can debate; (c) where there is disagreement we can talk of error. These three aspects of objectivity are captured by PCT
as follows. According to the PCT-account, subjects’ possessing desires doesn’t
make ought/must/may-statements true. Rather the truth of value-statements is
established through defence of motivational ⌸-properties, that is, through,
potentially, rational dispute. Participants in such intersubjective engagement are
often prepared to stand by their commitments. It is not an arbitrary matter what
subjects desire or prefer. Given U disapproves of lying, U, as a result, disapproves
of Fred’s lying. The former is a reason for the latter. The normativity of reasons,
as already noted in section 3, resides in our expressed commitment to acceptance/instantiation of causal laws in our cognitve/conative systems. So, in the
present case, U’s cognitive/conative system instantiates the following causal law:
(9) (⌸⬘) Disapproval of lying causes (⌸) disapproval of Fred’s lying.
So U’s defence of disapproval of Fred’s action is grounded by her instantiation
of (9) and the state of disapproving of lying. The normative force of this
defence derives from the fact that U is prepared to defend these states in turn.
That is, the PCT-analysis does not attempt to reduce normativity to causal
regularity, but rather to the intersubjective engagement of defending cognitive/
conative states that may involve causal regularities.
This account presupposes that motivational states can be defended, that is
accepted, and rejected, etc. For this to be possible in PCT’s terms, all that is
required is transitional laws like (9) and states, like disapproval of lying, that
can be instantiated or not, or which, in the case of rejection, are states that U’s
system is disposed not to instantiate, given certain constraints. These are essentially the same basic conditions that underpin defence of properties in the case
of non-value-assertions. PCT countenances no difference as such between the
truth of logical compounds, such as ‘There are no hippos in this room’, and
moral truths such as ‘Hippos ought not to be treated cruelly’. Both have equal
status to be called real, objective truths.
Note the objectivity of moral truth does not require convergence. It is not
built into the possibility of objective moral truth that there is agreement with a
Truth and the Expressing in Expressivism
315
community of moral judges. Moral truth can be essentially contested—there is
no guarantee that there is agreement to be had out there. What we have instead
is openness to contestation, and dialectical engagement. Value-assertions are
challenges to a potential other to debate. Because ‘is true’ is not a descriptive
predicate, with the content ‘believed by my peers’, a speaker cannot isolate herself
in her own community—immune from critique from others outside that
community. Rather the moral judge is open to challenge because built into
making any judgement at all is the act of taking a stance. In short, we don’t have
a descriptive, static conception of truth—truth as a kind of achievement of
correspondence or agreement—but a dynamic conception of truth. A judgement of moral truth is an expression of acceptance of a certain conative, or
related, property, potentially defendable by some agent. But this, as we have
seen, is essentially the same story with all truth.¹¹
Embedding, validity, and belief
The possibility of objective moral truth is one dimension of the semantics of moral
statements. A second important dimension to moral truth is embedding and
validity. The PCT-account has no problems treating embedded value-sentences.
Consider a conditional like (10):
(10) If Bill did it, he ought to confess.
Utterance of (10) expresses commitment to belief that the assertability for U of
the antecedent, given background beliefs, is a reason for accepting the assertability for U of the consequent. We include as a background belief, say, that
people who do wrong ought to confess, and the denotation of ‘it’ in (10) was a
wrong.
The validity of arguments like (11) below is perfectly explicable:
(11) If Bill did it, he ought to confess. Bill did it. Therefore, he ought to
confess.
Validity falls out straightforwardly in terms of truth—assuming that if-sentences
are truth-apt. We can say that an argument is valid if and only if any valuation
V of its premises that renders them true, will render its conclusion true. By
valuation I mean a set of judgements of truth and falsity of the logically simple
¹¹ How do we think of isolated moral agents like Robinson Crusoe? As we did for the agent isolated in
general. Moral judges don’t need audiences as such. To say something true is not constituted by being in
agreement with anyone else. (It is not constituted by anything, in fact.) In imaging Crusoe we assess the truth
of his moral utterances, which is to say, we express commitment to the motivational ⌸-properties he defends
or not.
Stephen Barker
316
sentences of the language concerned, with logically complex sentence receiving
values as a consequence, given their meanings, of this assignment.
In section 4, I argued that assertion-types are the principal objects of belief.
In the moral case, the object of belief is an expressive assertion-type. In the case
in which U believes that Bill ought to confess, the assertion-type has the structure of VO. U’s moral belief-state is a commitment to this assertion-type, one
in which U defends a motivation state: approving of Bill’s confessing. The
moral belief-state is then, effectively, a disposition to defend a motivational
state. It is not the motivational state itself.
‘Good’ and thick ethical terms
Sentences with modal auxiliaries such as ‘ought’, ‘must’, and ‘may’, etc. are not
the only kinds of ethical value-sentences. There are non-modal value-sentences
such as ‘T is good’ or ‘T is right’. One might attempt to treat these as corresponding to assertion-types of exactly the same kind as ought-sentences. In
other words, ‘T is good’ is like ‘S is true’; they are used in expressive assertions.
What, however, of thick ethical terms, such as courageous or generous?
In Barker (2000), I propose an implicature theory of value that is applied
to thick evaluative assertions. According to it, value-sentences such as ‘T is
courageous’ have a said-content, which is a factual assertion about T possessing
a certain property G; and an implicature that U is committed to approving of
things that have G. Such value-assertions have the same structure as utterances
of ‘Even T is G’, which feature the implicature operator ‘even’. This sentence
has a said-content that T is G and an implicature about the relative unlikeliness
of T being G. So, ‘T is courageous’ has the form:
C: If U asserts the sentence ‘T is courageous’ then U intends to denote
by ‘courageous’ a property F: being such as not to flee in situations of
adversity, etc, and:
(i) U reportively asserts that T is F;
(ii) U implicates that U is committed to approval of F-things.
This theory is expressivist because value-content is non-reported content,
albeit the value-assertion as a whole has a reportive, non-value component.
Implicature is a form of pragmatic presupposition. So in asserting that T is
courageous, U communicates that approval of F-things is to be taken for
granted by her audience. This means that H judges U’s utterance correct if and
only if she, H, judges that T has F and she, H, approves of F-things.¹²
¹² In fact, in Barker (2000) I supply a similar dual-content theory for ‘good’.
Truth and the Expressing in Expressivism
317
To fully develop this account we need to explain what implicature is as form
of non-truth-conditional content, embedding, and validity—I do this in
Barker (2003, 2004) using the PCT-framework.
Summary
In this paper I have presented an alternative framework for understanding new
wave value expressivism, which fully embraces the truth-aptness of valuesentences. I argued that the Minimalist approach, MP, which attempts to
analyse expressing in terms of non-representational discourse, is not tenable.
According to the alternative I defend, PCT—the pragmatic conception of
truth—truth-bearers are assertions, which come in two kinds, reportive and
expressive. This theory can be supplemented with an implicature account,
which applies to thick ethical terms.
References
Alston, W. (2000). Illocutionary Acts & Sentence Meaning. Ithaca, NY: Cornell.
Ayer, A. J. (1946). Language, Truth and Logic. 2nd edn. London: Victor Gollancz.
Barker, S. (2000). ‘Is Value Content a Component of Conventional Implicature?’
Analysis, 60/3: 268–79.
—— (2003). ‘Truth and Conventional Implicature’. Mind, 112: 1–31.
—— (2004). Renewing Meaning. Oxford: Clarendon Press.
Blackburn, S. (1998). Ruling Passions. Oxford: Clarendon Press.
Brandom, R. (1983). ‘Asserting’, Nous, 17: 637–50.
Gibbard, A. (1990). Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard
University Press.
Hale, B. (2002). ‘Can arboreal knotwork help Blackburn out of Frege’s abyss?’
Philosophy and Phenomenological Research, 65: 144–9.
Hare, R. (1952). The Language of Morals. Oxford: Oxford University Press.
Horgan, T., and Timmons, M. (2000). ‘Nondescriptivist Cognitivism: Framework
for a New Metaethic’, Philosophical Papers, 29: 121–53.
Koelbel, Max, (2002). Truth without Objectivity. London: Routledge.
Price, H. (1990). ‘Why “not”?’ Mind, 99: 221–38.
van Roojen, M. (1996). ‘Expressivism and Irrationality’. Philosophical Review, 105:
311–35.
Wright, C. (1998). Truth and Objectivity. Cambridge, Mass.: Harvard University Press.
This page intentionally left blank
14
Normative Properties
Allan Gibbard
When the young G. E. Moore chose the title of his first book, Principia Ethica,
roughly a hundred years ago, he was of course emulating Newton. And like
anyone who hopes to develop a systematic treatment of ethics, he first needed
to explain what questions he would be addressing. On this he found himself to
be in a minority of two, along with his old teacher Henry Sidgwick, whose
personality and lectures Moore had found dull back in his student days a
decade earlier.¹ Sidgwick was the only other writer who recognized that the
basic concept of ethics is indefinable, so that no full explanation of ethical questions in terms drawn entirely from outside ethics can be given.² ‘What definition
can we give of “ought”, “right”, and other terms expressing the same fundamental notion?’ asked Sidgwick. ‘To this I should answer that the notion which
these terms have in common is too elementary to admit of any formal
definition.’³ The century of metaethics we now celebrate, though, begins with
Moore, not with Sidgwick in 1874. That’s not because all of Moore’s arguments for the indefinability claim were sharply different from Sidgwick’s. It’s
not because Moore, any more than Sidgwick, meant the indefinability thesis to
be the centerpiece of his work; for both men it was a preliminary. But Moore
was vivid where Sidgwick was dull, and he not only attacked specific alternatives
but gave arguments intended to drive the thesis home in general.
Moore stressed the distinction in ethics between questions of meaning and
questions of substance, and thereby gave rise to a tradition in analytic philosophy
of separating the two parts of ethical theory: the metatheory and the substantive, normative part. Some philosophers have rejected the distinction; some
Kantians, for instance, think that if you get the metatheory right, substantive
ethical conclusions fall out as some kind of consequence, so that metaethics
and substantive ethics are not really separate. Then too, anyone who rejects
Sidgwick’s and Moore’s indefinability claim and thinks that ethical terms
¹ Moore, ‘An Autobiography,’ 16.
² Moore, Principia Ethica, 17.
³ Sidgwick, Methods of Ethics, I. iii . 3, p. 32.
320
Allan Gibbard
can be given analytic, naturalistic definitions thinks that the two putative
subdivisions are not really separate. Those who reject any systematic distinction between questions of meaning and questions of substance might likewise
reject a sharp, separate subject of metaethics. Still, in the analytic tradition, the
distinction between substantive ethics and metaethics has been tenacious, and
even philosophers who doubt it keep falling back into enquiries that seem
recognizably metaethical. Highlighting a hundred years of metaethics since
Moore’s book, then, is entirely appropriate.
The subject that came to be known as metaethics fits awkwardly into philosophical subdivisions. Its central questions concern meanings and concepts,
and how to treat these in general is a question not for the peripheral subfield of
philosophy that ethics is, but for the ‘core areas’; these are questions for the
philosophy of language and philosophy of mind, along with metaphysics and
epistemology. The dual nature of metaethics suited Moore himself perfectly:
Moore had a novel ethical vision to propound, with its organic wholes in place
of hedonism and its stress on the value of taking pleasure in true beauty and
true friendship. A correct metaethics, he thought, would clear the way to a
correct ethics. He also joined Russell to revolutionize English philosophy in the
core areas. Since Moore, metaethics has been driven both by people who want
to do ethics and by philosophers of the core areas, who ask whether ethics poses
any special problems for their philosophies—for theories of concepts, properties, and the like. Ayer, for instance, found philosophical ethics a nuisance,
since it threatened to offer a counterexample to the verifiability criterion of
meaning.⁴ Stevenson worked on metaethics, he told Frankena, because he
wanted a subject that Wittgenstein wasn’t working on;⁵ he developed a view he
had found in a one-page passage of Ogden and Richards on the meaning of
meaning.⁶ He and Ayer eventually converged on a view that included elements
drawn from each of them.⁷ Those in our recognized canon of the greatest
philosophers ever, Plato, Aristotle, Hume, and Kant, all worked seriously both
on core areas and on ethics, including what we’d now call metaethics, and their
ethical and core area theories are of a piece with each other. We recently sadly
lost the great David Lewis, whose interests centered on the core areas but whose
work in ethics, metaethics, and social philosophy fills a volume.⁸ Metaethics by
now is a subject that brings together two tribes of philosophers: specialists in
⁴ Ayer, Language, Truth and Logic, ch. 6.
⁵ This is my memory of what Frankena told us.
⁶ Ogden and Richards, The Meaning of Meaning, 125. Stevenson reproduces the passage at the start of
Ethics and Language, and in ‘The Emotive Meaning of Ethical Terms,’ he calls it ‘the source of the ideas
embodied in this paper’ (23).
⁷ See Stevenson, Facts and Values, 210–14, and Ayer, Language, Truth and Logic, 2nd edn. (1946). I thank
Kevin Toh for clarifying changes in Stevenson’s thought.
⁸ Lewis, Papers in Ethics and Social Philosophy.
Normative Properties
321
ethics and specialists in the ‘core areas.’ Since most of us can’t do both with a
specialist’s degree of sophistication, this makes for tensions. The ideal in
metaethics must be to combine experience and sensitivity in both areas.
Indeed in recent decades, aspects of ethics have begun to enter the core areas.
The special features of ethical concepts that Moore stressed are not, many of us
now think, special to narrow morality, with its stress on right and wrong, the
morally admirable and the blameworthy. We do of course need to understand
the concepts of narrow morality, to wield them effectively or to criticize them
(as do Nietzsche and Bernard Williams⁹). But part of what’s special about
morality is that it operates in the ‘space of reasons;’ it concerns justification and
oughts. So, broadly speaking, does epistemology, with its focus on what we
have reason to believe. The same goes for questions about sentiments or feelings: what we have reason to admire or be ashamed of. The term ‘normative’ is
central to much current philosophical discussion. There’s no agreement on
what this technical term in our discipline is to mean, but it involves, in a phrase
drawn from Sellars, being somehow ‘fraught with ought.’ And in recent
decades, it has been claimed—even proclaimed as an established finding—that
the concept of meaning is normative. It has been held that the concept of
mental content is normative, the concept, that is, of what a person is thinking;
thus, for instance, the predicate ‘thinks that snow is white’ is, according to
some, a normative concept. So not only is metaethics, in some ways, a branch
of core area philosophy, with all the privileges and duties pertaining thereto.
Crucial parts of the core areas now become branches of metaethics, more or
less. What’s crucially puzzling in metaethics extends to the puzzling topic of
normative concepts in general. That makes for a broader subject matter, the
metatheory of normativity, covering concepts of justified belief and justified
feelings. The metatheory of meaning and of mental content, the ‘meaning of
meaning,’ will, if some philosophers are right, be a part of all this. If so, then
substantive theory of meaning and mental content are branches of a broad
subject, substantive normative theory. Ethics and metaethics expand to take in
most all of philosophy.
I won’t explore in this paper whether, in some important sense, meaning
indeed is ‘normative.’ Some philosophers argue that it is and some that it
isn’t.¹⁰ I do think that normativity in general is an intelligible topic, worthy of
⁹ Nietzsche, The Genealogy of Morals; Williams, Ethics and the Limits of Philosophy.
¹⁰ Kripke’s words on p. 58 of Wittgenstein on Rules at least suggest a view that the concept of meaning and
the concept of what a person is thinking are normative concepts. McDowell endorses such a view in
‘Wittgenstein on Following a Rule,’ 336. Brandom makes such a view central to his theory of thought and
language in Making it Explicit. I propose an interpretation of such claims in ‘Meaning and Normativity.’ For
criticisms of such claims, see Horwich, Meaning, ch. 8, and Millar, ‘Normativity of Meaning.’
322
Allan Gibbard
study, and I have a hypothesis on what it involves. In advance of establishing
what normativity consists in, we can characterize it in a theory-light way that
allows for competing theories of its nature. Suppose Moore is right that the
concepts of good, morally ought, and the like have some important feature that
the empirical concepts of science and everyday life don’t share. And suppose
later philosophers are right that this puzzling feature is shared by a wider class
of concepts that somehow involve reasons or being ‘fraught with ought.’ Then
if there is some special, problematic feature that all these concepts share, our
technical term for it is ‘normativity’. If there isn’t any such crucial, distinctive
feature, then there’s no such special topic of study as normativity. If there is
such a feature, then this theory-light characterization tells us what competing
theories of normativity are at odds over: what the special, problematic feature
is that concepts of reasons and oughts and other concepts that seem laden with
these all distinctively share.¹¹
If there is such a distinctive thing as normativity in this sense, then we’d
expect metaethical enquiries to go in two directions. One is to develop, refine,
and criticize metatheories of normativity in general. The other is to enquire
into moral and ethical concepts in particular: within the broad field of normativity, we can ask, what’s special about ethics? These two broad areas of enquiry
must speak to each other: if ethics is a special branch of normative enquiry, then
a correct theory of the nature of ethics must fit with a correct theory of the
normative in general, and ethics will provide a crucial test case for any theory
of how normativity in general works. But the two flavors of enquiry will be
quite different. Developing metanormative theory must involve, ideally, a
close collaboration of thinkers from different subdisciplinary backgrounds,
ethics and the core areas. Asking what’s special about ethics within the normative in general, on the other hand, should lead to centrally ethical concerns.
In this paper, I grapple with normative concepts in general. I myself have a
view on how normative concepts work,¹² and in this paper I consider one
crucial aspect of the view. The view draws on resources from the classic
non-cognitivists Ayer and Stevenson, but it ends up with theses that sound
highly ‘realistic’ and metaphysical. There is a property, I say, that constitutes
being what one ought to do. This property is, in a broad sense, a natural
property. Concepts are another matter: normative concepts are distinct from
¹¹ I propose this theory-light characterization in ‘Meaning and Normativity,’ 97.
¹² My new book Thinking How to Live develops the aspects of my view that this paper explores further.
My ‘Normative and Recognitional Concepts’ was largely drawn from the manuscript of that book, and contains some of its central contentions. See also my ‘Reply to Hawthorne’ and ‘Knowing What to Do, Seeing
What to Do.’ My Wise Choices presents the main aspects of the view and is broader in its concerns, but on
the issues in this paper it says far less than does Thinking How to Live.
Normative Properties
323
naturalistic concepts; on this score Moore was quite right. But normative
and naturalistic concepts signify properties of the same kinds; indeed a normative
and a naturalistic concept might signify the very same property. What’s
distinctly normative, then, is not properties but concepts. It is this set of claims
about normative properties—that they exist and are, in a broad sense,
natural—that I scrutinize here.
1. The thesis and the argument for it
I distinguish concepts and properties. The property of being water turns out
just to be the property of being H2O, consisting of molecules made up in a
certain way. The concepts, though, are different; it came as a discovery that
water is H2O, and people could intelligibly disagree on whether it is—whereas
disagreement on whether water is water would be baffling. Moore’s arguments
for non-naturalism harped on such possibilities of intelligible disagreement,
and what arguments of this kind show, as with water and H2O, is not a difference of properties but of concepts. Moore himself insisted that what he called
‘the good’ is a natural object of thought; it is the natural property that all and
only good things must always have. We could emend Moore, then, to say that
whereas the concept of being good is distinct from any naturalistic concept—
from concepts fit for empirical science and its everyday counterparts—the
property of being good is a natural property, a property for which we could have
a naturalistic concept. Moore thought that this property involves complex
organic wholes, but his chapter I arguments in Principia Ethica were not the
ones he thought established this. The doctrine of organic wholes, as we would
now say, is a substantive ethical claim, not a conclusion of his metaethical
arguments. An ethical hedonist, for instance, would be mistaken but might be
intelligible, thought Moore. On the emended Moore-like picture I am proposing, such a metaethically unconfused hedonist would think that the concepts
are different but the properties are the same. The property of being good just is
the property of being pleasant.
Any being capable of action, I claim, is committed to a view with this
structure—not that being good is being pleasant, but that some natural property
is the property of being good, whether or not we know what property it is. To
show how the argument for this works, I switch from good to ought. (Good, on
my view, is a composite concept: to be good is to be desirable, such that it ought
to be desired.) The starting point of the argument owes much to Ayer: we say
what the word ‘ought’ means not directly, but by saying what is to think or
believe that a person ought to do something. What is it, say, to think that one
324
Allan Gibbard
ought to leave a building if the fire alarm rings? To think this, I propose, is to
plan for a contingency. I think, ‘Jack ought to leave the building,’ and accepting this amounts to adopting a contingency plan: a plan to leave the building
in the contingency of being in Jack’s exact circumstances. Now this is something on which we can agree or disagree; there is such a thing as disagreement in
plan. You and I, in other words, can disagree on what to do in Jack’s plight. And
this possibility of disagreement in plan leads to all the metaphysical-sounding
things that I want to say.
What, then, of more complex normative claims? That is the crucial
‘Frege–Geach’ worry for ‘expressivistic’ theories, theories in the spirit of Ayer
and Stevenson. Take, for instance, a disjunctive claim,
Jack is engaged in some matter of life and death or he ought to leave the building.
Now if we can make sense of disagreement in plan, then we can offer a general
account of what such complex normative claims amount to. We elicit the content of this claim by mapping what agreeing or disagreeing with it amounts to.
To do this, take an onlooker who is fully decided: she has a view as to how the
world is naturalistically and has a full contingency plan for living—or at least
one full enough to cover purposes at hand. I’m decided enough to agree or disagree with this claim if I’m settled on two matters: the naturalistic matter of
whether life or death hinges on what Jack is doing and the planning matter of
whether to leave the building if faced with his plight. The onlooker disagrees
with this claim just in case she both (i) thinks that Jack isn’t engaged in a matter of life and death, and (ii) plans not to leave the building if faced with his
plight. The content of the claim, then, amounts to which decided states it disagrees with, to which possibilities it rules out.
Now for the claim of Natural Constitution, the claim I explore in this paper.
Some broadly natural property constitutes being what one ought to do—that is the
claim. Any planner whatsoever is committed to this claim, I argue. Here is the
proof, quickly sketched¹³: Imagine first an onlooker who is decided enough to
agree or disagree with any normative claim whatsoever. Fantastically, she has
what we might call a hyperplan: a full contingency plan for living, which covers
any situation a person might conceivably be in. Now a full plan for a circumstance must be couched in terms that are empirical or naturalistic. A plan, say,
simply to do whatever is best would be no plan at all, until it was supplemented
with directions for recognizing what’s best. Infinitely, then, we could put this
onlooker’s hyperplan in naturalistic terms: it runs, ‘In circumstance C1, do A1;
in circumstance C2 do A2;’ and so on. According to this onlooker, then, there is
¹³ See my Thinking How to Live (forthcoming), ch. 5.
Normative Properties
325
a property that constitutes being what one ought to do—namely this one:
being act A1 in circumstance C1, or act A2 in circumstance C2, and so on. The
terms in this infinite disjunction are all couched naturalistically. Infinite
disjunctions of natural properties are ‘natural’ in a broad sense. This onlooker,
then, accepts Natural Constitution: that there is some broadly natural property
that constitutes being what one ought to do.¹⁴
Most of us, of course, aren’t hyperdecided on what to do and how things
stand. But here is a principle that applies to the rest of us too. Suppose there’s
something I’d accept no matter how I filled out my views, so long as I didn’t
change my mind about anything. I’m undecided, imagine, whether the temperature this afternoon will be in the high 80s or low 90s Fahrenheit. Any way
of filling out my views without changing my mind then involves thinking it
will be hot. And so, we can conclude, I’m already committed to the claim that
it will be hot this afternoon. This principle applies to the claim of Natural
Constitution. Any way I might fill out my plans for living completely without
changing my mind about anything would bring me to accept Natural
Constitution. I’d think that there is a broadly natural property that constitutes
being what one ought to do, because I’d have a view on what that property is.
And so Natural Constitution is a claim to which I’m already committed.
I now make a transcendental move and voice the claim to which I am
committed: Some broadly natural property constitutes being what one ought to do.
This quick argument needs far more elucidation and exploration that I can
give it here. In this paper, though, I explore just one set of questions it raises—
questions I don’t get far enough into in my new book Thinking How to Live.
2. Too many properties?
The proof I just sketched assumes a particular kind of conception of what a
property is. Any full contingency plan for living, I supposed, amounts to the
plan always to do whatever act has a certain property. The proof goes through
if a ‘possible worlds’ conception of properties is correct, a model-theoretic
conception. A property, on this conception, is a function of a particular kind.
A world-to-extension function, let’s say, is one whose domain is the set of all
¹⁴ Moore’s own arguments concerned a broader class of properties than the strictly natural. He thought
they applied not only to natural objects of thought but to supernatural and metaphysical ones too; he
thought he had refuted the possibility of a supernatural definition of good or a metaphysical one. Perhaps we
should use a term like ‘factual property’ to cover all of these—but that wouldn’t fit a minimalist reading of
‘fact’. For more discussion of this, see my ‘Reply to Hawthorne’; on minimalism for ‘truth’ and ‘fact’, see
Horwich, Truth.
326
Allan Gibbard
possible worlds, and which assigns to each possible world a set of entities in that
world. The proof I offered works so long as a property is just a world-to-extension
function. Or alternatively, I could have made a slightly weaker assumption:
that properties correspond one-to-one with world-to-extension functions. A
full contingency plan determines, for each possible world, a set of occasions for
action along with the act the plan prescribes for that occasion. It thus assigns to
each world a set of occasion-act pairs—and thus, on the possible world’s
conception of a property, determines a property of occasion-act pairs.
Now trivially, any property is characterized by a world-to-extension function. Redness is characterized by the function that assigns, to each possible
world, the set of entities in that world that are red in that world. In general, on
this pattern, any property P, we can say, is characterized by the world-to-extension
function that assigns, to each world w, the set of entities in w that have property P in w. My argument for Natural Constitution depended, though, on a
controversial assumption that goes the other way around: that every world-toextension function characterizes some property. I supposed too that no worldto-extension function characterizes more than one property.
I help myself freely to possible worlds, and I follow Stalnaker in thinking
that such talk is innocuous. A possible world is a way things might have been—
a way that is fully determinate.¹⁵ So the world-to-extension function for ‘red’
is the function that assigns, to each determinate way things might have been,
the set of all entities that would have existed and would have been red had
things been that way.
Given my assumptions, the property that constitutes being what one ought
to do is natural in the following sense. A natural world, we can say, is a completely determinate way that things might have been so far as basic natural
properties go. Given any natural world, there is a set of acts open in that world.
(Think of an open act as consisting in an occasion for action and one of the
alternatives open to the agent on that occasion.) Then there’s a world-to-extension
function defined as follows: to each natural world, it assigns the set of all acts
open in that world which an agent ought to perform. In other words, to each
determinate way things might have been naturalistically, it assigns the set of all
acts that would, if the things had been that way, be open to someone and be
what that person ought to do. Suppose then, as I did, that any world-to-extension
function corresponds to a unique property. With natural worlds in question,
this property is broadly natural in this sense: the worlds are ways things might
have been naturalistically, and the identity of the property is fixed by natural
aspects of how things would have been had they been that way.
¹⁵ Stalnaker, ‘Possible Worlds.’
Normative Properties
327
Indeed Natural Constitution in no way depends on my expressivistic
starting point. The principle holds so long as what one ought to do supervenes
on the natural properties of the situation, in this sense: in two possible situations, what an agent ought to do differs only if situations differ in some natural
feature.¹⁶ The point of the proof is that expressivism mimics a normative realism
like Moore’s in this regard: from expressivism follows the supervenience of the normative on the natural, and from this supervenience follows Natural Constitution.
This view of properties as world-to-extension functions, though, is probably
a minority view among metaphysicians who think about such matters. In a way,
the view may be too indiscriminate: it allows the wildest infinite disjunction of
properties to count as a property. In another way, it may be too restrictive, for
it doesn’t allow that there could be necessarily equivalent properties that are
distinct, such as being triangular and being trilateral.
I need to enquire, then, what happens if we question the assumption that
properties are world-to-extension functions—or at least correspond to them
one to one.
3. Plans that coincide with properties
The assumptions on which my argument depended, recall, were two: that (i)
for each world-to-extension function there is a property that it characterizes,
and (ii) no world-to-extension function characterizes more than one property.
This second assumption is, in other words, that necessarily equivalent properties are identical. Properties are necessarily equivalent just in case, no matter
how things had been, they would have been coextensional, that is, possessed by
the same set of things. I’ll explore the issue as follows: Take first any contingency plan that does coincide with a property, in the sense that some property
N satisfies this condition:
for any possible occasion for action and any act open on that occasion, the
plan prescribes that act for that occasion just in case the act has property N.
I’ll urge—though with misgivings—that it’s then idle to think anything but
that being what one ought to do consists in having property N. I’ll then turn to
assumption (i) and argue that we should be liberal with our attributions
of properties, allowing every world-to-extension function to characterize a
¹⁶ Of the many species of ‘supervenience’ that have been distinguished in the literature, this is roughly
‘strong supervenience’ in Kim’s sense, Supervenience and Mind, 149–56; he adds a vague dependence
requirement that seems to be met.
328
Allan Gibbard
property. My exploration, at the end of the paper, will arrive at an issue that has
been central to recent ethical theory and can be formulated independently of
the metaphysical questions I am asking about properties.
Suppose for now that assumption (i) fails, in that not every hyperplan coincides with a property. Consider, though, any hyperplan that does so coincide;
it directs all and only those acts that have a certain property N. Someone who
adopts this hyperplan, I’ll suggest, should accept that being what one ought to
do consists in having property N. She should consider the property of being
what one ought to do just to be property N.
Without loss of generality, we could take some particular hyperplan that
coincides with a property and ask what lessons generalize. Consider the hyperplan of a universal hedonist and the property of maximizing net pleasure in the
universe. Shouldn’t she claim that the property of being what one ought to do
just is the property of prospectively maximizing net happiness in the universe—
of being unihedonic, as I’ll say? Look again to our familiar paradigm of property
identity, water and H2O. The property of being water is the property of being
H2O—why say this? Natural properties, we can say, are properties open to
empirical investigation; they seem to earn their keep in empirically based
causal explanations.¹⁷ Being H2O is a matter of chemical structure, with two
hydrogen atoms bonded to an oxygen atom. Having this structure turns out
to explain the manifest phenomena we more familiarly explain wielding the
concept water. Thus we say that being water turns out to consist in being
H2O—that one property is in play, and molecular theory explains its nature.
The property identity helps explain many things about water: why, for
instance, passing an electric current through water produces two gases with
distinct properties.
Take now the property of being pleasant. It figures in causal explanations; pleasure helps to explain, for instance, why we eat chocolate. It’s for reasons like this
that being pleasant counts as a natural property, if commonsense psychology has
its explanations right. Now explaining why things do happen as they do is quite
different from explaining why people ought to do some things and not others.
(That’s a truism that falls out from my account of the concept of ought:
explaining why one ought to do something is just explaining why to do it; this
kind of explanation figures in deciding to do it or not.) So if properties somehow earn their keep in explanations, won’t different classes of properties figure
in these fundamentally different kinds of explanations?
¹⁷ This distinction between concepts and properties stems from Putnam, ‘On Properties’; he contrasts
‘predicates’ or ‘concepts’ with ‘physical properties’ or ‘physical magnitudes’—although he does not mean to
confine the distinction to physics in any narrow sense. He stresses the role of properties in what I would call
causal-explanatory theories.
Normative Properties
329
A concept like being pleasant, to be sure, figures in both kinds of explanation:
People do eat chocolate because it is pleasant, and because chocolate is pleasant,
other things equal we ought to eat it. But the puzzling claim here is the fundamental one: that other things equal, we ought to do what brings pleasure. Will
explaining this require two separate kinds of properties, natural ones like being
pleasant and normative ones like being what one ought to do?
The universal hedonist explains why to do things in terms of pleasure: once
it is shown that an act maximizes net pleasure in the universe, his explanation
of why to do it comes to an end. What explanatory purpose could he then serve
by saying that two distinct properties are in play? On a parallel with the case of
water and H2O, he can instead treat this as a case of property identity. He
thereby achieves a kind of explanatory power that parallels the explanatory
power of thinking that being water and being H2O are one and the same
property. In both cases, that’s where explanation comes to an end. In both
cases, the identity can be intelligibly rejected by those who reject the consequent explanations. In the one case, the explanations are causal/empirical, and
in the other case they are explanations of why to do some things and not others.
Pleasure figures in explaining, causally, why people do things, and it figures
too, according to the hedonist, in explaining why to do things. In either case,
no explanatory purpose would be served by supposing an extra property. A
further property of being what one ought to do would add nothing to the
explanation of why to eat chocolate; the identity claim itself works as the start
of the explanation.
Anti-hedonists, to be sure, will insist that more than one property is in play.
They will deny that the property of being what one ought to do is the property
of maximizing net pleasure in the universe. The universal hedonist can then be
asked why we should accept his claim of property identity. But that parallels the
case with water and H2O: the opponent of molecular theory denies the property identity, and the proponent can be challenged on why we should accept it.
As long as a universal hedonist doesn’t change his mind, though, he can still put
his view as follows: ‘Being what one ought to do consists in maximizing net
pleasure in the universe. The properties are one and the same, and that
explains, at base, why to do the things we ought to do. The concepts are distinct, and that explains why, nevertheless, my opponent’s view is intelligible.’
I now turn to a worry I don’t know how to quell. Do the things I have said
extend to any contingency plan whatsoever that coincides with a property?
Philosophers who think that necessarily equivalent properties may be distinct cite cases like being triangular and having three sides.¹⁸ Perhaps
¹⁸ See, for instance, Sober, ‘Why Logically Equivalent Predicates May Pick out Different Properties.’
330
Allan Gibbard
explaining the properties of triangles demands believing in both these
properties and believing them distinct. Of course neither of these properties
plausibly enters into basic explanations of why to do things, but imagine
someone with a bizarre view of how to live: one ought at base, this person
thinks, to accumulate triangles. The question might then arise, for this person, what property of triangles explains, at base, why we ought to accumulate
them. Ought we to accumulate them because they have three angles or
because they have three sides? If the latter, then trilaterality, not triangularity,
is what explains why we ought to do things. Perhaps we should then say this:
It’s a property involving trilaterality, not triangularity, that constitutes being
what one ought to do.
On this particular, bizarre view of how to live, necessarily coextensional
properties are sometimes distinct, but still, there’s a natural property that being
what one ought to do consists in. Still, once we admit that a natural property
might be necessarily coextensional with oughtness without explaining
which things we ought to do, we seem to allow this possibility: there’s a natural
property that’s necessarily coextensive with oughtness, but an act’s having that
property doesn’t explain why we ought to do it—and indeed no property
explains why one ought to do the things one ought to do.
The triangularity/trilaterality example suggests that properties have structure, with complex properties somehow composed of simpler properties and
structures. How a view of properties as structured might work for properties
like being alive or being overjoyed might seem baffling—what is their
structure? Suppose, though, that such a view is nonetheless right.¹⁹ Then to
establish that there is a property of being what one ought to do, we would have
to establish something like the following: when we ask why to do a thing, there
is a general answer with a structure of the kind that characterizes properties.
The argument I have given establishes no such thing.
Perhaps, however, another kind of argument could. We have an ideal of
ultimately correct reasoning on what to do and why; what would such reasoning be like? If hedonists are right, the answer is straightforward, whereas if
some pluralists are right, good reasoning on what to do can always be further
refined, so that the ideal right answer must be unlimited in its refinement. But
this ideal right answer will have a structure—even if it is infinite. That, then,
we might try saying, will be the structure of the property of being what one
ought to do.
¹⁹ Putnam, in ‘On Properties’ (313–15), proposes a way of treating such properties as second-order,
functional properties.
Normative Properties
331
4. A property for each function?
Normative pluralists maintain the thesis, sometimes, that no single property
always explains why to do those things we ought to do. Sometimes we ought to
do something because it’s keeping a promise at a small loss in general happiness,
sometimes because it’s crucial to one’s self-development and requires breaking
only a minor promise—and so on. Perhaps some such way as this is the best
way to put the views of a pluralist like Ross or of a current particularist.²⁰
Trivially, there’s a world-to-extension function that characterizes ought. But
why think it characterizes any single property at all?
My argument for Natural Constitution assumed such a liberal conception of
properties as to make the answer trivial: any world-to-extension function whatsoever, I took it, characterizes a property. Many philosophers, though, are more
sparing with the properties they countenance. Certainly if my assumptions are
right, then most properties are wild: they are unfamiliar, and in terms of
elementary properties and relations, they can be characterized only in ways that
are infinite. They don’t figure in causal explanations. Shouldn’t we be cautious
when it comes to believing in properties? And shouldn’t we be especially careful
in calling properties ‘natural,’ even in the broadest sense?
These considerations should indeed bear some weight in our thinking. We
need to ask, though, whether any alternative conception of a natural property
could answer these worries without introducing others. Will a more restrictive
conception of natural properties allow the familiar candidates we need but still
rule out being what one ought to do as a natural property?
Fundamentally, let’s assume, the causal patterns of the world stem from
elementary entities that have elementary properties and stand in elementary
relations to each other. These are the basic natural entities, properties, and relations. (I’ll from now on drop talk of ‘relations’ and include them in my talk of
‘properties.’) These can’t, though, be the entities and properties that figure in
everyday explanations: We explain the length of a shadow by the tallness of
the tree that casts it, damage to a house by the severity of a storm, and chocolateeating by cravings and expectations of pleasure. If tallness, severity, and expecting
count as properties, they somehow come down to elementary properties of
elementary entities, but not in a way we can specify in detail. We can talk
loosely of ‘higher order’ or ‘emergent’ properties, but how is such talk to be
understood? How do properties like these relate to elementary things and
properties and somehow ‘emerge’ from them?
²⁰ See Ross, The Right and the Good; Hooker and Little, Moral Particularism; Dancy, Ethics without
Principles (2004).
332
Allan Gibbard
Candidate answers come in at least two possible flavors. One is the flavor I
have been offering: the systematic, model-theoretic one. This offers us a
plethora of non-elementary properties, though most of them are of little or no
explanatory importance. The other is a thinner flavor: according to these views,
‘wild disjunctions’ of properties, when they explain nothing, are not properties
at all.
Now my puzzle with the latter kind of view is that many properties of great
explanatory importance could be presented as wild disjunctions: sex and
parenthood among sexually reproducing organisms, for instance, and even
some of the properties that physics studies—being a solid body or a coupled
harmonic oscillator. Coupled harmonic oscillators may be electronic or
mechanical; solid bodies can be composed of many different materials bonded
in a variety of ways. Or consider being yellow (in the sense in which dandelion
petals are yellow): from the point of view of pure physics, the property of being
yellow is wildly gerrymandered; it consists of a miscellaneous assortment of
reflectances a surface can have. But when human beings enter the picture,
being yellow can help explain much: why, say, the child hands you the second
block from the left when you say, ‘Give me the yellow one.’ The property exists
independent of us: there were yellow flowers before there were people and
would have been if our species had never evolved. What depends on our species
is the explanatory importance of color, not color itself.²¹
Being a property presents itself as an all-or-nothing matter, whereas explanatory
importance is a matter of degrees. If what makes for being a natural property is
helping in causal explanations of things that happen, then since helping comes
in degrees, there will be degrees of qualifying as a natural property. Why not
say instead that there are degrees of naturalistic importance? The contrast
shouldn’t be between properties and non-properties but between degrees of
importance in explaining events in the world.
Think again of color. Beings on a planet of Alpha Centauri have, perhaps, a
visual system as discriminating as ours, but one that differentiates classes of
reflectances that are quite unlike colors such as yellow. Their responses aren’t
explained by colors, then, but by a different set of properties that reflectances
can have. Like colors with us, those properties existed before the centaurs (as I’ll
call them), and like colors, they would be of zero explanatory importance had
the centaurs never evolved. Moreover, an indefinite number of such classifications must have equal claim to fix properties—properties that, in most cases,
are of no explanatory importance at all because no species sensitive to that
²¹ This view of color is sketched in my ‘Visible Properties of Human Interest Only.’
Normative Properties
333
classification has ever evolved. We seem forced to allow many more properties
than figure in explaining anything.
I conclude, tentatively, that we should accept a plethora of properties in the
universe, all but a small proportion of which have little or no explanatory
importance, either causal or normative. The possible worlds, model-theoretic
view of properties lets us do this.
5. Causal-explanatory importance
There remains a question that can be couched independently of the metaphysical
issues involved in the claim of Natural Constitution: Do normatively significant properties figure significantly in causal explanations? Or in deference to
those who reject the view of properties that I appealed to in deriving Natural
Constitution, here is the question put more cautiously: Take any important
normative concept, such as Moore’s concept of an intrinsically good state of
affairs. Is there a causally significant property, we can ask, such that being good
is necessarily equivalent to having that property? Is there, in other words, a
causally significant property N such that no matter what the natural world
were like, all and only good happenings would have property N?
Hedonists think that there is such a property, the property of being pleasant
on balance. They think that no matter how things went in the universe, all and
only states of affairs with more pleasure than displeasure in them would be
intrinsically good on balance. Now pleasure figures in causal explanations, and
being pleasant counts as a natural property if anything with a close bearing on
ethics does. That leaves open the question of quasi-metaphysics, as we might
call it, that I have been discussing: whether if hedonists are right on how to live,
then there is such a thing as the property of being good and it is the property of
being pleasant. But whatever the answer is to this, still, if universal hedonists
are right, we can truly say this: There is a natural property, a property of causalexplanatory importance, the having of which is necessarily equivalent to being
good.
Moore himself rejected substantive hedonism. Goodness involves organic
wholes, he thought; that a pleasure stems, say, from contemplating something
ugly can vitiate the claim that a state of affairs is good. Most current ethical theorists likewise reject ethical hedonism, and many think that ethical judgment
requires a refined sensibility that stymies codification. All this raises the
possibility that nothing equivalent to goodness plays a significant causal role in
our lives. There is of course a world-to-extension function that characterizes
being good, and this function is ‘natural’ at least in the sense of being infinitely
334
Allan Gibbard
constructible from clearly natural materials. But whether or not it characterizes
a property and does so uniquely—the issues I have been addressing—perhaps
it fails to characterize anything of causal-explanatory significance. Good happenings do, of course, have their effects, just as acts performed by red-haired
girls on Wednesday have their effects. Perhaps, though, being good as such has
nothing to do with explaining these effects.
If we think this to be so, we’ll be skeptics of a kind; we’ll be skeptical of the
importance of good and evil, justice and injustice themselves, in the causal
explanations of history and social science.²² Can we accept such a picture of
how the world works causally, though, and not become normative skeptics? I
paint normative skepticism as a state we’d be hard pressed to achieve: so long as
we come to views on what to do, we are committed to there being normative
truths and falsehoods, and so long as we place some faith in our own plans, we
ascribe to ourselves some capacity to know these truths, or some of them—
fallibly and imperfectly though it be. Still, isn’t coming to know a causal process?
And so if we are to know that a state of affairs is good, mustn’t goodness
somehow figure in the causal story of how we come to believe as we do?
Whether we plausibly count as responsive to goodness in our ethical judgments may be in part an empirical question. It turns out that we are less responsive to what’s better than we might have imagined: ethical judgments of
alternative policies in dealing with an epidemic, for instance, may respond to
whether the results are described in terms of lives lost or lives saved—even
when the two descriptions are logically equivalent.²³ Our first responses to
these cases, then, can’t all be veridical: we don’t reliably respond to a policy’s
being right with a judgment that it’s right. Once we note this systematic discrepancy in our judgments, though, we can revise our plans and work them
toward consistency. True, so long as we know our plans to be inconsistent, we
can’t reasonably think ourselves fully accurate judges of how to respond to such
dilemmas. But as we work to resolve our plans’ inconsistencies, we can think
ourselves increasingly responsive to goodness. In the limit as we refine our
judgments, we can think, the goodness of a state of affairs causes us to think it
good.
These reflections suggest a more sophisticated argument that goodness,
rightness, being the thing to do and the like go to causally explain things that
happen. Planning itself, we can say, supposes that there is such a possible state as
being a competent judge of what to do. This state might include, among other
²² Sturgeon hypothesizes such a significant causal role for moral properties in ‘Moral Explanations’;
Joshua Cohen argues for the plausibility of such explanations in ‘The Arc of the Moral Universe.’
²³ Tversky and Kahneman, ‘Framing of Decisions and the Psychology of Choice,’ 453.
Normative Properties
335
things, correcting for systematic biases and inconsistencies in unreflective
human judgments. Imagine, then, we did achieve ideal competence as normative
judges. Then goodness would have causal efficacy: we’d be responsive to it.
Goodness (or something necessarily equivalent to goodness) would then figure
in the causal explanation of our responses.
Does this line of reasoning show that goodness—or something equivalent—
is a property with causal-explanatory significance? I have laid out a mere hint
of an argument, which leaves many issues to be sorted through. One flaw,
though, looks damning: We are asking whether being good and the like figure significantly in causal explanations of how the world works. We have now
said that it would so figure if there were ideally competent judges of goodness.
We are committed to the possibility of ideally competent judges of goodness,
let’s agree, just by being planners. That, however, doesn’t establish that there
in fact are any ideally competent judges. It doesn’t, then, establish that
goodness enters into causal explanations of anything that really happens in the
world.
Take, after all, any world-to-extension function f, no matter how irrelevant
it is to anything of significance in explaining our judgments and social
phenomena. There is a possible state of finding something good, when one
contemplates a way the world might be, if and only if it is in the extension that
the function f assigns to that world. We might try arguing, then, that the
function f characterizes a causally significant property because it would if there
were an f-responsive judge in the world. But such an argument would clearly be
fallacious. That a property would be causally significant if things were different
fails to establish that the property is causally significant in actuality, as things
are. And this fallacious argument takes exactly the same form as the argument
I proposed for the causal significance of goodness.
The state of being an ideally competent normative judge is, of course, of
great normative significance: such a person would be the one to consult and
defer to in planning how to lead one’s life. And planning coherently must be
hostage to regarding oneself as at least on the right track as a normative judge—
for if I thought that I wasn’t in the least a competent judge of what to do, how
could I coherently follow through on the judgments I make? Perhaps this
requirement on planning commits us, somehow, to constraints on what being
the thing to do consists in. Perhaps it shows that the property can’t be too wild
and infinitely arbitrary; perhaps it even shows that the property must be
causally significant in human affairs. I haven’t found an argument that it does,
but I regard the issue as unsettled. This question arises from the way of analyzing normative concepts that I am proposing, and finding an answer to it would
be an important advance in normative theory.
336
Allan Gibbard
References
Ayer, A. J., Language, Truth and Logic. London: Victor Gollancz, 1936; 2nd edn. 1946.
Brandom, Robert, Making it Explicit. Cambridge, Mass.: Harvard University Press,
1994.
Cohen, Joshua, ‘The Arc of the Moral Universe.’ Philosophy and Public Affairs, 26/2
(1997): 91–134.
Dancy, Jonathan, Ethics without Principles. Oxford: Clarendon Press, 2004.
Gibbard, Allan, ‘Knowing What to Do, Seeing What to Do,’ in Philip Stratton-Lake
(ed.), Ethical Intuitionism: Re-evaluations. Oxford: Clarendon Press, 2002.
—— ‘Meaning and Normativity.’ In Philosophical Issues, 5: Truth and Rationality, ed.
Enrique Villanueva. Atascadereo, Calif.: Ridgeview Publishing Co., 1994: 95–115.
—— ‘Normative and Recognitional Concepts.’ Philosophy and Phenomenological
Research, 64/1 (2002): 151–67.
—— ‘Reply to Hawthorne.’ Philosophy and Phenomenological Research, 64/1 (2002):
179–83.
—— Thinking How to Live. Cambridge, Mass.: Harvard University Press, forthcoming.
—— ‘Visible Properties of Human Interest Only.’ In Philosophical Issues, 7: Perception, ed.
Enrique Villanueva. Atascadero, Calif.: Ridgeview Publishing Co., 1996: 199–208.
—— Wise Choices, Apt Feelings. Cambridge, Mass.: Harvard University Press, 1990.
Hooker, Brad, and Little, Margaret (eds.), Moral Particularism. Oxford: Oxford
University Press, 2000.
Horwich, Paul, Meaning. Oxford: Clarendon Press, 1998.
—— Truth. Oxford: Oxford University Press. 2nd edn. 1998.
Kim, Jaegwon, Supervenience and Mind: Selected Philosophical Essays. Cambridge:
Cambridge University Press, 1993.
Kripke, Saul, Wittgenstein on Rules and Private Language. Cambridge, Mass.: Harvard
University Press, 1982.
Lewis, David K., Papers in Ethics and Social Philosophy. Cambridge: Cambridge
University Press, 2000.
McDowell, John, ‘Wittgenstein on Following a Rule.’ Synthese, 58 (1984): 325–63.
Millar, Alan, ‘The Normativity of Meaning.’ In Anthony O’Hear (ed.), Logic,
Thought and Language. Cambridge: Cambridge University Press, 2002: 57–73.
Moore, G. E., Principia Ethica. Cambridge: Cambridge University Press, 1903.
—— ‘An Autobiography.’ In Paul Arthur Schilpp (ed.), The Philosophy of G. E. Moore.
LaSalle, Ill.: Open Court, 1942.
Nietzsche, Friedrich, Zur Genealogie der Moralen, trans. as On the Genealogy of
Morals, Walter Kaufmann and R. J. Hollingdale. New York: Vintage Books, 1967.
Ogden, C. K., and Richards, I. A, The Meaning of Meaning. New York: Harcourt
Brace, 1923.
Normative Properties
337
Putnam, Hilary, ‘On Properties.’ Mathematics, Matter and Method: Philosophical
Papers, vol. i. Cambridge: Cambridge University Press, 1975: 305–22.
Ross, W. D., The Right and the Good. Oxford: Clarendon Press, 1930.
Schilpp, Paul Arthur (ed.), The Philosophy of G. E. Moore. La Salle, Ill.: Open Court,
1942.
Sidgwick, Henry, The Methods of Ethics, 7th edn. London: Macmillan, 1907.
Sober, E., ‘Why Logically Equivalent Predicates May Pick out Different Properties.’
American Philosophical Quarterly, 19 (1982): 183–9.
Stalnaker, Robert, ‘Possible Worlds.’ Nous, 10 (1976): 65–75.
Stevenson, Charles L., ‘The Emotive Meaning of Ethical Terms.’ Mind, 46 (1937):
14–31.
—— Ethics and Language. New Haven: Yale University Press, 1944.
—— Facts and Values: Studies in Ethical Analysis. New Haven: Yale University Press,
1963.
Sturgeon, Nicholas L., ‘Moral Explanations.’ In David Copp and David
Zimmerman (eds.), Morality, Reason and Truth. Totowa, NJ: Rowman & Allanheld,
1985: 49–78.
Tversky, Amos, and Kahneman, Daniel, ‘The Framing of Decisions and the
Psychology of Choice.’ Science, 211 (1981): 453–8.
Williams, Bernard, Ethics and the Limits of Philosophy. Cambridge, Mass.: Harvard
University Press, 1985.
This page intentionally left blank
15
Moral Intuitionism Meets Empirical
Psychology
Walter Sinnott-Armstrong
. . . if this be all, where is his ethics? The position he is maintaining is
merely a psychological one.
(Moore 1903: § 11, 11)¹
G. E. Moore’s diatribe against the naturalistic fallacy in 1903 set the stage for
most of twentieth-century moral philosophy. The main protagonists over the
next sixty years were intuitionists and emotivists, both of whom were convinced by Moore that empirical science is irrelevant to moral philosophy and
common moral beliefs. Even in the 1970s and 1980s, when a wider array of
moral theories entered the scene and applied ethics became popular, few moral
philosophers paid much attention to developments in biology and psychology.
This isolation must end. Moral philosophers cannot continue to ignore
developments in psychology, brain science, and biology. Of course, philosophers need to be careful when they draw lessons from empirical research. As
Moore and his followers argued, we should not jump straight from descriptive
premises in psychology or biology to positive moral conclusions or normative conclusions in moral epistemology. That would be a fallacy.² Nonetheless, psychology
can still affect moral philosophy in indirect ways. That is what I want to illustrate
here. I will trace an indirect path from empirical premises to a normative conclusion
For comments on drafts and oral presentations, I thank Robert Audi, Paul Bloomfield, George Pappas,
Bernard Gert, Terry Horgan, Don Loeb, James Moor, Matthew Nudds, Diana Raffman, John Skorupski,
Roy Sorensen, and others in audiences at the Ohio State University, the University of Connecticut, the
University of Vermont, the University of Nebraska at Lincoln, and the University of Edinburgh.
¹ Here Moore was discussing the claim that the object of desire is not pleasure, but his charge against
naturalism extended much further.
² Although some such arguments are formally valid. See my 2000: 159–74. I add the qualification
‘positive’ because ‘Bertie and Madeleine are dead’ might entail ‘It is not the case that Bertie ought to marry
Madeleine.’
340
Walter Sinnott-Armstrong
in moral epistemology. In particular, I will argue that some recent research in
psychology and brain science undermines moral intuitionism.
1. What is moral intuitionism?
Some philosophers define moral intuitionism as the structural view that there
are many moral values or requirements with no systematic unification or ranking. Other philosophers see moral intuitionism as the metaphysical view that
moral properties are non-natural. Neither of these views concerns me here. I
mention them only to set them aside.
The kind of moral intuitionism that is my target here is a position in moral
epistemology, which is general epistemology applied to moral beliefs. The
deepest challenge in moral epistemology, as in general epistemology, is raised
by a skeptical regress argument: Someone is justified in believing something
only if the believer has a reason that is expressible in an inference with premises
that the believer is already justified in believing. This requires a chain of inferences that must continue infinitely, close into a circle, or stop arbitrarily.
Academic skeptics reject all three options and conclude that there is no way for
anyone to be justified in believing anything. The same regress arises for moral
beliefs (cf. Sinnott-Armstrong 1996: 9–14; 2002a).
The simplest way to stop this regress is simply to stop. If a believer
can work back to a premise that the believer is justified in believing without
being able to infer that premise from anything else, then there is no new
premise to justify, so the regress goes no further. That is how foundationalists
stop the regress in general epistemology. Moral intuitionists apply foundationalism to moral beliefs as a way to stop the skeptical regress regarding
moral beliefs.
The motivation behind moral intuitionism is not always to stop the
skeptical regress,³ but that use of moral intuitionism is common and is what
concerns us here, so we can use it to decide among possible definitions of moral
intuitionism. What we need to define is the weakest version of moral intuitionism
³ As George Pappas reminded me, part of foundationalism can be separated from the denial of skepticism. Foundationalists can be skeptics if foundationalism claims only that a believer is justified in a belief
only if the belief is either non-inferentially justified or inferable from non-inferentially justified beliefs.
Non-skeptical foundationalists merely add that some beliefs are justified. Analogously, moral intuitionism
could be seen as a claim about the structure of justification separate from any denial of skepticism.
Nonetheless, I will define moral intuitionism to include the denial of moral skepticism because almost all
actual moral intuitionists do deny skepticism and because I am concerned with whether moral intuitionism
can succeed as a response to skepticism.
Moral Intuitionism
341
that is strong enough to solve the regress problem that would lead to moral
skepticism. Here it is:
Moral intuitionism is the claim that some people are adequately epistemically
justified in holding some moral beliefs independently of whether those
people are able to infer those moral beliefs from any other beliefs.⁴
Several features of this definition are worth highlighting.
So defined, moral intuitionism is not about knowledge. It is about justified
belief. This makes it normative. Psychologists sometimes define intuitionism
as a descriptive claim about the nature and origins of moral beliefs (see Haidt
2001). Such descriptive claims are not intended to stop the skeptical regress; so
they do not concern me here.
More specifically, the defining claim of moral intuitionism is about what is
epistemically justified because moral skeptics win if the only justification for
holding moral beliefs is that those belief states have beneficial practical effects.
Similarly, moral skeptics win if some moral beliefs are inadequately justified
but none is adequately justified, that is, justified strongly enough that the
believer ought to believe it as opposed to denying it or suspending belief.⁵
Accordingly, I will henceforth use ‘justified’ as shorthand for ‘adequately
epistemically justified’.
To say that moral believers are justified independently of an inferential
ability is just to say that they would be justified even if they lacked that ability,
that is, even if they were not able to infer those beliefs from any other beliefs.
This independence claim can hold even when moral believers are able to
infer those moral beliefs from other beliefs as long as they do not need that
inferential ability to be justified.⁶ This notion of need will become prominent
later.
⁴ Although I define moral intuitionism in terms of ‘some’ believers and ‘some’ moral beliefs, all actual
moral intuitionists claim that a significant group of believers and beliefs can be justified non-inferentially. It
also might seem odd that a theory counts as moral intuitionism on my definition if it holds that some beliefs
based on testimony are justified independently of any inference or inferential ability. However, my arguments will apply to such views, so I see no pressing need to complicate my definition so as to avoid these
problems.
⁵ I will discuss pro tanto justifiedness in responses to objections below. For more detail on kinds of justifiedness, see my 2002b: 17–25.
⁶ Compare Ross 1930: 29: ‘without any need of proof ’. (Since not all inferences are proofs, and Ross does
not mention abilities, he might not deny the need for an inferential ability.) Contrast Moore (1903: 77), who
sees moral intuitions as unprovable. Moore’s stronger claim is not needed to stop the skeptical regress. Notice
also that opponents of moral intuitionism, who claim that an inferential ability is necessary, do not have to
claim that any inferential ability is sufficient to make any moral belief justified. At least the moral belief must
also be based on the inferential ability. Other necessary conditions might also have to be met. Opponents of
moral intuitionism can hold that an inferential ability is needed without specifying what, if anything, else is
needed and, so, without specifying what is sufficient for justified moral belief.
342
Walter Sinnott-Armstrong
I infer a belief when I go through a reasoning process of which the belief is
the (or a) conclusion and other beliefs are premises. A believer is able to draw
such an inference when the believer has enough information to go through a
reasoning process that results in this belief if he had enough incentive and time
to do so. This ability does not require self-consciousness or reflection about
the beliefs or abilities. All that is needed, other than general intelligence, is
for the requisite information to be encoded appropriately in the believer’s
brain at the time of belief.⁷
Some moral intuitionists claim only that certain moral beliefs are justified
independently of any actual inference. However, that weak moral intuitionism
is not enough to stop the skeptical regress. Even if whether certain moral beliefs
are justified does not depend on any actual inference, it still might depend on
the believer’s ability to infer them from other beliefs. The ability to draw an
inference cannot make a belief justified if beliefs in the inference’s premises are
not themselves justified. This requirement is enough to restart a skeptical
regress. Thus, to meet the skeptical challenge, moral intuitionists must make
the strong claim that some moral believers are adequately epistemically justified in holding some moral beliefs independently of any ability to infer the
moral belief from any other belief.⁸ So that’s what they claim.
Although this claim is strong, it has many defenders. Rational intuitionists
see basic moral beliefs as analogous to beliefs in mathematical axioms, which are
taken to be justified independently of inference. Moral sense theorists assimilate
particular moral beliefs to perceptual beliefs, which are supposed to be justified
independently of inference. More recently, reliabilists hold that any belief is justified if it results from a reliable process, regardless of whether that process has
anything to do with any inference. I group these views together under my broad
definition of moral intuitionism because my arguments will apply to them all.
Under my definition, there are at least two ways to deny moral intuitionism.
Moral intuitionists claim that some moral believers would be justified even if
they did not have any ability to infer their moral beliefs from any other beliefs.
Some opponents object that moral beliefs always depend on some inference or
inferential ability. However, the evidence (cf. Haidt 2001) strongly suggests
⁷ The relevant notion of ability, then, is not the same as when I am able to become justified in believing
that there are ten coins in my pocket because I could take them out of my pocket and count them. To have
an ability of the relevant kind, I must be able to infer the belief from other beliefs that I already have without
gaining any new information. I hope that it is also clear that, when I write about needing ‘an inferential ability’, I am not referring to a general ability to draw just any old inference, or any inference of a certain form.
What is at issue is the ability to infer the specific moral belief from other beliefs.
⁸ The stronger claim is also needed for moral intuitionism to contrast with its traditional opponent,
moral coherentism, since coherentists do not claim that believers must actually draw inferences in order to
be justified.
Moral Intuitionism
343
that people often have moral beliefs that do not result from any actual
inference. It is harder to tell whether any moral beliefs are independent of any
ability to infer. Nonetheless, I will grant for the sake of argument that some
moral beliefs are spontaneous in the sense that they are independent of any
inference or inferential ability.
Other opponents of moral intuitionism deny that moral believers are ever
justified in holding such spontaneous moral beliefs if they lack certain inferential abilities. This conclusion follows if inferential abilities are always needed
for a moral believer to be justified. That is what I will try to show.
2. When is confirmation needed?
We cannot answer this question directly. If a moral intuitionist baldly asserts
that we do not need inferential abilities to back up our spontaneous moral
beliefs, then this assertion begs the question. Similarly, if a critic of moral
intuitionism baldly asserts that we do need inferential abilities to back up our
spontaneous moral beliefs, then this assertion also begs the question. Neither
side can win so easily. We need a less direct method.
One alternative uses analogies to non-moral beliefs. This path is fraught
with peril, but it might be the only way to go. What this approach does is
appeal to non-moral cases to develop principles of epistemic need and then
later apply those principles back to moral beliefs. Let’s try it.
I will formulate my principles in terms of when confirmation is needed, but
they do not claim that the believer needs to go through any process of confirming the belief. The point, instead, is only that some confirmation needs to
be available at least implicitly as information stored somehow in the believer
that gives the believer an ability to infer the belief from some other beliefs.⁹ The
question is when some such confirmation is needed in non-moral cases.
Suppose that I listen to my daughter in a piano competition. I judge that she
played great and her rival was mediocre. Am I justified in trusting my judgment? Not if all I can say is, ‘Her performance sounded better to me.’ I am too
biased for such immediate reactions alone to count as evidence. I still might be
justified, if I am able to specify laudable features of her performance, or if I
know that others agree, but some confirmation seems needed. Generalizing,
Principle 1: confirmation is needed for a believer to be justified when the
believer is partial.
⁹ Confirmation need not always be evidence, since I want confirmation to include defeator defeators,
that is, reasons to discount what would otherwise keep a belief from being justified.
344
Walter Sinnott-Armstrong
This principle also applies to direct perceptual judgments, such as when I
believe that my daughter played middle C at just the right time in the midst of
her piece. This partly explains why we prefer umpires, referees, and judges not
to be parents of competitors. Even reliabilists can admit this principle because
partiality often creates unreliability.
Second, imagine that each of us adds a column of figures, and I get one sum,
but you get a different sum. Maybe I would be justified in believing that I am
right if you were my child and I was helping you with your homework.
However, if you are just as good at arithmetic as I am, then, when we get
different answers, we need to check again to find out who made a mistake
before either of us can be justified in believing that his or her answer is the correct one. We owe each other that much epistemic respect. The best explanation
of this natural reaction seems to be
Principle 2: confirmation is needed for a believer to be justified when
people disagree with no independent reason to prefer one belief or believer
to the other.
This principle also applies when the person on the sidewalk looks like Tom
Cruise to me but not to you. If I have no reason to believe that I am better than
you at this identification, then I am not justified in believing that your belief is
incorrect or that mine is correct.
A third principle concerns emotions. When people get very angry, for example,
they tend to overlook relevant facts. They often do not notice excuses or apologies
by the person who made them angry. We should not generalize to all emotions,
but we can still endorse something like this:
Principle 3: confirmation is needed for a believer to be justified when the
believer is emotional in a way that clouds judgment.
This explains why jurors are dismissed from a case that would make them too
emotional. This principle applies even if their emotions do not bias them
towards either side, so it is distinct from Principle 1, regarding partiality.
Next consider illusions. At least three kinds are relevant here. First, some
illusions are due to context. Objects look larger when they are next to smaller
objects, and they look smaller when they are next to larger objects. Since our
estimates of their sizes are affected by their surroundings, we are not justified in
trusting our estimates until we check their sizes in other circumstances or by
other methods.
A second group of illusions arises from generalizations. For example, an oval
that is shaded on top looks concave, but an oval that is shaded on the bottom
looks convex. The explanation seems to be that our cognitive apparatus
Moral Intuitionism
345
evolved in circumstances where the light usually came from above, which
would produce a shadow on the top of a concave oval (such as a cave opening)
and on the bottom of a convex oval (such as an egg). Since we often overextend
generalizations like this, we are not justified in trusting beliefs that depend on
such generalizations until we check to determine whether our circumstances
are exceptional.
The third kind of illusion involves heuristics, which are quick and simple
decision procedures. In a passage with a thousand words, how many seven-letter
words have the form ‘_ _ _ _ in _’? How many seven-letter words have the form
‘_ _ _ _ ing’? Most people estimate more words have the latter form, although
that is impossible, since every word of the form ‘_ _ _ _ ing’ also has the form
‘_ _ _ _ in _’. Why do people make this simple mistake? They seem to test how
likely something is by trying to imagine examples and guessing high if they easily think of lots of examples. This is called the availability heuristic (Kahneman
et al. 1982). Most people easily produce words that end in ‘ing’, but they have
more trouble coming up with words that end in ‘in _’ because they do not think
of putting ‘g’ in the last place. In cases like this, the availability heuristic is
misleading. Accordingly, they do not seem adequately epistemically justified in
trusting beliefs based on such heuristics until they check on whether they are in
circumstances where the heuristics work.
This quick survey of three common kinds of illusion suggests
Principle 4: confirmation is needed for a believer to be justified when the
circumstances are conducive to illusion.
This principle would apply as well to many other kinds of illusions.
A fifth and final principle considers the source of a belief. If you believe that
George Washington never told a lie, and if this belief comes from a legend
spread by Washington’s allies to gain power, then you are not justified in believing the legend, though it still might be true. Even if you believe only that
Washington was unusually honest, this belief might be a lingering effect of this
childhood story, and then its origin makes this belief need confirmation. The
point can be generalized into something like
Principle 5: confirmation is needed for a believer to be justified when the
belief arises from an unreliable or disreputable source.
This principle also explains why we do not view people as justified in beliefs
based only on prejudice and stereotypes.
These five principles, although distinct, complement each other. When a
belief is partial, controversial, emotional, subject to illusion, and explicable by
dubious sources, then all of these principles apply. In such cases, they work
346
Walter Sinnott-Armstrong
together to make it even clearer that confirmation is needed for justified belief.
Even if not all of these principles apply, the more that do apply, the clearer it
will be that there is more need for more confirmation. We might think of this
as a sixth principle.
I do not claim that these principles are precise or that my list is complete.¹⁰
What I do claim is that these principles or some close relatives seem plausible
to most people and are assumed in our shared epistemic practices. I also claim
that they make sense because they pick out features that are correlated with reliability and other epistemic values.
Most importantly, I claim that these principles apply in all areas of belief. My
illustrations include beliefs about arithmetic, language, history, identity, value,
sound, size, and shape, but the same principles apply in scientific research, religion, and so on. The main question here is whether they apply to moral beliefs.
Admittedly, morality might be a special case where these principles do not
apply. However, unless someone can point to a relevant difference between
these other areas and moral beliefs, it seems only fair to apply these same standards to moral beliefs when asking whether moral beliefs are justified. So that’s
what I will do.
3. When are moral beliefs justified?
Some of these principles can be applied only with the help of empirical
research. Others are easier to apply. Let’s start with the easy ones.
Partiality
Principle 1 says that partiality adds a need for confirmation. But what is partiality? A judge is often called partial when the judge’s self-interest is affected by
the outcome of the case. However, if the judge’s self-interest does not influence
the judge’s decision, so the judge would have made the same decision if the
judge’s self-interest had not been involved, then it is natural to say that the
judge’s decision is not partial, even if the judge is partial. Analogously, believers can be called partial whenever their beliefs affect their self-interest either
directly or indirectly (by affecting the interests of people whom they care
about). Beliefs are then partial only when the believer’s self-interest influences
whether the believer holds that belief. Thus, a partial believer can hold an
¹⁰ One additional principle might claim that confirmation is needed when errors are costly. This principle applies to moral beliefs insofar as moral errors are costly.
Moral Intuitionism
347
impartial belief (or can hold it impartially) if the believer has an interest in
holding the belief, but that interest does not influence whether the believer
holds that belief.
Partiality of a belief can’t be all that triggers Principle 1. To see why, recall the
examples that motivated Principle 1. Because I am biased in favor of my daughter, even after watching her play in the piano competition, I need confirmation
to be justified in believing that my daughter played better than her rival. Maybe
my interest in her victory did not influence my assessment, but the danger of
such influence is enough to create a need for confirmation. Admittedly, if I can
rule out such influence, then I can be justified in believing that my daughter
played better, but the only way to rule out such influence involves independent
confirmation. Thus, confirmation seems needed when the believer is partial,
even if the believer is not actually influenced by that partiality, so the belief is
not partial. Since confirmation is also needed when the belief is partial,
Principle 1 requires confirmation when either the believer or the belief is
partial.
To apply this principle to moral beliefs, we need to determine whether moral
beliefs affect our self-interest either directly or indirectly. The answer seems
clear: Moral beliefs affect us all. It can be very expensive to believe that we are
morally required to help the needy, but it can be even more expensive if others
do not believe that they are morally required to help us when we are in need. It
can also cost a lot to believe that we have to tell the truth or to keep a promise.
And all of us know or should know that, if killing, stealing, lying, cheating, and
promise-breaking were generally seen as morally permitted, then we would be
more likely to get hurt by others doing such acts. Life would be more ‘solitary,
poor, nasty, brutish, and short’, as Hobbes put it. Moreover, if an individual did
not see such acts as immoral, then he or she would be more likely to do them
and then to be punished in various ways—‘if not by law, by the opinion of his
fellow creatures’, as Mill said. Special interests also arise in special cases:
Women and men know or should know that, if abortion is not seen as morally
permissible, then they, their friends, or their daughters will be more likely to
suffer more. Moral beliefs about affirmative action affect the interests of the
preferred groups and also the non-preferred groups. And so on. Indeed, on
many views, what makes an issue moral in nature is that interests are significantly affected by the judged actions. Since moral beliefs about actions affect
those actions, our moral beliefs themselves affect our interests at least
indirectly. Finally, social groups often form around and then solidify moral
beliefs (cf. Chen and Tetlock et al. as discussed in Haidt 2001). People who
believe that homosexuality is immoral find it harder to get along with homosexuals and easier to get along with homophobes. Conversely, people who
348
Walter Sinnott-Armstrong
believe that homosexuality is not immoral find it harder to get along with
homophobes and easier to get along with homosexuals. Some might try to fake
moral beliefs in order to get along, but few of us are good enough actors, and
those who believe that homosexuality is immoral usually also believe that they
ought not to pretend otherwise in order to get along with homosexuals. Thus,
our moral beliefs affect our social options as well as our actions.
Many moral beliefs might seem to have no effect on us. If I believe that it was
immoral for Brutus to stab Caesar, this moral belief by itself will not change my
social options or rule out any acts that I could do today. Still, given universalizability, my judgment of Brutus seems to depend on a principle that does apply
in other cases where my self-interest is involved more directly. Arguments from
analogy also might force me to take a moral stand that affects my interests.
Thus, any moral belief can affect my self-interest indirectly.
Because our moral beliefs affect our self-interest so often in so many ways at
least indirectly, we cannot be justified in assuming that any of us is ever fully
impartial as a moral believer. Even if our self-interest is not involved in some
exceptional case, we still need a reason to believe that our self-interest is not
affected in that case, since we know or should know that such effects are very
common and often hidden. The facts that partiality is so common in this area
and so difficult to detect in ourselves are what create a need for confirmation of
all moral beliefs, according to Principle 1.
Disagreement
Principle 2 says that disagreement creates a need for confirmation. Many
people seem to think that this principle is easy to apply to moral beliefs because
moral disagreement is pervasive. In their view, people from different cultures,
time periods, social classes, and genders disagree about a wide variety of particular
moral judgments and general moral principles.
Actually, the extent of moral disagreement is not obvious. One reason is that
people who seem to disagree are often judging different actions or using different concepts. Also, many apparently moral disagreements are really factual,
since those who seem to disagree morally would agree in their moral judgments
if they agreed about the facts.
Still, straightening out concepts and non-moral facts seems unlikely to
resolve all apparently moral disagreements. One reason is that people often
express different moral beliefs about hypothetical cases where all of the facts are
stipulated, so these moral believers seem to accept the same non-moral facts.
Admittedly, descriptions of these situations usually leave out important facts,
and moral believers might interpret the hypothetical cases in light of different
Moral Intuitionism
349
background beliefs. But there still seem to be lots of cases where all relevant
non-moral facts are agreed upon without leading to agreement in moral belief.
This claim is supported by a study in which Jana Schaich Borg and I
surveyed fifty-two undergraduates at Dartmouth College, using thirty-six
scenarios, including the well-known side-track and fat-man trolley cases. In
both cases, five people tied to a track will be killed by a runaway trolley if you
do nothing, and the only way to save the five is to kill one other person.¹¹ In the
side-track version, you can save the five only by pulling a lever to divert the trolley onto a side-track where it will run over one victim. In the fat-man variation,
you can save the five only by pushing a fat man in front of the trolley so that his
body will stop the trolley before it hits the five. In two rounds, 35 per cent then
43 per cent of our subjects said that it would be wrong to divert the trolley onto
the side-track. When the same scenario was described with more vivid
language, 61 per cent then 45 per cent judged diversion wrong in the two
rounds. In contrast, 76 per cent in the first round then 88 per cent in the
second round judged it wrong to push the fat man. (Interestingly, there were
still 35 per cent in the first round and 18 per cent in the second round who said
that they would push the fat man.) These percentages did not change much
with more vivid language. Thus, we found significant disagreement about the
very cases that philosophers often cite to support their theories.¹²
There is, admittedly, more agreement about other cases: Would it be wrong
to push the fat man in front of the trolley just because you are angry at him for
beating you at golf when killing him will not save or help anyone else? I hope
and expect that 100 per cent would answer, ‘Yes.’ But what would that show?
The universality of moral beliefs about cases like this one could hardly be used
to justify any moral theory or any controversial moral belief.¹³ Such cases
cannot get moral intuitionists all that they seem to want.¹⁴
Moral intuitionists might respond that all they claim is that some moral
beliefs are non-inferentially justified. One case seems enough to establish that
¹¹ These cases originate from Foot 1967. If such cases seem unrealistic, see the real case at
www.cnn.com/2003/US/West/06/21/train.derail.ap/index.html.
¹² For more evidence of disagreement, see Haidt et al. 1993.
¹³ Compare Descartes’s ‘I think,’ which is nowhere near enough to ground science. Notice also that I am
talking about actual moral beliefs, not possible moral beliefs. There might be an infinite number of possible
moral beliefs that would garner agreement from everyone who understands them. However, there still might
be a high rate of disagreement among the actual moral beliefs that people bother to form. That is what matters when we ask whether our actual moral beliefs are justified.
¹⁴ Some intuitionists might claim agreement on qualified general principles, such as that it is morally
wrong to kill anyone in a protected class without an adequate reason. Of course, those who accept this formula still disagree about which class is protected and which reasons are adequate. Similarly, although many
people (today!) agree that all moral agents deserve respect, different people count different acts as violating
that rule by showing disrespect. It is not clear whether to count agreement on such indeterminate formulas
as real moral agreement. See Snare 1980.
350
Walter Sinnott-Armstrong
claim. However, the fact that there is so much disagreement in other cases
affects the epistemology of cases where there is no disagreement. Compare a
box with one hundred thermometers. We know that many of them don’t work,
but we are not sure how many. If we pick one thermometer arbitrarily from the
box, and it reads 77 degrees, then we are not justified in believing that the
temperature really is 77 degrees, even if we were in fact lucky enough to pick a
thermometer that works. Of course, if we confirm that this thermometer
works, such as by testing it against other thermometers, then we can use it to
form justified beliefs, but we cannot be justified in trusting it before we confirm that it works.¹⁵ Similarly, if we know that many moral intuitions are unreliable because others hold conflicting intuitions, then we are not justified in
trusting a particular moral intuition without some reason to believe that it is
one of the reliable ones. If we know that everyone agrees with that particular
moral intuition, then we might have reason to trust it. But that is just because
the known agreement provides confirmation, so it does not undermine the
point that some confirmation is needed, as Principle 2 claims.
Emotion
Next consider Principle 3, which says that emotions that cloud judgment
create a need for confirmation. It is hard to tell whether this principle applies
to moral beliefs. Philosophers and others have argued for millennia about
whether moral beliefs are based on emotion or on reason. They also argue
about which emotions, if any, cloud judgment. How can we resolve these
debates? Luckily, some recent empirical studies suggest an answer.
Haidt and his group have been accumulating an impressive body of
behavioral evidence for what they call the social intuitionist model:
This model suggests that moral judgment is much like aesthetic judgment: we see an
action or hear a story and we have an instant feeling of approval or disapproval. These
feelings are best thought of as affect-laden intuitions, as they appear suddenly and
effortlessly in consciousness, with an affective valence (good or bad), but without any
feeling of having gone through any steps of searching, weighing evidence or inferring a
conclusion. (Greene and Haidt 2002: 517)¹⁶
¹⁵ According to Bayesians, if the temperature feels to us as if it is about 70 degrees, so we start with that
assumption, then the fact that this thermometer reads 77 degrees should make us move our estimate towards
77 degrees. How much our estimate should increase depends on our prior assumption about how many thermometers work. This might make it seem as if the thermometer reading can lead to justified belief. However,
if our initial estimates (about temperature and the percentage of working thermometers) are unjustified,
then I doubt that one reading can ground justified belief. Besides, many of us do not form any of the initial
estimates that are needed to start Bayesian reasoning.
¹⁶ See also Haidt 2001. Of course, many other judgments cause emotional reactions ‘suddenly and effortlessly’.
But Haidt argues that emotions drive or constitute moral judgments rather than being effects of those judgments.
Moral Intuitionism
351
Haidt’s behavioral evidence dovetails nicely with independent brain studies.
Moll’s group found that brain tissue associated with emotions becomes more
activated when subjects think about simple sentences with moral content
(e.g. ‘They hung an innocent’) than when they think about similar sentences
without moral content (e.g. ‘Stones are made of water’) (Moll et al. 2001) or
disgusting non-moral sentences (e.g. ‘He licked the dirty toilet’) (Moll et al.
2002a).¹⁷ Similar results were found with pictures in place of sentences (Moll
et al. 2002b).
Studies by Joshua Greene and his colleagues are even more fascinating
because they distinguish kinds of moral beliefs (2001).¹⁸ Greene’s group
scanned brains of subjects while they considered what was appropriate in three
kinds of dilemmas: non-moral dilemmas, personal moral dilemmas, and
impersonal moral dilemmas. A moral dilemma is personal if and only if one of
its options is likely to cause serious harm to a particular person other than by
deflecting an existing threat onto a different party (Greene and Haidt 2002:
519). A standard personal moral dilemma is the fat-man trolley case. A paradigm impersonal moral dilemma is the side-track trolley case. These different
moral cases stimulated different parts of the brain. While considering appropriate action in impersonal dilemmas, subjects showed significant activation in
brain areas associated with working memory but no significant activation in
areas associated with emotion. In contrast, while considering appropriate
action in personal dilemmas, subjects showed significant activation in brain
areas associated with emotion and under-activation (below the resting
baseline) in areas associated with working memory. It is not obvious what
to make of these results. Brain scientists do not know how to interpret underactivation in general. Nonetheless, one natural speculation is this: When asked
about pushing the fat man, subjects react, ‘That’s so horrible that I can’t even
think about it.’ Emotions stop subjects from considering the many factors in
these examples. If this interpretation is correct, then many pervasive and
fundamental moral beliefs result from emotions that cloud judgment.¹⁹
Some moral intuitionists might argue that there is no need to consider
anything else when the proposed action is the intentional killing of an innocent
fat man. It might even be counterproductive to consider additional factors,
since they might lead one away from the correct belief. Such responses,
¹⁷ It would be interesting to test reactions to negations, such as ‘They did not hang an innocent’ and ‘He
did not lick the dirty toilet.’
¹⁸ This article also reports timing studies that confirm the different roles of emotion in different moral
beliefs.
¹⁹ Philosophers should notice that what Greene calls ‘personal dilemmas’ include most proposed counterexamples to consequentialism. If those intuitions are unjustified, then Greene’s study might help consequentialists defend their moral theory, even if other intuitions are not affected.
352
Walter Sinnott-Armstrong
however, assume that it is morally wrong to push the fat man, so they beg the
question here. When asking whether a moral belief is justified, we should not
assume that the only relevant factors are those that would be relevant if the
belief were true. Ridiculous moral beliefs could be defended if that method
worked.
Still, moral intuitionism is hardly refuted by these experiments because
Greene’s results must be replicated and interpreted much more carefully.
(Some initial replication can be found in Greene 2004.) All I can say now is that
such brain studies seem to provide some evidence that many moral judgments
result from emotions that cloud judgment.
Additional evidence comes from Wheatley and Haidt (2005). They gave participants the post-hypnotic suggestion that they would feel a pang of disgust
whenever they saw either the word ‘take’ or the word ‘often’. Participants were
later asked to make moral judgments about six stories designed to elicit mild to
moderate disgust. When a story contained the word that elicited disgust in a
participant, that participant was more likely to express stronger moral
condemnation of acts in the story. Moral judgments were then affected by elements of the story that could not determine the accuracy or acceptability of
those moral judgments. In that sense, emotions clouded their judgment.
Because independently caused emotions can distort moral beliefs in such ways,
moral believers need confirmation in order to be justified in holding their
moral beliefs.
Illusions
To apply Principle 4 to moral beliefs, we again need empirical research, but this
time in cognitive science rather than brain science. I mentioned three kinds of
illusions that should be considered separately.
The first kind of illusion occurs when appearances and beliefs depend on
context. An interesting recent example comes from Peter Unger, who found
that the order in which options are presented affects beliefs about whether a
given option is morally wrong. He also claims that people’s moral beliefs about
a certain option depend on whether that option is presented as part of a pair or,
instead, as part of a series that includes additional options intermediate between
the original pair (Unger 1996: 88–94).²⁰ Since order and intermediate options
²⁰ Unfortunately, Unger does not describe the method or precise results of his informal survey, so there is
room for more careful empirical work to test his claims. Some philosophical support comes from moral paradoxes, which often arise through the mechanisms that Unger describes. One example is the mere addition
paradox of Parfit 1984, in which B seems worse than A when the two are compared directly, but it seems not
worse than A when Parfit interjects A⫹ and Divided B as options intermediate between A and B.
Moral Intuitionism
353
are not morally relevant factors that could affect the moral wrongness of the
judged option, the fact that moral beliefs are affected by these factors shows
that moral beliefs are unreliable in such cases. That is why confirmation is
needed. One still might confirm one’s moral belief by reconsidering the issue
in several contexts over time to see whether one’s moral belief remains stable,
but that is just a way of seeking confirmation, so it does not undermine my
point that confirmation is needed.
The second kind of illusion arises from overgeneralization. Such illusions
also affect moral beliefs. Jonathan Baron even argues that all ‘nonconsequentialist principles arise from overgeneralizing rules that are consistent with
consequentialism in a limited set of cases’ (1994: 1). But one need not accept
consequentialism in order to admit that many people condemn defensible
lying, harming, and love-making because they apply generalizations to
exceptional cases. We probably disagree about which moral beliefs are overgeneralizations, but we should agree that many people overgeneralize in ways
that create illusions of moral wrongness. In any such case, the moral believer
could argue that this case is not an exception to the generalization, but, as
before, that is just a way of seeking confirmation, so it does not undermine my
point that this kind of illusion creates a need for confirmation.
Heuristics, which are quick and simple decision procedures, also create
illusions in morality. One reason is that many moral beliefs depend on consequences and probabilities, for which we often lack adequate evidence, and then
we have to guess these probabilities. Such guesses are notoriously distorted by
the availability heuristic, the representative heuristic, and so on.²¹ Even when
moral beliefs do not depend on probability assessments, moral beliefs are
affected by the so-called ‘I agree with people I like’ heuristic (cf. Chaiken and
Lord, Ross, and Lepper as discussed by Haidt 2001). When people whom we
like express moral beliefs, we tend to go along and form the same belief. When
people whom we dislike oppose our moral beliefs, we tend to hold on to them
in spite of contrary arguments. This heuristic often works fine, but it fails in
enough cases to create a need for confirmation.
In addition to these three kinds of illusions, moral beliefs also seem subject
to framing effects, which were explored by Kahneman and Tversky (1979). In
one famous experiment, they asked some subjects this question:
Imagine that the U.S. is preparing for an outbreak of an unusual Asian disease which is
expected to kill 600 people. Two alternative programs to fight the disease, A and B,
have been proposed. Assume that the exact scientific estimates of the consequences of
²¹ Kahneman et al. 1982. Lackey 1986: 634, suggests how such heuristics might explain conflicting
moral intuitions about nuclear deterrence.
354
Walter Sinnott-Armstrong
the programs are as follows: If program A is adopted, 200 people will be saved. If
program B is adopted, there is a 1/3 probability that 600 people will be saved, and a 2/3
probability that no people will be saved. Which program would you choose?
The same story was told to a second group of subjects, but these subjects had to
choose between these programs:
If program C is adopted, 400 people will die. If program D is adopted, there is a 1/3
probability that nobody will die and a 2/3 probability that 600 will die.
It should be obvious that programs A and C are equivalent, as are programs B
and D. However, most subjects who chose between A and B favored A, but
most subjects who chose between C and D favored D. More generally, subjects
were risk averse when results were described in positive terms (such as ‘lives
saved’) but risk seeking when results were described in negative terms (such as
‘lives lost’ or ‘people who die’).
The question in this experiment was about choices rather than moral
wrongness. Still, the subjects were not told how the policies affect them personally, so their choices seem to result from beliefs about which program is
morally right or wrong. If so, the subjects had different moral beliefs about programs A and C and about programs B and D. The only difference within each
pair is how the programs are framed or described. Thus, descriptions seem to
affect moral beliefs. Descriptions cannot affect what is really morally right or
wrong. Hence, these results suggest that such moral beliefs are unreliable.
Moral intuitionists could claim that moral intuitions are still reliable when
subjects have consistent beliefs after considering all relevant descriptions. But
then moral believers would need to know that their beliefs are consistent and
that they are aware of all relevant descriptions before they could be justified in
holding moral beliefs. Framing effects distort moral beliefs in so many cases
that moral believers need confirmation for any particular moral belief.
To see how deeply this point cuts, consider Warren Quinn’s argument for
the traditional doctrine of doing and allowing, which claims that stronger
moral justification is needed for killing than for letting die. In support of this
general doctrine, Quinn appeals to moral intuitions of specific cases:
In Rescue I, we can save either five people in danger of drowning at one place or a single person in danger of drowning somewhere else. We cannot save all six. In Rescue II,
we can save the five only by driving over and thereby killing someone who (for an
unspecified reason) is trapped on the road. If we do not undertake the rescue, the
trapped person can later be freed. (1993: 152)
Most people judge that saving the five is morally wrong in Rescue II but not in
Rescue I. Why do they react this way? Quinn assumes that these different
Moral Intuitionism
355
intuitions result from the difference between killing and letting die or, more
generally, doing and allowing harm. However, Tamara Horowitz uses a different
distinction (between gains and losses) and a different theory (prospect theory)
to develop an alternative explanation of Quinn’s moral intuitions:
In deciding whether to kill the person or leave the person alone, one thinks of the
person’s being alive as the status quo and chooses this as the neutral outcome. Killing
the person is regarded as a negative deviation. . . . But in deciding to save a person who
would otherwise die, the person being dead is the status quo and is selected as the
neutral outcome. So saving the person is a positive deviation. . . . (1998: 153)
The point is that we tend to reject options that cause definite negative deviations from the status quo. That explains why subjects rejected program C but
did not reject program A in the Asian disease case (despite the equivalence
between those programs). It also explains why we think that it is morally wrong
to ‘kill’ in Rescue II but is not morally wrong to ‘not save’ in Rescue I, since
killing causes a definite negative deviation from the status quo. This explanation clearly hinges on what is taken to be the status quo, which in turn depends
on how the options are described. Quinn’s story about Rescue I describes the
people as already ‘in danger of drowning’, whereas the trapped person in
Rescue II can ‘later be freed’ if not for our ‘killing’ him. These descriptions
affect our choice of the neutral starting point. As in the Asian disease cases, our
choice of the neutral starting point then affects our moral intuitions. Horowitz
adds, ‘I do not see why anyone would think the distinction [that explains our
reactions to Quinn’s rescue cases] is morally significant, but perhaps there is
some argument I have not thought of. If the distinction is not morally significant, then Quinn’s thought experiments do not support one moral theory over
against another’ (1998: 155).
Admittedly, Horowitz’s explanation does not imply that Quinn’s moral
intuitions are false or incoherent, as in the Asian disease case. It does not even
establish that his moral intuitions are arbitrary. As Mark van Roojen says,
‘Nothing in the example shows anything wrong with treating losses from a
neutral baseline differently from gains. Such reasoning might well be appropriate where framing proceeds in a reasonable manner’ (1999).²² Nonetheless,
the framing also ‘might well’ not be reasonable, so the epistemological dilemma
remains: If there is no reason to choose one baseline over the other, then our
moral intuitions seem arbitrary and unjustified. If there is a reason to choose
one baseline over the other, then either we have access to that reason or we do
not. If we have access to the reason, then we are able to draw an inference from
²² Van Roojen might admit that Horowitz’s argument undermines moral intuitionism, since he defends
a method of reflective equilibrium that is coherentist rather than foundationalist.
356
Walter Sinnott-Armstrong
that reason to justify our moral belief. If we do not have access to that reason,
then we do not seem justified in our moral belief. Because framing effects so
often lead to incoherence and error, we cannot be justified in trusting a moral
intuition that relies on framing effects unless we at least can be aware that this
intuition is one where the baseline is reasonable. So Horowitz’s explanation
creates serious trouble for moral intuitionism whenever framing effects could
explain our moral intuitions.
The doctrine of doing and allowing is not an isolated case. It affects many
prominent issues and is strongly believed by many philosophers and common
people, who do not seem to be able to infer it from any other beliefs. If moral
intuitions are unjustified in this case, doubts should arise about a wide range of
other moral intuitions as well.
Origins
Some previous principles look at origins of individual moral beliefs, but
Principle 5 considers the social origins of shared moral beliefs. The two issues
are related insofar as many of our moral beliefs result from training and social
interaction.
Specifically, Principle 5 claims that problematic social origins create a need
for confirmation. To apply this principle, we need to ask whether moral beliefs
have problematic social origins. The social origins of moral beliefs might be
problematic in two ways. First, moral beliefs might be caused by factors that are
unrelated with the truth of those beliefs. Second, the origins of moral beliefs
might be immoral according to those moral beliefs. I will focus on the latter case.
Are the origins of our moral intuitions immoral by their own lights?
Friedrich Nietzsche suggests as much when he argues that Christian morality
results from slaves cleverly overcoming their superiors by re-evaluating values.
Insofar as Christian morality condemns such subterfuge and self-promotion,
Christian morality condemns its own origins, if Nietzsche is correct (Nietzsche
1966).²³ Similarly, Michel Foucault argues at length that moral beliefs express
or result from social power relations. Yet these moral beliefs themselves seem to
condemn the very kind of power that leads to these beliefs. But I don’t want to
rely on Nietzsche or Foucault, at least not in this context, so I will consider
Gilbert Harman’s explanation of the common moral belief that harming
someone is much worse than failing to helping someone in need:
whereas everyone would benefit equally from a conventional practice of trying not to
harm each other, some people would benefit considerably more than others from a
²³ I am not, of course, endorsing Nietzsche’s speculations.
Moral Intuitionism
357
convention to help those who needed help. The rich and powerful do not need much
help and are often in the best position to give it; so, if a strong principle of mutual aid
were adopted, they would gain little and lose a great deal, because they would end up
doing most of the helping and would receive little in return. On the other hand, the
poor and the weak might refuse to agree to a principle of non-interference or noninjury unless they also reached some agreement on mutual aid. We would therefore
expect a compromise [that] would involve a strong principle of non-injury and a
weaker principle of mutual aid—which is just what we now have. (1977: 110; cf.
Scheffler 1982: 113)
Remember also that rich and powerful people have always controlled the
church, the media, and culture, which in turn affect most people’s moral
beliefs. In this context, Harman’s claim is that the self-interest of the rich and
powerful in making everyone believe that harming is worse than failing to help
can explain why so many people believe that harming is worse than failing to
help. But our moral beliefs also seem to condemn such self-serving indoctrination by the rich and powerful, since morality is supposed to consider everyone’s
interests equally. Thus, if Harman is correct, morality condemns its own
origins, as Nietzsche and Foucault claimed.
The point is not that such moral views are internally inconsistent,
self-condemning, or even self-defeating. The point is only that there are
grounds for doubt when beliefs come from disreputable sources. Defenders
of such moral beliefs must admit that the sources of their beliefs are disreputable if Harman’s explanation is accurate. Then they need additional
support for their beliefs beyond the mere fact that those beliefs seem correct
to them.
These speculations about the origins of moral beliefs are mere armchair
psychology. Perhaps more support could be obtained from the literature on
sociobiology or evolutionary psychology. Still, these explanations are likely to
remain very controversial. Luckily, I don’t need to prove them here. I claim only
that these undermining accounts are live possibilities. They seem plausible to
many people and have not been refuted.
That would not be enough if I were arguing for the falsehood of a certain
moral belief, such as Christian morality (from Nietzsche) or the prevalence of
non-injury over mutual aid (from Harman). However, I am not drawing any
substantive moral conclusion. To do so would commit a genetic fallacy, but my
argument is different. My point lies in moral epistemology, and I reach it indirectly. If these disreputable origins are live possibilities, then moral believers
need some independent confirmation that their beliefs are not distorted by
such disreputable origins. This need for independent confirmation then
undermines moral intuitionism.
358
Walter Sinnott-Armstrong
Togetherness
Don’t forget that Principles 1–5 complement each other. If I am right, moral
beliefs are partial, controversial, emotional, subject to illusion, and explicable
by dubious sources, so all of the principles apply. However, even if not all but
only several of them apply, these principles still work together to make it clear
that confirmation is needed for justified moral belief. That undermines moral
intuitionism. It also shows how empirical research can be indirectly relevant to
normative moral epistemology.
4. Objections
None of my arguments is conclusive, so opponents can object at several points.
Here I cannot respond to every objection or to any objection thoroughly. But
I will quickly run through the most formidable objections.
Confirmation
One common objection is that, even if some confirmation is needed, that does
not show that any inference is needed. If we can confirm color beliefs just by
looking again in different light, perhaps we can confirm moral beliefs simply
by reflecting on the moral issue again in a different mood without involving
any substantive moral principle from which we infer our moral belief.
I grant that confirmation does not require an actual inference. To avoid the
skeptical regress, however, moral intuitionists must deny more than the need
for an actual inference. They must deny the need for any ability to infer the
moral belief. It is hard to see how you could confirm a moral belief without
gaining information that makes you able to draw some kind of inference to the
moral belief. Even if you just think about the moral issue several times in
different moods, after such rethinking you have all the information you need
for a simple inference like this:
I hold this moral belief after reflecting on the issue several times in
different moods.
If I hold a moral belief after reflecting on the issue several times in different
moods, then it is usually true.
So, probably, this moral belief is true.
Admittedly, this inference is not a deductive proof. Nor does it infer the moral
belief from a more general substantive moral principle. But no specific kind of
Moral Intuitionism
359
inference is needed. Any kind of inference can lead to a skeptical regress, so
moral intuitionists have to deny dependence on any kind of inference or ability
to infer. And an ability to draw the above kind of inference is needed, since, if
the moral believer does not believe its premises (or something like them), then
it is hard to see why the moral believer is justified in holding the moral belief.
At this point, externalists (including reliabilists) sometimes accuse me of
confusing whether a belief is justified with whether the believer knows that it is
justified. I plead innocent. I do not assume that justified believers must know
or be justified in believing (or even be able to know or be justified in believing)
that they are justified. I claim only that justified moral believers must be able to
infer their moral beliefs from something. Some externalists still deny this, but
their denial is implausible, as I have argued elsewhere (Sinnott-Armstrong
2002b). Besides, most externalists are mainly concerned about non-moral
beliefs, including perceptual beliefs. Principles 1–5 do not apply to perceptual
beliefs in the same way as to moral beliefs. Most perceptual beliefs are not
partial, controversial, emotional, or explicable by dubious sources. Perceptual
illusions are common, but they are normally easier to detect than moral illusions, and they do not affect anything as widespread and fundamental as the
doctrine of doing and allowing. Consequently, externalism and reliabilism
might work for perceptual beliefs even if, as I have argued, moral beliefs need
confirmation of a kind that requires an inferential ability.²⁴
Children
This response leads to another objection. It might seem too strict to require an
inferential ability because then young children cannot have justified moral
beliefs, since they cannot formulate the needed inferences.
I love children. I grant that they can have justified beliefs in other areas, such
as beliefs about food and toys. However, it is not as clear that very young children (say, 1–3 years old) can be justified in holding moral beliefs. After all,
young children often base their normative beliefs on fear of punishment. If
someone believes that stealing is wrong just because he believes that he is likely
to get punished if he steals, then it is not even clear that the belief is a moral
belief. It might be purely prudential. Instead of fear, the basis for some young
children’s moral beliefs might be deference to authorities (or peers). But if
children accept their parents’ word that an act is wrong without any idea of
what makes that act wrong, then these children might not believe that the act
is wrong morally, since they might not believe that there is any specifically
²⁴ I do not claim that moral judgments are the only ones that need confirmation according to Principles
1–5. Several of these principles also apply to beliefs about what is prudent, rational, wise, and beautiful.
360
Walter Sinnott-Armstrong
moral reason not to do the act. Moreover, authorities can make someone
justified only if she is justified in trusting those authorities. Maybe young children are justified in trusting their parents, for example. But then they can infer:
My parents are trustworthy.
My parents tell me that I shouldn’t pull my sister’s hair.
Therefore, I shouldn’t pull my sister’s hair.
If a young child is not able to draw any inference like this, then this child does
not seem justified (even if her parents are trustworthy and even if her belief is
true). Those who think otherwise are too soft on their kids.
Ignorance
Another objection claims that, if a moral believer could not know that moral
beliefs are subject to controversy, partiality, illusion, and so on, then that moral
believer does not need to guard against these problems by getting confirmation.
Children and some adults (such as isolated medieval peasants) might have no
way of discovering such problems for moral beliefs. They certainly lack access
to the psychological research that I cited. So maybe these moral believers do not
need confirmation for their moral beliefs.
This objection confuses two claims. To call a believer unjustified is often to
criticize that believer. Such criticism seems misplaced when the believer is not
responsible for any epistemic failures. Believers are not responsible when they
have no way of knowing that their beliefs are problematic. Thus, if children and
medieval peasants cannot know that moral beliefs are problematic, it seems
odd to call them unjustified.
In contrast, to say that a believer is not justified is not to criticize the believer.
It is only to withhold the praise of calling the believer justified. There is nothing unfair about withholding praise when a believer is not responsible. Thus,
even if children and medieval peasants are not responsible for their epistemic
failures, that does not undermine my claim that they are not justified in their
moral beliefs.
Moreover, even if children and medieval peasants were justified in their moral
beliefs, that would not save moral intuitionists or my readers from the need for
confirmation. Moral intuitionists and my readers are neither children nor
medieval peasants. They are modern educated adults. Modern educated adults
can know that moral beliefs are problematic in the ways that I outlined (at least
if they have read this far). So my readers and other modern adults need confirmation for moral beliefs, regardless of what you think about other people.
Moral Intuitionism
361
Moral intuitionists might seem to avoid this point if they claim only that
some moral believers are justified. However, moral intuitionists always include
themselves among those who are justified. Similarly, I assume that my readers
want to know whether they themselves are justified moral believers. If it turns
out that the only moral believers who are justified without confirmation are
children, medieval peasants, and others who are ignorant of the empirical
research in this paper, then it is not so great to be justified.
Defeasibility
Some moral intuitionists accuse me of forgetting that a moral believer can be
defeasibly justified without being adequately justified. Again, I plead innocent.
To say that a moral believer is defeasibly justified is to say that the believer
would be justified in the absence of any defeater. Defeaters come in two kinds.
An overriding defeater of a belief provides a reason to believe that the belief is
false. For example, if one newspaper predicts rain tomorrow, but a more
reliable newspaper predicts clear skies, then the latter prediction overrides the
former, even if I still have some reason to believe the former. In contrast, an
undermining defeater takes the force out of a reason without providing any
reason to believe the opposite. If I find out that the newspaper that predicts rain
based its prediction on a crystal ball, then this new information keeps the
prediction from making me justified in believing that it will rain, but the new
information does not make me justified in believing that it will not rain, since
a crystal ball is just as likely to lead to a true prediction. When my justification
is undermined completely in this way, I have no reason left for believing that it
will rain or that it will not rain.
The factors in Principles 1–5 cannot be overriding defeaters, since they do
not provide any reason to believe the moral belief is false. Even when moral
beliefs are partial, controversial, emotional, subject to illusion, and due to disreputable sources, that does not show that those beliefs are false. Thus, the factors in Principles 1–5 seem to be undermining defeaters. That suggests that we
have no reason to trust our spontaneous moral beliefs before confirmation.
Admittedly, some defeaters might not completely undermine a justification.
They might leave some weaker reason that makes believers partially justified.
However, the manifold underminers in Principles 1–5 add up, so that it is hard
to see why there is any reason left to hold spontaneous moral beliefs without
confirmation.
Moreover, I am not just talking about possible underminers. I argued in
section 3 that the underminers in Principles 1–5 actually exist for many moral
362
Walter Sinnott-Armstrong
beliefs. Actual moral believers are partial and emotional. They actually do disagree
often. Cultures actually are disreputable in ways that affect moral beliefs. There
is even empirical evidence for actual widespread illusions in morality.
Moral intuitionists can still say that spontaneous moral believers are prima
facie justified if that means only that they would be adequately justified if their
moral beliefs were not undermined by the factors in Principles 1–5. This counterfactual claim is compatible with their actually not being justified at all, but
only appearing to be justified. They might have no real reason for belief but
only the misleading appearance of a reason (as with the newspaper’s rain
prediction based on a crystal ball). In contrast, to call a believer pro tanto
justified is to indicate some actual positive epistemic force that is not cancelled
or undermined even if it is overridden. If the factors in Principles 1–5 are
underminers, as I argued, then spontaneous moral believers are not even pro
tanto justified. At most they misleadingly appear to be justified when they are
not really justified at all.
Besides, even if moral intuitions were pro tanto justified independently of
any inferential ability, this status would not make them adequately justified. As
I said, skeptics win if no moral belief is adequately justified. So moral intuitionists cannot rest easy with the claim that moral intuitions are merely pro
tanto justified.
Some
Many opponents object that, even if Principles 1–5 apply to some moral
beliefs, they do not apply to all moral beliefs. As I admitted, some moral beliefs
are not controversial. For example, almost everyone (except moral nihilists)
agrees that it is morally wrong to push the fat man in front of the trolley just
because you are angry with him for beating you in a game. Such cases also do
not seem due to context, heuristics, overgeneralization, or framing effects. Still,
such moral believers are partial and emotional (as Greene’s experiments
suggested). So Principles 1 and 3 do seem to create a need for confirmation
even in such clear cases.
Furthermore, if very many moral beliefs need confirmation, the others
cannot be immune from this need. To see why, compare a country with lots of
barn façades that look just like real barns when viewed from the road
(Goldman 1976). If someone looks only from the road, then he is not justified
in believing that what he sees is a real barn, at least if he should know about the
barn façades. The barn façades are analogous to situations that produce
distorted moral beliefs. Since such distortions are so common, morality is a
land of fake barns. In such areas, confirmation is needed for each justified
Moral Intuitionism
363
belief, even for those beliefs formed in front of real barns. Analogously,
confirmation is needed for each spontaneous moral belief, even when the
common distorting factors are absent. We need to get off the road and look
closer. At least when we should know that moral beliefs in general are so often
subject to distortion, we cannot be justified in trusting any moral belief until
we confirm that it is an exception to the rule that most moral beliefs are
problematic. So moral intuitionists cannot claim that any moral believers are
justified without confirmation.
This point can be presented as a dilemma: If a moral believer is an educated
modern adult, then she should know that many moral beliefs are problematic
in the ways indicated by Principles 1–5. She either knows or does not know
that her moral belief is an exception to the trend. If she does not know this, she
should accept a significant probability that her belief is problematic. Then she
cannot be justified without confirmation. Alternatively, if she does know that
her moral belief is exceptionally reliable, then she has enough information to
draw an inference like this: My moral belief is exceptionally reliable. Exceptionally
reliable beliefs are probably true. Therefore, my belief is (probably) true. If this
moral believer does not have the information in these premises, then it is hard
to see why we should call her justified. So, either way, moral intuitionism fails.
Skepticism
A common objection is that my argument leads to general skepticism, since
every inference has premises, so the demand for an inference cannot always be
met. However, my argument does not generalize so easily. If my belief that a
pen is in front of me is not subject to disagreement or illusions and has no
disreputable sources, and if I am neither partial nor emotional about pens, then
I might be justified in holding that non-moral belief without being able to
support it with any inference. Thus, my argument against moral intuitionism
does not lead to general skepticism.
My argument still might seem to lead to moral skepticism. If so, and if moral
skepticism is unacceptable, then something must be wrong with my argument.
However, my argument does not by itself lead to moral skepticism. My thesis is
not that spontaneous moral beliefs are not justified, but only that they are not
justified non-inferentially because they need confirmation. Such confirmation
still might be possible somehow. Even if moral intuitionism is rejected, there are
other non-skeptical methods in moral epistemology, including coherentism,
contractarianism, contractualism, contextualism, and naturalism (SinnottArmstrong 1996: 31–41). Moral skepticism arises only after all of these other
approaches fall. So my argument does not by itself support moral skepticism.
364
Walter Sinnott-Armstrong
Besides, even if these other approaches also fail, so my argument plays a role
in a larger argument for moral skepticism, that does not show that anything is
wrong with my argument, unless one assumes that moral skepticism is unacceptable. Why assume that? I accept a limited Pyrrhonian version of moral
skepticism. So I, at least, will not be dismayed if my argument takes one step in
that direction.²⁵
Anyway, my goal here has not been to argue for moral skepticism. My goal
has been to argue against moral intuitionism. More generally, I tried to show
one way in which empirical research in psychology and brain science might be
relevant to normative moral epistemology. If I succeeded in that enterprise, I
will happily leave moral skepticism for another occasion.
References
Baron, J. (1994). ‘Nonconsequentialist Decisions’. Behavioral and Brain Sciences, 17:
1–42.
Foot, Philippa (1967). ‘The Problem of Abortion and the Doctrine of Double
Effect’. Oxford Review, 5: 5–15.
Goldman, A. (1976). ‘Discrimination and Perceptual Knowledge’. Journal of
Philosophy, 73: 771–91.
Greene, J., et al. (2001). ‘An fMRI Investigation of Emotional Engagement in Moral
Judgment’. Science, 293: 2105–8.
—— and Haidt, J. (2002). ‘How (and Where) does Moral Judgment Work?’ Trends
in Cognitive Science, 6: 517–23.
—— et al. (2004). ‘The Neural Bases of Cognitive Conflict and Control in Moral
Judgment.’ Neuron, 44: 389–400.
Haidt, J. (2001). ‘The Emotional Dog and its Rational Tail: A Social Intuitionist
Approach to Moral Judgment’. Psychological Review, 108: 814–34.
—— et al. (1993). ‘Affect, Culture, and Morality, or is it Wrong to Eat Your Dog?’
Journal of Personality and Social Psychology, 65: 613–28.
Harman, G. (1977). The Nature of Morality. New York: Oxford University Press.
Horowitz, T. (1998). ‘Philosophical Intuitions and Psychological Theory’. In
M. DePaul and W. Ramsey (eds.), Rethinking Intuition: The Psychology of Intuition
and its Role in Philosophical Inquiry. Lanham, Md.: Rowman & Littlefield.
²⁵ See Sinnott–Armstrong 2006. My version of Pyrrhonism denies that any particular contrast class is the
relevant one in the sense that believers need to rule out all alternatives in that class in order to be justified
without qualification. That view might seem to conflict with my claim here that moral beliefs need inferential confirmation. However, there might be no need to rule out any particular contrast class, even if there is
a need to give evidence of a certain kind.
Moral Intuitionism
365
Kahneman, D., Slovic, P., and Tversky, A. (eds.) (1982). Judgment under
Uncertainty: Heuristics and Biases. Cambridge: Cambridge University Press.
—— and Tversky, A. (1979). ‘Prospect Theory: An Analysis of Decision under Risk’.
Econometrica, 47/2: 263–92.
Lackey, D. (1986). ‘Taking Risk Seriously’. Journal of Philosophy, 83: 633–40.
Moll, J., et al. (2001). ‘Frontopolar and Anterior Temporal Cortex Activation in a
Moral Judgment Task: Preliminary Functional MRI results in Normal Subjects’.
Arq. Neuropsiquiatr, 59: 657–64.
—— et al. (2002a). ‘Functional Networks in Emotional Moral and Nonmoral Social
Judgments’. Neuroimage, 16: 696–703.
—— et al. (2002b). ‘The Neural Correlates of Moral Sensitivity: A Functional
Magnetic Resonance Imaging Investigation of Basic and Moral Emotions’. Journal
of Neuroscience, 22: 2730–6.
Moore, G. E. (1903). Principia Ethica. Cambridge: Cambridge University Press.
Nietzsche, F. (1966). Genealogy of Morals. In Basic Writings of Nietzsche, ed. and trans.
W. Kaufmann. New York: Random House.
Parfit, D. (1984). Reasons and Persons. Oxford: Clarendon Press.
Quinn, W. (1993). ‘Actions, Intentions, and Consequences: The Doctrine of Doing
and Allowing’, repr. in Morality and Action. New York: Cambridge University Press.
Ross, W. D. (1930). The Right and the Good. Oxford: Oxford University Press.
Scheffler, S. (1982). The Rejection of Consequentialism. Oxford: Clarendon Press.
Sinnott-Armstrong, W. (1996). ‘Moral Skepticism and Justification’. In W.
Sinnott-Armstrong and M. Timmons (eds.), Moral Knowledge? New York: Oxford
University Press.
—— (2000). ‘From “Is” to “Ought” in Moral Epistemology’. Argumentation, 14:
159–74.
—— (2002a). ‘Moral Skepticism’, in Edward N. Zalta (ed.), The Stanford
Encyclopedia of Philosophy (Summer 2002 edn.), http://plato.stanford.edu/
archives/sum2002/entries/skepticism-moral/.
—— (2002b). ‘Moral Relativity and Intuitionism’. Philosophical Issues, Realism and
Relativism, 12: 305–28.
—— (2006). Moral Skepticism. New York: Oxford University Press.
Snare, F. (1980). ‘The Diversity of Morals’. Mind, 89: 353–69.
Unger, P. (1996). Living High and Letting Die. New York: Oxford University Press.
van Roojen, M. (1999). ‘Reflective Moral Equilibrium and Psychological Theory’.
Ethics, 109: 846–57.
Wheatley, T., and Haidt, J. (2005). ‘The Wisdom of Repugnance: HypnoticallyInduced Disgust Makes Moral Judgments More Severe’ (unpublished manuscript,
University of Virginia).
This page intentionally left blank
16
Ethics Dehumanized
Panayot Butchvarov
1. Ethics and anthropology
It is too early to judge how twentieth-century philosophy ended, but its
beginning was remarkable. Both Moore’s Principia Ethica and Russell’s
Principles of Mathematics appeared in 1903, the first volume of Husserl’s
Logical Investigations in 1900–1, and four of William James’s major philosophical books in 1902–9. There was not a significant difference, except in style and
temperament, between Anglo-American and European philosophers. The
analytic/continental schism came much later. Both Russell and Husserl began
as mathematicians. Moore wrote in the preface of Principia that his ethics was
closest to Brentano’s. Russell studied and discussed Frege and Meinong in
detail. James was admired in Britain and in Europe, influenced Husserl and
Wittgenstein, and was the subject of articles by Moore and Russell.
The present occasion is devoted to Moore’s Principia and what his heirs in
analytic ethics accomplished. But we must not neglect the rest of its historical
context, the important, often illuminating, similarities Moore’s ethics bears to
the views not only of Brentano but also of his close associates Russell and
Wittgenstein, as well as of predecessors such as Plato and Kant and his immediate successors H. A. Prichard and W. D. Ross. The relation of analytic ethics
to Moore’s book rested largely on his thesis in chapter I that the property good
is indefinable. Yet Moore made clear he had no interest in what he called verbal
and the tradition calls nominal definitions. They are the business of lexicography, he wrote. But it is just such definitions that analytic philosophers
sought, sometimes calling them analyses. The most familiar example comes from
analytic epistemology, not ethics: the definitions of ‘S knows that p’ in the
1960s, 1970s, and early 1980s. They were not even lexicographic definitions,
which record lexical fact and are tested by empirical investigation of speech and
writing. Rather, they recorded impressions of lexical fact, and were tested by the
author’s ‘intuition’ of what would or would not be said in some hypothetical
368
Panayot Butchvarov
situation, called a ‘counterexample’ if it did not fit the intuition. The question
‘How do I know what one would say in that situation, given that I am not in
it?’ was usually ignored. For it could be answered properly only by appealing to
what I and, especially, others have said in relevant situations, and this would be
to appeal to lexical fact. Even the Oxford English Dictionary is valuable mainly
for the examples of usage it lists, not the definitions it distills from them.
The kind of definition Moore did seek was an account of the constitution of
the thing, res, that is defined. It was closer to what the tradition calls real definition, though it gave not the genus and differentia but the parts of the thing.
Such a definition can be called an analysis, in a sense reasonably similar to that
employed in chemistry. In later years analyses were offered mainly of facts and
propositions, which were taken to be nonlinguistic entities categorially different from those chemistry analyzes. Their analysis was intended to reveal logical
form and, for this reason, was called logical analysis. It was in such analyses that
analytic philosophy took root, beginning in 1905 with Russell’s theory of
definite descriptions and culminating in Moore’s claim two decades later, in ‘A
Defense of Common Sense,’ that he knew the proposition ‘This is a hand’ to be
true but did not know how to analyze it. In Principia, however, his example was
the definition of a horse and consisted of an anatomical inventory. Our example
might be the account of water as H2O. Moore in effect agreed with Kant that ‘in
matters of morality it is always real definitions that must be sought.’¹
Despite its inattention to what he meant by ‘definition,’ analytic ethics did
begin and develop in relation to Moore’s ethics, though by way of sustained disagreement, not agreement. Discussions of Principia seldom ventured beyond
chapter I, which alone was included in most anthologies. Usually ignored were
the crucial preface, where Moore explained what he meant by ‘intuition’ and
‘self-evidence’ and, thus, what anyone calling him an intuitionist and a foundationalist ought to mean. Also usually ignored were the beginning of chapter II,
where he explained what he meant by ‘natural’ and ‘nonnatural,’ thus what
anyone calling his ethics nonnaturalist ought to mean, and chapter V, where he
explained his theory of right on the basis of the theory in chapter I.
By ‘intuitions,’ Moore wrote, he meant self-evident propositions, and ‘nothing whatever as to the manner or origin of our cognition of them.’ And a selfevident proposition, he explained, is one that is evident but not by virtue of
inference from other propositions. He did not say what he meant by ‘evident,’
perhaps thinking it unnecessary. A proposition is evident, of course, if it is, or
can readily be, seen to be true, either literally or metaphorically. Therefore, it
¹ Immanuel Kant, Logic, trans. Robert S. Hartman and Wolfgang Schwartz (New York: Dover,
1974), 144.
Ethics Dehumanized
369
may also be said to be known, in the serious and traditional sense of ‘know.’
The noun ‘evidence,’ as used in court or in the lab, has a wider meaning, but
the same root. Moore used ‘self-evident’ for the propositions stating ‘what kind
of things ought to exist for their own sakes,’ that is, are intrinsically good.
As to the meaning he attached to calling something ‘natural,’ he wrote he
meant that the thing is in time. Thus a paradigm of a nonnatural thing would
be a number, an abstract entity. The mere fact that a thing falls outside the subject
matter of physics does not make it nonnatural. For example, irreducibly
mental states would be in time and thus natural. If we said that a natural thing
is one belonging in the province of the natural sciences, as Moore himself did
on occasion, we would need a noncircular account of what is meant by calling
a science natural, as he doubtless was aware and so did not offer this as his
definition of ‘natural.’ The fact is that the distinction between the natural and
the nonnatural did not play a central role in his book, though the phrase
‘naturalistic fallacy’ of course did. As Moore made clear in the also ignored
chapter IV, which was devoted to what he called metaphysical ethics, even ethical
theories concerned with the ‘supersensible’ committed the fallacy. The fallacy was
just that of confusing two things: the property good and some other property.
In chapter I Moore held that the property good is nonnatural and simple,
therefore (given his account of definition) indefinable, that almost all earlier
ethical theories had committed the naturalistic fallacy of confusing it with
another property, and that they could be refuted with the so-called open question argument, which in effect encouraged the reader to pay close attention to
the property such a theory confuses with the property good in order to see that
they are two properties, not one. But his contemporaries in the Society of
Apostles and the Bloomsbury Circle, who included Russell, Keynes, and
Virginia Woolf, found more important not these metaphilosophical generalities
but the substantive views, defended in chapter VI, that personal affection (love,
friendship) and aesthetic appreciation (contemplation of beauty, in art and in
natural objects, human and nonhuman) are the greatest goods. In contrast with
Kant’s position, it is they that for Moore were the Ideal. And it is they that
prompted Keynes to rate Moore higher even than Plato. That chapter, too, has
been ignored in analytic ethics, which has focused instead on the preliminary
discussions in chapter I, especially the objectivity of value it took Moore to be
defending there. But, in a recent book, Brian Hutchinson points out that
‘Moore never even entertained doubts about the objectivity of value.’
Hutchinson acknowledges that for us this may be ‘a mystery difficult to fathom,’
but wisely suggests that the mystery ‘is to be savored rather than solved.’²
² Brian Hutchinson, G. E. Moore’s Ethical Theory: Resistance and Reconciliation (New York: Cambridge
University Press, 2001).
370
Panayot Butchvarov
While the central tenet of Moore’s theory of good was that it is a simple,
indefinable, and nonnatural property, the central tenet of his theory of right
was that duty is the action that ‘will cause more good to exist in the Universe
than any possible alternative,’³ the action that ‘is the best thing to do,’ that
‘together with its consequences presents a greater sum of intrinsic value
than any possible alternatives,’ either because it ‘itself has greater intrinsic value
than any alternative’ or because ‘the balance of intrinsic value’ of its consequences
does, so that ‘more good or less evil will exist in the world’ if it is adopted
(Principia Ethica, 76–7). Of course, the action need not do so on a grand scale.
To think that it must, or even could, would be human conceit of cosmic
proportion. And ‘cause’ or ‘produce’ should be understood broadly in the sense
of ‘contribute,’ since the action might be the best thing to do because of its own
goodness or its organic relationships. Moore’s was an ideal utilitarianism,
which unlike Bentham’s, Mill’s, and Sidgwick’s presupposed a theory of good
that placed no limits on what items might enjoy intrinsic goodness, thus allowing
that some may be actions.
Moore’s theory of right may be called cosmological. It tells us that we ought
to do what would be best, all things in the universe considered. It accords with
Aquinas’s first principle of natural law: ‘Good is to be done and promoted, and
evil is to be avoided.’⁴ It does imply, as Moore noted, that justice is not to be
done if the heavens should fall—unless, he wryly added, ‘by the doing of justice the Universe gains more than it loses by the falling of the heavens’ (Principia
Ethica, 197). The ethical views of Russell and Wittgenstein, the other two
founders of analytic philosophy, were also nonnaturalist and cosmological.
But, with the exception of Prichard, a thinker of unsurpassed acuity, and Ross,
whose terminology and distinctions we still find indispensable, later AngloAmerican ethics diverged in both respects. They are related. If ethics is naturalistic, then it is not likely to be cosmological. And if it is cosmological, then it is
not likely to be naturalistic.
Naturalistic ethics is almost certainly ethics humanized, a sort of anthropology: it
is about humans, not cats or bats. So it is not cosmological. Not only does it ignore
the good of the universe, it ignores that of gods, angels, and extraterrestrials, if there are
any, and usually also that of rivers, plants, and even nonhuman animals. Thus it
lacks the supreme generality and abstraction distinctive of philosophy and probably alone justifying its existence alongside the other cognitive disciplines. A
cosmological ethics can be expected, of course, to have application to humans, just
as chemistry and mathematics do. But this makes none of them about humans.
³ G. E. Moore, Principia Ethica, 2nd edn. (Cambridge: Cambridge University Press, 1993), 198.
⁴ Thomas Aquinas, Summa Theologica, Part Two, Question 94, Article 2.
Ethics Dehumanized
371
To be sure, we all feel what Cora Diamond calls the ‘heart-breaking specialness’
of the human.⁵ We all are human, enormously interested in ourselves and in
other humans, especially those we love or hate. It seems unfriendly, indecent,
inhumane to suggest that our ethics should be dehumanized, that it should not
be about us. Many demand that even space research be funded only if it leads
to cures for our diseases. But we are not the center of the universe, much as we
crave center stage. As Russell insisted, man does not have ‘the cosmic significance assigned to him by traditional philosophers.’ Man only has cosmic vanity.
To think that philosophy should be about us is like thinking that astronomy
should be about us. If this was not evident in the past, perhaps the reason was
the belief that, though God did see to it that all things he created were good, he
created only man in his own image.
Of course, there is a special, deep, and often misunderstood sense in which
humans may be cosmically central, namely, that leading to views such as Kant’s
transcendental idealism and its recent versions in Goodman and Putnam. They
rest on the virtual tautology that how we perceive and understand (conceive of )
the world, and thus the world itself as perceived and understood, depend on
our faculties of perception and understanding (conception). It does not follow
that there is nothing else. As Kant remarked, we can at least think of things
in themselves, for the notion of such things is not self-contradictory.⁶ If we
denied that we can, we would be committing ourselves to a peculiar sort of
epistemic creationism. Nevertheless, in that special sense, Kant, Putnam, and
Goodman may be said to have humanized even astronomy. But they did not
hold that astronomy is about humans. Although for Kant ‘the ultimate end of
the pure use of our reason’ was ethical, he resolved to ‘[keep] as close as possible to
the transcendental and [to set] aside entirely what might . . . be psychological,
i.e., empirical’ (Critique of Pure Reason, A 797/825–A 801/B 829), since ‘the
metaphysic of ethics is really the pure morality, which is not grounded on any
anthropology’ (A 841/42–B 869–70).
Ethics humanized thus is unphilosophical. It also lacks competence. Quine,
who took up the case for epistemology naturalized, in effect epistemology
humanized, often mentioned the role in cognition of ‘surface irritations’ but
wisely left the study of those irritations to neurology. In both epistemology
humanized and ethics humanized, we would be frivolous to compete with the
sciences specializing in humans, their cognitive functions and capacities, or
⁵ Cora Diamond, The Realistic Spirit: Wittgenstein, Philosophy and the Mind (Cambridge, Mass.: MIT
Press, 1991), 352.
⁶ Immanuel Kant, Critique of Pure Reason, trans. Paul Guyer and Allen W. Wood (Cambridge:
Cambridge University Press, 1998), B xxviii.
372
Panayot Butchvarov
their ways of acting and well-being.⁷ If humans are natural objects, a species of
animal, we can hardly expect to have special philosophical knowledge of them,
just as we can hardly expect to have special philosophical knowledge of stars or
bats. Accounts of human well-being and searches for the best explanation of
human conduct do not belong in philosophy departments, just as accounts
of human anatomy and human evolution do not. My point does not depend
on a narrow use of the words ‘natural’ and ‘science.’ If mental states are not
reducible to physical states, there could still be a natural science of them, in
Moore’s sense of ‘natural’ and the traditional sense of ‘science’ in which history
and political geography are social sciences. In fact there was such a science in
Moore’s time, namely, the largely introspective psychology of James, Wundt,
and Titchener. My point does depend, however, on taking competence seriously,
whether in forensic pathology and medieval history—or in ethics and
epistemology. Genuine competence requires serious training, for example, in
chemical analysis or parsing Latin. Nothing analogous with respect to humans
occurs in philosophy seminar rooms. If employed in hospitals, medical ethicists
are expected to learn some medicine. The reason is that their concern is properly
with humans, and it is naturalistic in Quine’s sense of being continuous with
natural science. They are often invaluable, not because they know something physicians do not, but because they are Socratic—they ask questions physicians do not.
We need not go to hospitals for examples. How to achieve happiness, in the
ordinary sense, recognized by both Kant and Mill, of enjoyment or satisfaction
of our needs and desires, has been a stock question in ethics, with Epicurus and
even Plato offering much advice; but arguably the invention of aspirin and
contraceptives, tractors and pesticides, air conditioning and spreadsheets,
answered it better. This is especially evident in politics. In Buddhist ethics,
sadly but realistically, suffering seems the primary concern, not pleasure, as in
Western ethics. Indian Benthamites hoping to learn from Americans how to
reduce suffering presumably go to American colleges of agriculture and schools
of public health, not to American philosophy departments.
One may ask, indignantly, what about loftier goods, not Bentham’s perhaps
but certainly Plato’s and Kant’s, such as justice, authenticity, salvation?
Especially in India, a deeply religious country, they are often thought far more
important. But these loftier goods call for nonanthropological, nonzoological,
considerations. Of course, philosophers who avow allegiance to naturalistic
ethics do write about some of them, at least about justice, not about gustatory
delights. Do they think they have access to human nature that zoologists lack?
⁷ Cf. my ‘Epistemology Dehumanized,’ in Quentin Smith (ed.), Epistemology: New Essays (Oxford:
Oxford University Press, forthcoming).
Ethics Dehumanized
373
Of course, they do not. Long ago, they took the conceptual turn. They
adopted a view of ethics far removed from both naturalism and nonnaturalism:
ethics as a ‘conceptual,’ not ‘factual,’ discipline. This allowed them to avoid
both commitment to nonnatural facts and responsibility for competence
about natural facts. Such ethics may be called analytic in the metaphorical
sense in which we do speak of analyzing concepts. And it may be called ‘naturalistic’ in the secondhand sense that it analyzes ethical concepts by referring
only to properties it deems ‘natural’ or at least to properties ‘supervenient’ on
such properties (see below, section 3). Conceptual analysis was the descendant
of the seventeenth-century ‘new way of ideas,’ which philosophers took in
search for a place not already occupied by Copernicus and Galileo. They chose
as their subject matter the human ‘mind’ and its ‘ideas,’ at that time unexplored
by other disciplines, and wrote books such as ‘An Essay Concerning Human
Understanding,’ ‘A Treatise Concerning the Principles of Human Knowledge,’
and ‘A Treatise of Human Nature.’ But if concepts are in nature—presumably
in human languages or human brains—conceptual analysis remains a part of
anthropology, in the broad but literal sense of this word. Concepts, so
understood, also lie outside philosophers’ competence today: there are linguistics and the lexicography that through painstaking research produces
dictionaries like the OED, and there are the rapidly growing brain sciences.
Philosophers have no more special competence in human languages or brains
than earlier they had in human minds or the solar system. On the other hand,
if concepts are not in nature, then Moore’s venture into the nonnatural at least
was straightforward.
It also, unlike conceptual analysis, was not dated, though its critics relish
calling it ‘obsolete.’ Like the seventeenth-century way of ideas, conceptual
analysis has been out of date since 1787 when Kant pointed out that our business is not merely to analyze concepts but to extend our knowledge (Critique
of Pure Reason, B 18). It has been out of date since 1951 when Quine pointed
out that ‘meaning is what essence becomes when it is divorced from the object
of reference and wedded to the word.’⁸ Like the seventeenth-century way of
ideas, the twentieth-century ‘way of concepts’ inherited its rationale from the
medievals’ concern with essences, but left out the grounding of essences in
things, which the medievals had taken for granted. Without such grounding
the rationale is opaque, even if we say, in often unwitting imitation of
Wittgenstein, that our concern is with how language or discourse ‘works.’
What special qualifications do philosophers have for research in the workings
of language? To be sure, because of their interests, sometimes they do have a
⁸ W. V. O. Quine, ‘Two Dogmas of Empiricism,’ Philosophical Review, 60 (1951): 20–43.
374
Panayot Butchvarov
better ear than professional lexicographers for the nuances of some segments of
speech. J. L. Austin’s work half a century ago is an example. But, as Austin vigorously argued, having such an ear is not a substitute for empirical knowledge.
Research in concepts or meanings, or in the ‘workings of our language,’ calls
for the competence of linguists, philologists, and lexicographers, today also of
neurologists, even computer scientists. It requires meticulous empirical
descriptions and fruitful, empirically verifiable hypotheses, not definitions or
‘iff ’ statements. If concepts are brain states, to focus on definitions of them
would be alien to neurology, just as such a focus would be to the other sciences.
That water is H2O was a discovery of chemistry, not a definition, lexical or
stipulative, and it concerned a substance, not a concept. And if concepts are
meanings or uses of words, to attempt to capture them in definitions would be
alien even to current philosophy of language, owing to the three trail-blazing
developments in it half a century ago. The first was Quine’s already mentioned
attack on appeals to meanings. It is widely accepted today, but usually only pro
forma. Phrases such as ‘conceptual question,’ ‘conceptual content,’ and ‘conceptual
connection’ still abound in the literature. The second development, also widely
accepted just pro forma, was Wittgenstein’s relentless argument in the
Philosophical Investigations, posthumously published shortly after Quine’s
article, that words are not used in accordance with necessary and sufficient
conditions. He gave ‘game’ as an example, but the argument applies also to ‘good,’
‘right,’ ‘reason,’ ‘know,’ ‘exist,’ and other denizens of the philosopher’s lexicon,
which, like ‘game,’ are everyday words, not technical terms introduced as abbreviations of multi-clause descriptions. The third development was Chomsky’s linguistics, announced four years later. It marked a striking advance by stressing the
biological, largely inherited, core of linguistic competence and urging the use in
the study of language of the standard methods of scientific research.
The project of defining knowledge, which I gave as an example of conceptual
analysis, was already dated at its birth in the late 1950s, when Ayer’s Problem of
Knowledge and Chisholm’s Perceiving appeared. A paper by Edmund Gettier, a
student of Wittgenstein’s disciple Norman Malcolm, made this evident in the
early 1960s. Few of those who wrote the thousands of pages devoted to it seemed
aware that, whatever its author’s intentions, the paper called not for greater
diligence, sophistication, or imagination in pursuing the project but for its abandonment. Thirty years earlier Wittgenstein had written: ‘If I was asked what
knowledge is, I would list items of knowledge and add “and suchlike.” There is
no common element to be found in all of them, because there isn’t one.’⁹
⁹ MS 302, ‘Diktat für Schlick’ 1931–3. Quoted by David Stern, ‘Sociology of Science, Rule Following
and Forms of Life,’ in M. Heidelberger and F. Stadler (eds.) History of Philosophy and Science (Dordrecht:
Kluwer, 2002), 347.
Ethics Dehumanized
375
Like post-Gettier analytic epistemology, post-Moorean analytic ethics was
unfazed by misgivings such as Kant’s, Quine’s, and Wittgenstein’s. It clung to
conceptual analysis, and thus, despite its professions, was really neither naturalistic nor nonnaturalistic. It went through several stages. The first began in
Vienna, soon after the publication of Wittgenstein’s Tractatus LogicoPhilosophicus and with some personal involvement by him. Ethical statements
were rejected as nonsense, or at least as lacking cognitive sense. The subtlety of
Wittgenstein’s verbally similar position, however, was missed altogether. The
second was to offer a positive characterization: they are expressions of emotion.
But the rich literature already in existence on the emotions in psychology (from
James to Arnold) and in phenomenology (from Meinong to Sartre) was
ignored, though it seemed to show that they are not, as the emotivists thought,
self-contained subjective episodes, Humean ‘impressions of reflexion,’ but
intentional states, directed upon objects, with character dependent on that of
their objects, and thus in principle cognitive. The third stage, probably
motivated by the experience of the Second World War, which made both the
outright rejection and the emotivist interpretation of moral statements seem
jejune, was to suppose that they express a special ‘moral point of view,’ something psychologically as genuine as emotion but less subjective, and that their
function is to guide, not goad. In effect, it was to deny them a full-fledged,
unqualified cognitive status, yet concede that their function is not merely
imperative or exclamatory. Taken for granted in all three stages was that the job
of ethics is to describe the meanings or uses of moral words, or the content of
moral concepts, or the features and workings of moral discourse. We cannot
give people what really interests them, namely, an ethics that says what they
should do, Moore’s heirs held, but we can give them an ethics that says what
they mean—a ‘metaethics.’ This was the message even in the more recent
fallback positions of projectionist antirealism and supervenience realism,
where the focus remained metaethical, not substantive. Few worried that the
very idea of telling people what they mean seemed paradoxical—except
perhaps to psychoanalysts.
2. The good and the world
By taking the conceptual turn analytic ethics did not provide a genuine
alternative to ethics humanized. Its new, ‘conceptual,’ subject matter was either
specious, or genuine but beyond its competence. The alternative provided by
Moore remained. Let us return to some of its details. I suggested that the place
in it of the indefinability and nonnaturalness of the property good was relatively
376
Panayot Butchvarov
minor, given his explanation of what he meant. Less familiar is that Moore
proposed a criterion, a test, for determining whether something has that property: ‘the method of isolation.’ It consisted in asking whether a world, a whole
world, which contains the thing but otherwise is just like a world that lacks it,
would be better (Principia Ethica, 135–6, 143–7, 236–8, 245–7). The two
worlds might be wholly inanimate and considered even, as Moore said, ‘apart
from any possible contemplation . . . by human beings.’ He applied the criterion to the intrinsic goodness of beauty, in opposition to Sidgwick’s contention
that nothing ‘appears to possess this quality of goodness out of relation to
human existence, or at least to some consciousness or feeling’ (133).
The method of isolation implies important similarities of Moore’s views to
Kant’s and Wittgenstein’s, which are seldom noticed. I am not suggesting historical connections, though as a student Moore did study Kant assiduously,
even attending a course on him in Germany, and Wittgenstein, who had read
Principia and heard Moore’s lectures at Cambridge (but liked neither—he
found them repetitious), went camping with him in Norway, presumably discussing not just logic and the Norwegian landscape. Similarities between major
philosophers are especially enlightening when their views are reached independently. Kant, Moore, and Wittgenstein shared a dehumanized conception of
ethics despite their fundamental differences in most other respects.¹⁰
In Moore’s case that conception was rooted in his method of isolation, in his
conviction that, contrary to Sidgwick, the focal good is that of the world, not
that of the human or sentient parts of it, and that it is independent even of
possible human consciousness. This conviction led to his ideal utilitarianism. It
also led to his principle of organic wholes: ‘the value of a whole may be different from the sum of the values of its parts’ (Principia Ethica, 40). For the
method of isolation suggests that the world itself is an organic whole. The goodness of a thing overall, we may say, is a function of (1) its intrinsic goodness,
determined by the method of isolation, (2) the intrinsic goodness of the totality
of its consequences, which are determined, insofar as this is possible, empirically, and (3) its noncausal contribution to the intrinsic goodness of the organic
wholes, including the world, of which it is a part, which also are determined,
insofar as this is possible, empirically. The overall goodness of a particular item,
whether an action or not, depends thus on the actual or possible goodness of
the whole world, the ‘universe.’
I believe that the usual objections to Moore’s theory of a nonnatural
property good are no more properly motivated or philosophically astute than
¹⁰ I discuss Wittgenstein’s view in detail in ‘Saying and Showing the Good,’ in Heather Dyke (ed.), Time
and Ethics (Dordrecht: Kluwer Academic Publishers, 2003).
Ethics Dehumanized
377
the objections to Plato’s theory of forms as a commitment to a ‘heaven’ or as
growth of a philosophical ‘beard.’ But more to the point here is that even if
good were a natural property, right would remain nonnatural if it is understood
in terms of good. For, so understood, it would still involve reference to all the
consequences of an action and all the organic wholes of which it is a part—to
all space and time, to the whole world. Scrupulous moral thought sets no
time or place beyond which it cares not what happens. Some Americans do care
about the floods in Bangladesh, and many people, wherever they may be, care
about the climate on earth a century from now. Authentic environmentalists
do not say that when humans become extinct, whales and prairie grass might
just as well. Many believe honesty would be owed to, and expected of, also
gods, angels, and extraterrestrials, should they exist. But these totalities of consequences and organic wholes, indeed the world itself, might not be natural
objects. They would not be natural in Moore’s precise sense if they are not in
time, even if they consist only of things that are in time. But they might not be
natural also in a larger sense. Wittgenstein held that, although sentences about
such totalities show what is higher, they say nothing.
At a meeting of the Apostles in 1912 Wittgenstein heard Moore’s paper ‘Is
Conversion Possible?’ which Moore had first read to them in 1900 while working on Principia. That Moore read the paper again suggests he had not abandoned its ideas. Moral conversion, he said, ‘is not unlike religious conversion,’
even though it ‘is not necessarily connected with any religious ideas.’ It is ‘both
a great good in itself and it secures all other goods which depend on one’s own
mind alone. . . . You see “life steadily and whole” and can feel neither desire nor
fear of what you see to be bad in it.’¹¹ We may note that in 1903, when
Principia was published, Bertrand Russell had written: ‘Man’s true freedom . . . [lies] in the determination to worship only the God created by our own
love of the good,’¹² and that in 1914, after (but probably not because of ) two
years of intense discussions with Wittgenstein, he attributed to ‘the ethical
work of Spinoza . . . the very highest significance,’ as ‘an indication of some new
way of feeling towards life and the world.’¹³ This new way of feeling, Russell
added, lay outside the scope of ‘the scientific method.’
At about the same time, Wittgenstein wrote in his Notebooks: ‘To believe in a
God means to understand the question about the meaning of life . . . to see that the
facts of the world are not the end of the matter . . . to see that life has a meaning.’¹⁴
¹¹
¹²
¹³
¹⁴
Tom Reagan, Bloomsbury’s Prophet (Philadelphia: Temple University Press, 1986), 144.
Bertrand Russell, ‘A Free Man’s Worship,’ in Mysticism and Logic (London: Allen & Unwin, 1917), 50.
Bertrand Russell, ‘Scientific Method in Philosophy,’ in Mysticism and Logic, 109.
Ludwig Wittgenstein, Notebooks, 1914–1916, trans. G. E. M. Anscombe (Oxford: Blackwell, 1961), 74.
378
Panayot Butchvarov
And later, in the Tractatus: ‘The sense [Sinn] of the world must lie outside the
world. In the world . . . no value exists. . . . If there is any value that does have
value, it must lie outside the whole sphere of what happens and is the case. For
all that happens and is the case is accidental’ (6.41).¹⁵ The sense (or meaning)
of the world is not something in the world because it is the sense of the whole
world. It constitutes ‘[t]he solution of the riddle of life in space and time,’ but
that solution ‘lies outside space and time’ (6.4312). Later, in 1929,
Wittgenstein explained: ‘What is good is also divine. Queer as it sounds, that
sums up my ethics. Only something supernatural can express the Supernatural.’¹⁶
And elsewhere, also in 1929: ‘[Attributions of ] absolute value are nonsensical
but their nonsensicality [is] their very essence . . . [A]ll I wanted to do with
[those attributions] was to go beyond the world and that is to say beyond
significant language.’¹⁷ He meant, however, not that the attributions are
gibberish, but only that they are not logical pictures, in the sense required by
his rather exacting theory of meaning.
What Wittgenstein called the riddle of life presumably concerns the sense or
meaning of life, and this has been a central topic in serious ethics. To ask about
ultimate value is to ask about the meaning of life, what makes life worth living.
And the meaning of life does involve the meaning of the world. One who asks
about the meaning of one’s life does sometimes phrase the question as asking
about ‘the sense of it all.’ Life can hardly be fully meaningful in a meaningless
world. Indeed, that the world exists at all, that there is something rather than nothing, may be the ultimate object of joy (or sorrow) and certainly of wonder (6.44).
But to ask about the sense of the world requires, as Wittgenstein put it,
‘view[ing] the world sub specie aeterni,’ even ‘feeling the world as a limited
whole,’ which, he wrote, is something ‘mystical’ (6.45). Ethics does ask what
makes life good, but ‘the good life is the world seen sub specie aeternitatis’
(Notebooks, 83e). Realization of value, whether goodness or rightness, does not
consist in the occurrence in the world of some particular event or events, but in the
world itself being different, at its limits, in its waxing and waning as a whole (6.43).
However, all this can only show itself. For there cannot be ethical propositions.
The reason is not that, as Wittgenstein’s early followers thought, there is nothing for such propositions to be about, but that what they purport to say cannot
be said, in the precise sense that it cannot be pictured, not even ‘logically.’ It is
‘the higher,’ which can only be shown (6.42).
¹⁵ Ludwig Wittgenstein, Tractatus Logico-Philosophicus, trans. D. F. Pears and B. F. McGuinness
(London: Routledge, 1972). References use the decimal system in the original.
¹⁶ Ludwig Wittgenstein, Culture and Value, ed. G. H. von Wright, trans. Peter Winch (Chicago:
University of Chicago Press, 1980), 3e.
¹⁷ Ludwig Wittgenstein, Philosophical Occasions 1912–1951 (Indianapolis: Hackett, 1993), 40, 44.
Ethics Dehumanized
379
Why is this so? Even if we refrain from calling the world mystical, we should
acknowledge that it is mysterious. The reason is logical, not mawkish or cabbalistic. It is not that the world is too big or too unlike what we take it to be.
Not its size or content, but its logical/ontological category, or rather its failing
to fall in any category, is what makes it mysterious. This is why genuine
propositions about it, and thus ethical propositions, are impossible. We may say
that the world is everything, but this would only acknowledge its peculiarity. For
to speak of everything is to employ the predicate ‘is a thing’ or ‘is a fact,’ depending on whether we think the world is the totality of things or of facts. Both predicates, Wittgenstein noted, express only formal concepts, corresponding to
formal or internal properties, and thus the sentences in which they occur say
nothing, though they show much. ‘[T]he variable name “x” is the proper sign for
the pseudo-concept object,’ he wrote. This is why ‘it is nonsensical to speak of
the total number of objects.’ And he added: ‘The same applies to the words “complex”, “fact”, “function”, “number”, etc. They all signify formal concepts and are
represented in conceptual notation [only] by variables . . .’(4.1272).
The concept of a world also is formal and thus unsuited for any ‘saying,’
though available for ‘showing,’ because it involves the formal concept of fact
(the world is the totality of facts). But it is formal for yet another reason: it
involves the formal concept of generality (the world is the totality of facts). An
important, especially relevant to this paper, application of the distinction
between saying and showing was Wittgenstein’s attempt to avoid both naive
realism and superficial antirealism with respect to the account of general
propositions. A general proposition does not assert an irreducibly general fact
that is ‘out there in the objective world,’ as Russell held in his Lectures on
Logical Atomism.¹⁸ But neither is it reducible to the conjunction of its singular
instances, as (following Frege) Russell showed in that same work.
Wittgenstein rejected both realism and antirealism regarding general
propositions by proposing a remarkably original and sophisticated third
alternative. In a 1919 letter to Russell, which replied to Russell’s objection that
in an account of a general proposition in terms of elementary (i.e. singular)
propositions ‘It is necessary also to be given the proposition that all elementary
propositions are given,’ he wrote: ‘There is no such proposition! That all
elementary propositions are given is shown by there being none having an elementary sense which is not given.’¹⁹ There is no such proposition, presumably,
¹⁸ Bertrand Russell, The Philosophy of Logical Atomism (Chicago: Open Court, 1996), 103.
¹⁹ Notebooks, 130. See also Letters to Russell, Keynes and Moore, ed. G. H. von Wright (Ithaca, NY: Cornell
University Press, 1974), 73. In Principia Mathematica (Part I, Summary), an elementary proposition is
defined as one that ‘contains no reference to any totality.’
380
Panayot Butchvarov
because ‘proposition,’ and thus also ‘elementary proposition,’ belong on
Wittgenstein’s list of words signifying formal concepts. To be sure, he wrote
that all propositions are truth-functions of elementary propositions (Tractatus,
5), implying that general propositions are truth-functions of their elementary
instances. He thus acknowledged what makes the antirealist position regarding
generality plausible. But he also wrote that the concept ‘all’ is not a truth-function
(5.521), thus acknowledging what makes the realist position plausible.
Presumably, his reason was that, as Frege and Russell had shown, a general
proposition is not a molecular proposition: its singular instances are not components of it. ‘It is surely clear,’ Frege wrote, ‘that when anyone uses the sentence “all men are mortal” he does not want to assert something about some
Chief Akpanya, of whom perhaps he has never heard.’²⁰ And Russell pointed
out that ‘in order to arrive [by “complete induction”] at the general proposition
“All men are mortal”, you must already have the general proposition “All men
are among those I have enumerated.” ’²¹ The quantifier ‘all’ does not gather its
singular instances in the way the paradigm truth functions, namely, the
propositional connectives, gather the propositions they connect. The horseshoe requires two propositions—an antecedent and a consequent—however
complex they may be. The quantifier requires only a propositional function.²²
The realism/antirealism debate, in metaphysics and logic as well as in ethics,
must be bypassed wherever it involves putative statements employing formal
concepts about the totality of things or of facts, that is, about the world. If
ethics involves such statements, as Wittgenstein held, then both moral realism
and moral antirealism are to be rejected. Ethical statements both say nothing
and show something. The controversy between moral realism and moral antirealism is thus a special case of the controversy between metaphysical realism
and metaphysical antirealism. Insofar as the latter concerns the world, namely,
its reality or nature apart from our cognition of it, neither alternative can be
stated properly. Thus we do not face the stark choice between them.²³
Antirealism is usually a negative position, merely denying the reality of whatever
items are in question, and today usually asserting that with respect to them
‘language is all there is.’ This is why it is deeply unsatisfactory, whether in metaphysics or in ethics. Wittgenstein’s distinction between saying and showing
²⁰ Peter Geach, and Max Black (eds.), Translations from the Philosophical Writings of Gottlob Frege
(Oxford: Blackwell, 1970), 83.
²¹ Bertrand Russell, The Philosophy of Logical Atomism (Chicago: Open Court, 1996), 101.
²² In the first edition of Principia Mathematica we find: ‘Our judgment that all men are mortal collects
together a number of elementary judgments. It is not, however, composed of these, since, e.g., the fact that
Socrates is mortal is not part of what we assert . . .’ (Cambridge: Cambridge University Press, 1964), 45.
²³ In Skepticism about the External World (New York: Oxford University Press, 1998), I argue for an
analogous conclusion in epistemology regarding the existence of bodies, of an ‘external world.’
Ethics Dehumanized
381
offered an alternative to antirealism that did not constitute a return to realism,
to the equally unsatisfactory acceptance of the items in question as unqualifiedly
‘out there.’ It is often dismissed as obscurantist. Tough-minded philosophers
ask, How can there be things that only show themselves? But the question
misses the point of the distinction. It is an alternative to both realism and
antirealism. What only shows itself is not part of reality. But neither is it unreal,
like Hamlet, the golden mountain, or the round square.²⁴
Contrary to received opinion, Wittgenstein’s distinction between saying
and showing is reasonably clear. What can be said is what can be pictured, at
least logically. What can only be shown is what the picturing involves but
cannot itself be pictured—not even by a ‘meta-picture.’ Consider the
statement ‘Socrates is white.’ It does say something. What it says can even be
pictured literally, not just logically. But the putative statement ‘Socrates is an
individual,’ in the sense ‘individual’ has in logic, says nothing. Yet it is not
gibberish. Nor is it ‘metalinguistic,’ i.e. asserting that ‘Socrates’ is an individual constant, just as ‘Socrates is white’ is not about ‘Socrates.’ ‘Socrates is
an individual’ shows something about the very individual it purports to be
about. It shows the logical category to which the individual belongs, and a
logical category is ontologically and cognitively much ‘higher’ than color.
‘Socrates is an individual’ presupposes what it purports to say, its having
sense depends on its being true. You can picture an individual’s being white
but not its being an individual—as is obvious in the case of literal pictures,
e.g. paintings. You can use color to picture a face, but you cannot picture the
color—you can put a splash of it on a canvas, but as Nelson Goodman
pointed out this would be a sample, not a picture, of the color, even if the
canvas is touted as a ‘painting.’
Wittgenstein’s distinction between saying and showing was thus a natural
consequence of his picture theory of meaning. And, again contrary to received
opinion, this theory is also reasonably clear, indeed familiar in the history of
philosophy. It was a descendant of the traditional theories according to which
meaning depends on the presence in the mind of ‘ideas,’ ‘representations,’
sometimes explicitly held to be ‘mental images,’ even ‘copies,’ of what is meant.
And the idea that a representation need not be literal is standard in present-day
accounts, neurological or computational, of cognitive states as involving representations. Wittgenstein’s picture theory of meaning is no more questionable
than such theories and accounts. Of course, it also is no less questionable, as in
his later works he himself argued.
²⁴ See my ‘Metaphysical Realism and Logical Nonrealism,’ in Richard Gale (ed.), Blackwell Guide to
Metaphysics (Oxford: Blackwell, 2002), 282–302.
382
Panayot Butchvarov
As we saw, Moore also held that ethics involves reference to the world as a
whole. For this reason and in this sense, he might have agreed with Wittgenstein
that, like logic, ethics is transcendental, that it concerns the limits of the world,
not its contents (Tractatus, 6.13, 6.421). Such a view of ethics was not novel.
For Plato the philosophic life culminated in a glimpse of the Form of the Good,
which he held to be indescribable. Aquinas placed Good in the company of
Being, One, Truth, and Beauty, the so-called transcendentals, which were said
to range across the categories, that is, the highest genera, and thus to lack even
the status of categories of things in the world, much less the status of things. In
philosophical theology God was described as a being of infinite goodness that
is the source and measure of all other goodness, earthly and unearthly. And
Kant, as if using words from the Tractatus, held that, unlike what he called
practical anthropology, moral thought is concerned not with what happens but
with what ought to happen, even if it never happens (Critique of Pure Reason, A
802/B 830). The ‘supersensible’ was as central to Kant’s ethics as the ‘nonnatural’ was to Moore’s and the ‘supernatural’ to Wittgenstein’s.
Kant gave ‘Thou shalt not lie’ as an example of an imperative of duty, but
promptly explained that it ‘does not apply to men only, as if other rational
beings had no need to observe it.’ For ‘the ground of obligation here must not
be sought in the nature of man or in the circumstances in which he is placed,’
Kant wrote, and he urged that ‘it is a matter of the utmost necessity to work out
for once a pure moral philosophy completely cleansed of everything that can
only be empirical and appropriate to anthropology.’²⁵ Kant’s distinction
between what happens and what ought to happen was in tune with
Wittgenstein’s distinction between saying and showing. If ethics is not about
what happens or is the case, yet truth is correspondence to what is the case, then
ethics contains no truths. Kant did not explicitly draw this conclusion, but
Wittgenstein did. The logical positivists also drew it, attacked Moore with it,
but misunderstood it.
Whether we ourselves should draw it depends on how circumspect we are in
wielding the notion of truth. That truth is correspondence (Übereinstimming,
‘agreement,’ in Kant’s terminology) to fact is a truism of common sense, but as
a philosophical theory it is too crude for ethics, as well as for logic and mathematics, indeed, for the reasons mentioned earlier, even for ordinary general
statements. Wittgenstein saw this, perhaps Kant also did, and Moore might
have seen it, at least in the case of statements predicating goodness overall, had
he considered the matter. To be sure, all three accepted the truism. But Kant
²⁵ Immanuel Kant, Groundwork of the Metaphysics of Morals, trans. H. J. Paton (New York: Harper &
Row, 1964), 57.
Ethics Dehumanized
383
called it a ‘mere verbal explanation’ (Logic, 55), a nominal definition
(Namenklärung) that proffers no criterion of truth (Critique of Pure Reason, A
58/B 72), and, paving the way to Hegel, he held that the proper and sufficient
criterion of empirical truth presupposes the idea of the systematic unity of
nature (ibid., A 651/B 679). There were no Russellian facts for Kant, or anything else not already epistemic and thus, by implication, alethic, to which
judgments might correspond if they were to be called true without circularity.
It is not our cognition that must conform (richten) to objects; rather, objects
must conform to our cognition (ibid., p. xvi). Wittgenstein also endorsed a
correspondence view of truth, but only for impossibly impoverished sentences
about configurations of simple objects of which no example could be given,
not for the sentences of logic, mathematics, ethics, or even everyday discourse.
Moore discussed the nature of truth extensively but inconclusively in lectures
delivered in 1910–11 and not published in his lifetime. He wrote that ‘to say
that a belief is true is to say that it corresponds to a fact,’²⁶ that this means that
‘the fact to which it refers is, or has being’ (ibid. 267), but also that the notion
of a fact itself can be understood as standing for what corresponds to true belief
(ibid. 298), and he acknowledged that the truth of conditional statements did
not seem to fit the account at all (ibid. 268). Moore, too, seemed confident
about the definition of truth as correspondence only if understood as merely
nominal.
The moral, however, is not that we should accept ordinary noncognitivism
or stampede into a coherentist or some other standard theory of truth, but that
we should recognize that the ways of knowledge and truth are not neat and tidy.
A doctor’s orders are neither true nor false, but their legitimacy and authority
are cognitive through-and-through. The reason, of course, is that their ground
is taken to be cognitive. Religious and theological thought often rests the
authority of the will of God on his omniscience. But a sophisticated theology
can also hold that it is grounded nonepistemically in God’s status as our creator.
It is in this latter way, presumably, that the authority of what Kant called
practical reason (Vernunft) and his description of ethical judgments as both
imperatives (Imperative) and cognitions (Erkenntisse) should be understood.²⁷
We may call ethical judgments ‘valid’ instead of ‘true,’ as Kant often did, just as
we may describe both the doctor’s orders and the propositions grounding them
as valid, though only the latter as true. Or we may follow Nelson Goodman and
just use ‘right’ for both. Goodman’s antirealism (‘irrealism’) closely resembled
²⁶ G. E. Moore, Some Main Problems of Philosophy (London: Allen & Unwin, 1953), 277.
²⁷ Immanuel Kant, Critique of Practical Reason, trans. Lewis White Beck (Indianapolis: Bobbs-Merrill,
1956), 68.
384
Panayot Butchvarov
Kant’s transcendental idealism. And Kant’s ethics cannot be detached from his
transcendental idealism, the first two sections of the Groundwork from the
third. If we cannot understand his transcendental idealism, then we cannot
understand the moral yet nonepistemic authority of a noumenal self, but we
should be clear that nothing merely human can enjoy such authority.
The autonomy Kant thought essential to morality required membership in the
intelligible, not just the sensible, world.²⁸
3. The right and the good
It is often asked whether a dehumanized ethics such as Moore’s could be ‘relevant’ to action. The question is as ancient as Aristotle’s complaints about Plato’s
Form of the Good. But it is ambiguous. It may be asking (1) whether Moore’s
property good can bear a relation to the rightness of actions, (2) whether one
can be motivated to action by it, (3) whether one can be motivated by cognitive states of which it is an object, by itself or as a constituent of states of affairs,
or (4) whether there can be such cognitive states in the first place.
The connection between good and right cannot be just happenstantial, but
neither can it be trivially definitional. In Principia Ethica Moore did define
duty as ‘that action, which will cause more good to exist in the Universe than
any possible alternative’ (198), but he also described what ‘is good in itself or
has intrinsic value’ as what ‘ought to exist for its own sake’ (34). In Ethics he
repeated that ‘it is always our duty to do what will have the best possible consequences,’ but denied that this is ‘a mere tautology.’²⁹ And in the preface to the
second edition of Principia he wrote that he had used ‘good’ in a sense that
bears an ‘extremely important relation to the conceptions of “right” and
“wrong” ’ (4). While Kant did not allow for a concept of moral good that is
independent of the concept of right, he did take for granted the independence
of the general concept of good. He held that ‘all imperatives . . . say that
something would be good to do or leave undone’ (Groundwork, 81), that ‘the
necessary object of a will which is determined by [the moral law],’ as ‘given to
it a priori,’ is ‘the highest good’ (Critique of Practical Reason, 4), and that ‘It is
the concepts of the good and evil which first determine an object for the will’
(ibid. 70). As John Rawls remarks, ‘the priority of right . . . does not mean that
²⁸ ‘The causality of [actions which can be done by disregarding all desires and incitements of sense] lies
in man as intelligence and in the laws of such effects and actions as accord with the principles of an intelligible world . . . [H]e is there his proper self only as an intelligence (while as a human being he is merely an
appearance of himself ) . . .’ (Groundwork, 125).
²⁹ G. E. Moore, Ethics (Oxford: Oxford University Press, 1912), 73.
Ethics Dehumanized
385
Kant’s moral doctrine includes no conception of the good, nor does it mean
that the conceptions used are somewhat deduced from a previously specified
concept of right. . . . No moral doctrine can do without one or more
conceptions of the good.’³⁰
The idea of being motivated to action by Moore’s property good indeed does
have doubtful coherence, but only because the idea of being motivated by any
property may be incoherent for purely logical, not ethical or psychological,
reasons. Dyadic relations do not hold between individuals and properties. But
it is not incoherent to ask whether one can be motivated by cognitive states
that have that property or states of affairs involving it as object. Would such
cognitive states be ‘reason-giving’? Would they be ‘action-guiding’? This
question is difficult to answer because the specialists in such matters, namely,
psychologists working on motivation, still know too little, probably because
they cannot engage in serious experimentation with human subjects, while
philosophers engage only in speculation. When Hume announced that reason
is and ought to be the slave of the passions, we may ask how he knew all this,
how he knew that it is their slave, whatever we think of his adding that it ought
be. Perhaps he could be excused for thinking that the ‘passions’ are discoverable
through introspection. But his own views should have kept him from thinking
that their being motives, i.e. their motivating, could be so discovered.
Nevertheless, there have been advances in genuine, scientific psychology. The
preoccupation with ‘primary drives’ seems to have ended. There is growing
recognition of the spontaneity of the young child’s artistic and play behavior,
the importance of surprise for its cognitive development, its preference for the
novel, the presence, even in nonhuman animals, of curiosity, a tendency to
explore, a desire to know for the sake of knowing. Progress has been made away
from the egoism and hedonism presupposed by most so-called ‘rationality’ theories.³¹ The genetic basis of motives like empathy is readily acknowledged.
Chomsky accepts Plato’s thesis in the Meno about the innateness of much of
our knowledge.
Can there be cognitive states of which the property good, or a state of affairs
that includes it as constituent, is the object? Those who give a negative answer
usually rely on causal or quasi-causal metaphysical and epistemological theories, accepted, if not because they seem ‘scientific,’ then because of thoughtexperiments about what we would or would not ‘say.’ They ask, ‘Would you say
that S knows (perceives, sees, is aware of ) x if x bears no relation to S?’ And then
they ignore the obvious answer, ‘No, I would not, but x does bear a relation to
³⁰ John Rawls, Lectures on the History of Moral Philosophy (Cambridge, Mass.: Harvard University Press,
2000), 231.
³¹ Cf. Philip Kitcher, ‘The Evolution of Human Altruism,’ Journal of Philosophy, 90 (1993): 497–516.
386
Panayot Butchvarov
S, that of being known by S,’ in favor of unobvious answers, as if the right
answer could not be the obvious one. They think they understand the nature
of causality better than they understand the nature of knowledge and would
deny even the cognitive status of arithmetic, a paradigm of unquestionable
knowledge, if it is taken to be about numbers. They would find a mathematician’s intellectual life as inexplicable as the mathematician’s moral life when
understood in Moore’s way, if arithmetic turned out to be about numbers.
Elsewhere I have suggested that the property good, described by Plato as
blinding, by Moore as nonnatural, and by Wittgenstein as unsayable, is best
considered a generic property, though one on the highest level of generality.³²
It is the genus to which Moore’s personal affection and aesthetic appreciation,
Aristotle’s eudaimonia, Mill’s pleasure, and other goods reasonably proposed
by reasonable people belong as species. This is why we cannot ‘see’ it in the way
we see a shade of yellow. But then neither can we see even Color in that way,
though Color is a generic property on an incomparably lower level of generality. We might say (though Wittgenstein did not) that Goodness only shows
itself in its species, just as Color only shows itself in yellow, red, blue, etc., and
is not seen as a separate property additional to them. We might also say that the
genus supervenes on its species, and thus provide the idea of supervenience
with content that avoids reliance on an otherwise puzzling relation such as
‘determination.’
Whether or not this suggestion is right, the theorists denying that we can
have cognitive access to Moore’s property good need to pay more attention to
the epistemological and metaphysical details. Their epistemology is open, of
course, to the familiar objection to externalism, namely, that it cannot answer
the skeptic, although finding such an answer was the raison d’être of modern
epistemology. But they also ignore, or require us to count as unintelligible,
Kantian accounts of cognition, which even if we do not accept we ought to be
able to understand and respect. According to Kant, Putnam, Goodman, and
many others, though as a fact in the world human cognition is a zoological
matter, and thus subject to the demands of scientific causal explanation, the
world itself is comprehensible only as an object of a cognition that is not
zoological.
The theorists also owe us answers to numerous metaphysical questions, such
as those explored in detail by David Armstrong³³ and Evan Fales.³⁴ The first, of
course, is ‘What is causation?’ They cannot just revert to Hume and in effect beg
the question against Moore by saying that causation is constant conjunction in
³² I do so in Skepticism in Ethics (Bloomington: Indiana University Press, 1989).
³³ David M. Armstrong, Universals and Scientific Realism (Cambridge: Cambridge University Press,
1978).
³⁴ Evan Fales, Causality and Universals (London: Routledge, 1990).
Ethics Dehumanized
387
time, even if ‘nomological.’ Is it then, as Armstrong and Fales hold, a relation
of universals, properties, or at least based on one? But are properties universals
in the first place, or are they rather particulars, tropes, or perhaps both, as
Moore in fact held?³⁵ How would the arguments against Moore read in the case
of each possibility? Fales argues that there must be properties we can identify
independently of their causal powers if a vicious infinite regress is to be
avoidable. His examples are the properties characterizing the content of sense
perception, though his ultimate concern is with the identification of properties
in physics (Causality and Universals, 221–4). But if some properties can be
identified or known without reference to their causal powers, so might Moore’s
property good, whether or not it has causal powers. If our theorists deny this on
the ground that it is nonnatural, then they must revisit the distinction between
natural and nonnatural properties and give a serious, detailed account of it. Is
a property nonnatural because it is not in time, as Moore held? What is it for a
property to be in time? Would they say, in a vicious circle, that to be in time is
to have causal powers? Moreover, since a property, natural or nonnatural,
enters in causal relations only indirectly, as a constituent of states of affairs, we
would need to be told a great deal about the nature of states of affairs and how
and what properties might be constituents of them. For example, can a state of
affairs that is in time and part of the causal order have a nonnatural property as
a constituent? The latter would not itself be in time and enter in causal
relations, but is not this the case with all properties?
Yet another question the theorists need to consider is whether there are
uninstantiated properties. If there are, do they have causal powers? Are moral
properties, even though real, uninstantiated? If they have no causal powers, is
this so just because they are uninstantiated? Being an angel perhaps has no
causal powers, but is this true of being a circle, as this property, though never
instantiated, must be understood at Michelin and Goodyear? Is Moore’s property good like that of being an angel or like that of being a circle, or is it rather
like the shape of my pen? And if the latter, is it also, like it, a specific property,
or a generic property, like Shape, or a transcendental and thus not even generic?
Perhaps generic properties as such have no causal powers but their species do.
A tire’s having shape helps no car roll but its being round does. A traffic light’s
being colored stops no driver but its being red sometimes does. Are Moore’s
ideal goods, personal affection and aesthetic appreciation, species of goodness,
as I suggested? Do they have causal powers? If they do but their genus, goodness, does not, is this so because the latter is a nonnatural property or just
because it is a generic one, like Shape or Color? Or are Moore’s ideal goods
³⁵ G. E. Moore, “Identity,” Proceedings of the Aristotelian Society, ns 1, (1900–1).
388
Panayot Butchvarov
uninstantiated properties, mere ideals? If so, can mere ideals—moral, political,
religious—have causal powers? If they cannot, how are we to understand our
striving and sometimes dying because of them?
Of course, I shall not attempt answers to these questions here, but answers are
needed, detailed and carefully worked out, if we are to take seriously the complaint that Moore’s property good is irrelevant to action. Appeals to ‘naturalism’
or ‘the scientific point of view,’ let alone to our ‘intuitions,’ are not enough.
Without such answers, the complaint might be like the seventeenth-century
natural philosophers’ complaint that Newton appealed to occult and immaterial gravitational forces, rather than to intelligible and robust bumping, or
H. A. Prichard’s complaint that Einstein’s theory of relativity was unintelligible.
How a body could ‘motivate’ another body at a distance without the help of
intervening bodies was incomprehensible to the natural philosophers, and
Prichard could not visualize a non-Euclidean space. Some of Moore’s heirs have
been unable to comprehend or visualize the property good as doing any
pushing or pulling.
So, I shall ignore the metaphysical and epistemological concerns behind the
usual questions about the relevance of Moore’s ethics. But a different question
does arise. It is both legitimate and deep. By requiring reference to the whole
world in judgments of duty, Moore could tell us nothing specific about how we
ought to act in any particular situation, just as Wittgenstein could not. This is
why Moore virtually admitted that his theory of right was profoundly skeptical.
There might be an action we ought to do, but we could not know which or what
it is. In view of the mind-boggling range of its consequences and organic relationships, ‘throughout an infinite future’ (Principia Ethica, 202), even probability
statements about them could not be seriously made. According to Wittgenstein,
we could not make genuine statements about them at all, since this would require
reference to totalities determined by formal properties. For both Moore and
Wittgenstein, radical moral skepticism seemed inevitable, though in Moore’s case
the reasons were empirical, while in Wittgenstein’s they were logical.
But Moore’s moral skepticism does not lead to amoralism. Ideal utilitarianism is not mere consequentialism. An action may be intrinsically good even if
it ought not to be done, even if it did not make the world better. In Ross’s terminology, if not meaning, it may be a prima facie duty even if not an actual
duty. This is why respect for the good, for Aquinas’s first principle of natural
law, may continue to inform the ideal utilitarian’s actions. Such respect would
be akin to love, whether practical or pathological, not to calculation. This love
can have as its object only the intrinsic goodness of the action, its being a prima
facie duty, not its being an actual duty. Only a part of a world, not a whole
world, can be loved.
Ethics Dehumanized
389
I am not suggesting an inference, surely specious, from the intrinsic goodness
of an action to its rightness. No claim is made that the former makes the latter
‘probable,’ even to a tiny degree, or that it ‘justifies’ or is a ‘reason’ for the
action. In Hutcheson’s useful terminology, if not meaning, it is at most an
‘exciting reason’—not a ‘justifying reason,’ i.e. an item one may appeal to in
reasoning, what Kant called Grund and Moore called evidence. What I am
suggesting is that if in acting one is motivated and guided only by respect for
the good, and only the intrinsic goodness of an action is intellectually visible,
then one is motivated and guided only by respect for the intrinsic goodness
of the action. One has no knowledge of the totality of its consequences and
organic relationships, indeed not even a genuine conception of it. Thus, qua
agent, the ideal utilitarian can only be a deontologist, not a consequentialist.
This is why Moore’s ideal utilitarianism was not inimical to moral common
sense, which views with distaste the spirit of calculation consequentialism cultivates. The ideal utilitarian has no ‘justifying reasons’ but plenty of ‘exciting
reasons’ for doing good particular actions without guile: their plain goodness.
Thus Moore’s dehumanized ethics may be seen as the marriage—of love, not
convenience—of the two great ways of moral thinking: the utilitarian and the
deontological.
This page intentionally left blank
INDEX
Note to readers
Because G. E. Moore and his works are discussed throughout the text, an entry for
Moore would not be useful and is not included. To find discussion of Moore and his
works readers should search for relevant subject entries, e.g., organic unity; goodness,
whether simple property; etc.
a priori moral knowledge 212, 217–18
actions
instrumental value of 29–32
intrinsic value of 29–30, 370
aesthetic value 80, 89–90
agency 31, 65–67, 71, 108–9, 323
relation to personal good 113–17,
124–27
agent neutral reasons 30
and personal good 127–28
agent relative reasons 31–32
Alston, W. 303
analytical ethical philosophy 18. See also
philosophical ethics
methodology of 19–20
antirealism 380–81
Aquinas, T. 370
Aristotle 84, 90
Armstrong, D. 386
assertions 302–5
compared to orders 304
expressive versus reportive 303–6
and intentions to assert 305–6
and judgments of truth 304
of value sentences 312–13
assertoric nondescriptivism. See cognitivist
expressivism
authority 35
of motivational states 31
availability heuristic 345, 353
Ayer, A. J. 234, 245, 255, 320, 323
Barker, S. 300, 308, 316, 317
Baron, J. 353
beliefs. See also, moral judgments
functional role of 266–67
is-commitment versus ought-commitment
270–72, 281–82
phenomenology of 263–65
semantic assessability of 265–66
whether descriptive 287–88
Blackburn, S. 163, 235, 236, 244 n. 12,
256, 257, 300
blameworthiness 34–35
Borg, J. S. 349
Boyd, R. 120
Brandom, R. 303 n. 5, 321 n. 10
Brandt, R. B. 126
Brink, D. 227
Broad, C. D. 97, 195, 200–2
Broome, J. 42 n. 4
buck passing. See goodness, buck passing
account of
Butchvarov, P. 96–97
Butler, J. 31
Carnap, R. 255
causation
causal power of normative properties 98
causal test of ontological credibility 225,
228–29
and moral facts 224–30
and normative facts generally 226–27
Chaiken, S. 353
Chen, S. 347
choiceworthy life 91, 95–96, 103
Chomsky, N. 374
Christian morality 356
392
Index
Church, A. 173
Clark, R. 252
cognitive goals 66–67, 69, 72 n. 15, 74
cognitivist expressivism 255–297 passim
competence of philosophy 372–75
concepts 323
normative versus natural 323
conceptual analysis 18, 373–75
consensus 24
consequentialism 21, 25, 31–34. See also,
utilitarianism
act 33–5
analytic 30
demandingness objection to 33
and over generalizing 353
and relativizing value 134
rule 21
self-defeating objection to 33
constitutive value. See inherent value
contextualism
concerning predications of ‘good’ 90–91
semantic 276
contractualism 33
contributory reasons 39–40
defined in terms of desires of rational self
54–56
defining overall ought 51–53
expressivist analysis of 56–59
and relation to overall ought 40–45,
50, 58
Copp, D. 150
Cornell realism 191
Crisp, R. 161–62
Dancy, J. 157–61
Darwall, S. 108 n. 5, 117 n. 20, 122 n. 29,
128 n. 37, 170
defending commitments
to evaluative sentences 314
generally 308–09
definitions
lexicographic 367
and logical analyses 368
nominal 367
real 368
demandingness 33
deontology 25, 50–51
descriptivism 92–93, 97–98. See also,
representational discourse
how related to cognitivism 256–57,
270–75
Diamond, C. 371
disagreement. See moral disagreement
disgust 352
doing versus allowing 354–55
Dworkin, R. 113 n. 14
Einstein, A. 182
embedability 235–36, 237–38, 266–67,
277–78, 315–16, 324
of proto-assertions 310
emotionally clouded judgment 344
emotivism 255, 375
empirical science
competence of 371–72
relevance to ethics 24, 339, 371–72
ends justify the means 133
enticing reasons 43–44, 51
epistemic confirmation 343, 345–46,
362–63
and disagreement 344
and emotionally clouded judgment 344
and illusion 344–45
and partiality 343–44
and suspect sources 345
epistemic defeaters 361–62
epistemic justification 341
and adequate justification 341
and causation 387
and externalism 359, 386
and inference 341–42, 358–59
of moral beliefs 346
in children 359–60
and disagreement 348–50
and emotionally clouded judgment
350–52
and illusions 352–56
and partiality 346–48
and suspect origins 356–57
prima facie justification compared to
pro tanto justification 362
unjustified versus not justified 360
error theory 273–75
essentialism of normative properties 98–100
Index
ethical non-naturalism 210–31, 233,
243–46
objections to 212–14
similarities to philosophy generally
215–18
ethics dehumanized 370
and competence of philosophy 371–73
evaluative incompleteness 93
experience 82–83, 94–95, 98–100
as internal 83
relational versus nonrelational 83,
88–89
expressivism 56–59, 73–74, 191, 205–6,
234–35, 240–47, 255–57, 299–300,
312. See also, proto-expressivism
and motivational internalism 240–41,
299
and naturalism 327
and reporting versus expressing
301–02
Fales, E. 386
Foot, P. 36, 192
formal concepts 379
formal logic
assigning truth values 280, 290–97
for is and ought commitments 278–80,
289–90
Foucault, M. 356
foundationalism 340
framing effects 353–56
Frankena, W. 28, 81–82, 172
games 374
Geach, P. 90, 235, 237–38, 246 n. 14,
251
genetic fallacy 357
Gibbard, A. 34, 205, 56–57, 170, 187, 206,
235, 241 n. 10, 255–57
Glanzberg, M. 239
Goldman, A. 362
Goodman, N. 371, 383
goodness. See also, intrinsic value; open
question argument; sortal
attributions
analyses of 159–62
buck passing account of 149–63
393
definability of 170, 175–78, 258,
319, 367
as empty property 246
as generic property 386
teleological account of 157–58
whether simple property 90–91, 96–97,
163, 175–76, 258
Greene, J. 350–52
habituation 114, 116
Haidt, J. 341, 342, 347, 350–51, 352, 353
Hampton, J. 40–41, 45
happiness 372. See also, personal good
Hare, R. M. 108 n. 3, 205, 206, 259, 283,
285 n. 17
Harman, G. 63, 224, 263–64, 356–57
hedonism 328, 333, 385
Hobbes, T. 347
Horgan, T. 276, 300, 303
Horowitz, T. 355–56
Horwich, P. 238 n. 5, 321 n. 10
hot cognitions. See moral judgments, as
motivationally hot
Hudson, W. D. 20 n.10
human nature 107–9, 123–25
philosophy’s study of 372–73
Hume, D. 262 n. 9
Humean motivation 240–41. See also,
motivational internalism
Hurka, T. 17
Hutcheson, F. 35
Hutchinson, B. 369
hyperplans 324
coinciding with properties 328
in natural terms 324–25
hypothetical requirement 117
illusion 344–45
in context 352–53
and heuristics and biases 345
overgeneralizing 353
imperialism of the moral 247–48
implicature 316–17
inherent value 84–106 passim
and fitting attitudes 87–90
intrinsic properties 200
compared to intrinsic value 200–4
394
Index
intrinsic value 20–23, 26–29, 32, 79–106,
253. See also, inherent value; organic
unity
of actions 29–30, 370
compared to goodness in itself 83–84
and consciousness of pleasure 80–82
epistemology of 100–2, 369
experiences as bearers of 82–84, 88–89,
91, 102–03
and isolation test 376
of objects 83–86
and ought to be 26–27
and reasons for action 91–92, 102–5,
159
intuitions
and definitions 368
and framing effects 354–55
as self-evident propositions 368
irreducibility of ethics 17–18
Jackson, F. 137–45, 147–48, 237
Kahneman, D. 345, 353
Kant, I. 371, 382
Kim, J. 327 n. 16
knowledge 374. See also, epistemic
justification
Korsgaard, C. 72 n. 13
Kripke, S. 173–75, 321 n. 10
Lemos, N. 94 n. 22
Lewis, C. I. 84–85
Lewis, D. 320
Lewy, C. 172–73
Mackie, J. L. 185, 244, 274, n. 14
Malcolm X 112–13, 115–18
Malvo, John Lee 121
Mandelbaum, M. 260, 268–69, 284
McDowell, J. 321 n. 10
meaning of life 378
meaning 320. See also, semantic
incompleteness
picture theory of 381
study of 373–74
metaethics
generally 319–20
related to normative ethics 19, 22–23,
32–37, 25, 319–20
metaphysical ethics 20–21, 196–99
being compared to existing 196–97
Mill, J. S. 34, 347
Millar, A. 321 n. 10
Moll, J. 351
Mooreanism 191–92
Moore’s methodological maxim 259
moral dilemmas 351
moral disagreement 218–24
compared to philosophical disagreement
generally 219–21, 223–24
as epistemic defeater 221–24
in ideal agents 219
and inference to the best explanation
218–19
and non-moral disagreement 344, 348–49
moral intuitionism 53, 340, 342–43
and inference 342–43
and moral epistemology 340–41
moral judgments 260, 267–70
and belief-independent assertions 282
as beliefs 263–67
content of 261–62
as motivationally hot 269, 271–72,
273–74, 284–85
phenomenology of 268, 272–73
semantic assessability of 275–77
moral methodology 282–83
moral nihilism 210
moral progress 285–86
moral sense theory 342
motivational internalism 145–47, 185–86,
214, 240–43, 269, 385
natural properties 150, 192–96, 199–204,
210–11, 245, 328, 368. See also,
normative properties
disciplinary conception of 211–12, 369
epistemic conception of 212
and isolation in time test 194–96, 369,
377, 387
and natural constitution of normative
statements 324, 326, 328, 331
and normative concepts 322–23
and response-dependence 249
naturalism 73, 96–98, 308. See also, ethical
non-naturalism; metaphysical
ethics
and ethics as a natural science 211–12
Index
and natural constitution 324–25
and naturalistic ethics 20, 370
naturalistic fallacy 20, 149–51, 172–73,
339, 369
applied to metaphysical ethics 198–99
and the paradox of analysis 173
Newton, I. 319
Nietzsche, F. 321, 356
noncognitivism 125–26, 255, 322, 383.
See also, expressivism
and the open question argument 151
nondescriptivism 227–28
nondescriptivist cognitivism. See cognitivist
expressivism
normative disjunctivism 97–8
normative properties
and explanations 328–29, 332–35
normative thoughts. See also, hyperplans
as contingency plans 323–24
normativity
as central ethical notion 26–32
generally 308–9, 321–322
moral compared with other kinds of
183–87
outside of ethics 321–22
of reasons 50–51
and truth value 234, 235–36, 238–39
objectivity 313–15, 369
and agreement 315
Occam’s razor 225
Ogden, C. K. 320
open question argument 30–31, 125–27,
149–52, 164–67, 169–70, 233–34,
369
and following rules 171–73, 175–88
and irrealism 170–72, 183
motivating expressivism 234–35
organic unity 21–22, 54–56, 58, 80, 323, 376
of inherent and intrinsic value 89–90,
93–95
of reasons 104–105
ought implies can 48
ought to be 26–30
versus ought to do 29, 32
Parfit, D. 185, 352 n. 20
partiality 343–44
Pascal, B. 114, 116
395
personal good 107–29 passim. See also,
agency, and choosing personal
goods
evil things as components of 121–22
and fitness relation 109–11, 123
and personal relationships 119–20
and talents 118
personal relationships 119–20, 123–24
philosophical ethics 25
philosophical method 215–16. See also,
moral methodology
and the a priori 216–17
physics 24. See also, empirical science
Plato 49
pluralism 22, 26
about goodness 156
generally 331
about reason giving properties 156–57
practical rationality 71–72
practical reasoning 46–50
prescriptivism 255, 259. See also,
expressivism
Price, H. 300
Prichard, H. A. 27 n. 20
prima facie duties 52–53
as parti-resultant attribute 52
prima facie goodness
related to intrinsic goodness 93–94
priority of good 20, 22, 29
projectionist antirealism 375
properties. See also, natural properties;
normative properties
and causal powers 387
and explanatory importance 332–35
extensional equivalence 329–30
individuation of 332
possible worlds analysis 325–26
proto-expressivism 205–7
Putnam, H. 328 n. 17, 330 n. 19, 371
quasi-cognitivist expressivism 287–88
quasi-realism 257
queerness 170–71, 179–81, 183–86,
244
Quine, W. O. 242, 354, 373
Rabinowicz, W. 165 n. 6
Railton, P. 110 n. 8, 170
rational advisor 53–54
396
Index
rational beings 63–65
and epistemic limitations 64
rational intuitionism 342, 368
rationalism 214–15
rationality. See also, rational beings; practical
rationality
and decision making 144–48
justificatory and critical force of 77–78
and machines 65–66
qua agent 65–67, 71
recognitional view of 73–74
relativized to one’s purposes 71–76
theoretical rationality as a virtue 62
Rawls, J. 23, 384–85
Raz, J. 108 n. 5
realism 96–98, 209–31 passim 256, 322,
380
reasons 286–87. See also, contributory
reasons; enticing reasons; normativity
enablers and disablers 47–50
as favorers 45–50
intensifiers and diminishers 49–50
welfarist analysis of 161–62
reasons internalism 34–35
reduction 73, 199
of evaluative terms 92–93, 259, 288, 309
reliabilism 342, 359
representational discourse 300–2
and truth 302
respect for persons 27
revisionary ethics 299–300
Richards, I. A. 320
right action
as action-guiding 139, 144–45, 384
cosmological conception of 370
grounding criticism 141–42, 144–45
as maximizing expected utility 138–45
as maximizing value 135–36, 137, 370,
384
as non-natural property 377
skepticism about 388
risk aversion 354
Rønnow-Rasmussen, T. 165 n. 6
Ross, W. D. 31, 41, 50–52, 93–94, 159–60,
331, 341 n. 6
rule following 178–82. See also, open
question argument
compared to family resemblance
174–75
compared to private language argument
173–74
Russell, B. 320, 368, 371, 377, 379–80
saying versus showing 378–81
compared to transcendental ethics 382
Scanlon, T. M. 28 n. 25, 91 n. 18, 149,
151–67 passim
self-evidence 368–69
Sellars, W. 321
semantic incompleteness
and determinables 251–52
and metaphor 252–53
of normative sentences 246, 250–53
Sen, A. 107 n. 2
Sidgwick, H. 17–19, 26, 319, 376
simple property. See also, goodness, whether
simple property
compared to non-natural property
193–94
yellow as 193
skepticism 334, 340, 363–64
and regress argument 340, 368
Skorupski, J. 34
Smith, M. 53–54, 126, 127
Sober, E. 329, n. 18
social intuitionism 350
socratic method 372
sortal attributions 246–47, 250–53
Stalnaker, R. 326
Stevenson, C. L. 19, 255, 286, 320
Stratton-Lake, P. 163
Sturgeon, N. 150, 211, 227
supervenience 90, 327. See also, reduction
of goodness 201–4
of reasons 163–64
and thick terms 152
supervenience realism 375
Tetlock, P. E. 347
theoretical rationality
and aimless activity 69–71
compared to practical rationality 67–71
defined 71
and working assumptions 68
thick ethical terms
truth values of 316
Thomson, J. J. 90
Timmons, M. 205, 236, 276, 300, 303
Index
transcendental idealism 371, 384
trolley thought experiments 349, 351
truth
and communicative purposes 304
and contextual semantics 276
and correspondence 276–77, 280, 307,
382–83
minimalism 235–37, 300–1, 307
and non descriptivism 275
pragmatic conception of 302
and relativism 276
and saying versus showing 381
semantic conception of 302
T schema 236–37, 275, 300–1
truth aptness
and assertion types 310–12
and commitment properties 307
of value sentences 313–15, 383–84
Tversky, A. 345, 353
unanalyzability 79–80
uncertainty
concerning right action 133–48
concerning what is valuable 142–45
Unger, P. 352
utilitarianism
and deontological rules 389
ideal 370, 376, 388
valuing
versus having value 152–53
van Inwagen, P. 184–85
van Roojen, M. 355
Velleman, D. 126
virtue theory 35–36
water and H2O 328, 368, 374
welfare. See personal good
Wheatley, T. 352
Williams, B. 117 n. 22, 162, 321
Wingrave, O. 162–63
Wittgenstein, L. 169, 171–75, 178–88,
374, 377–80
Wright, C 300
Zagzebski, L. 187
Zimmerman, M. 93 n. 19, 251,
253
397